Upload
arka-dutta
View
45
Download
4
Tags:
Embed Size (px)
DESCRIPTION
DEVICE ENGINEERING
Citation preview
System Drivers TEXT
UPDATED
DesignTEXT
TEXT ADDED
UPDATED
TEXT UPDATED
TEXT
TEXT
TEXT
TEXT
TEXT
Test and Test EquipmentTEXT
UPDATED
UPDATED
UPDATED
UPDATED
TEXT
2008 FOCUS ITWG TABLES:System Drivers, Design, Test & Test Equipment, RF and AMS for Wireless, and Process Integration, Devices, & Structures (PIDS)
Link to file for Emerging Research Devices (ERD), Emerging Research Materials (ERM), Front-end Processes (FEP), Lithography, Interconnect, Factory Integration, and Assembly & Packaging
Link to file for Environment, Safety, & Health (ESH), Yield Enhancement, Metrology, Modeling & Simulation
Link to the 2008 Update Overview
Link to the 2007 ITRS chapters
Table SYSD1
Table SYSD2
Table SYSD3
Table SYSD4
Table DESN1
Table DESN X
Table DESN2
Table DESN3
Table DESN4
Table DESN5
Table DESN6
Table DESN7
Table DESN8
Table DESN9
Table DESN10
Table DESN11
Table DESN12
Table TST1
Table TST2
Table TST3
Table TST4
Table TST5
Table TST6
Table TST7
Table TST8
Table TST9
Table TST10
Table TST11
ORTCINDEX
2007 ITRS Chapters
TEXT
RF and AMS for WirelessUPDATED
UPDATED
UPDATED
UPDATED
UPDATED
Process Integration, Devices, and Structures (PIDS)TEXT
UPDATED
UPDATED
UPDATED
UPDATED
TEXT
TO ACCESS THESE TABLES LISTED BELOW, USE THE LINKS AT THE TOP OF THIS PAGE
Emerging Research DevicesTEXT Please refer to the ERD summary in the 2008 Update Overview
Emerging Research MaterialsTEXT Please refer to the ERM summary in the 2008 Update Overview
Front End ProcessesTEXT Table FEP1
UPDATED Table FEP2
UPDATED Table FEP3
UPDATED Table FEP4a
UPDATED Table FEP4b
UPDATED Table FEP5
UPDATED Table FEP6
UPDATED Table FEP7
UPDATED Table FEP8
Table TST12
Table TST13
Table TST14
Table TST15
Table RFAMS1
Table RFAMS2
Table RFAMS3
Table RFAMS4
Table RFAMS5
Table RFAMS6
Table RFAMS7
Table RFAMS8
Table PIDS1
Table PIDS2
Table PIDS3a and b
Table PIDS3c and d
Table PIDS4
Table PIDS5
Table PIDS6
Table PIDS7
Table FEP9
LithographyTEXT Table LITH1
TEXT Table LITH2
UPDATED Table LITH3
UPDATED Table LITH4AB
Table LITH4C
UPDATED Table LITH5AB
UPDATED Table LITH5CD
UPDATED Table LITH5EF
Table LITH6
InterconnectTEXT Table INTC1
UPDATED Table INTC2
UPDATED Table INTC3
Table INTC4
TEXT Table INTC5
UPDATED Table INTC6
Table INTC7
Factory IntegrationTEXT Table FAC1
TEXT Table FAC2
Table FAC3
UPDATED Table FAC4
Table FAC5
UPDATED Table FAC6
Table FAC7
TEXT Table FAC8
TEXT Table FAC9
Assembly and PackagingTEXT Table AP1
Table AP2
UPDATED Table AP3
UPDATED Table AP4
ADD Table AP4B
UPDATED Table AP5A
UPDATED Table AP5B
UPDATED Table AP5C
Table AP6
Table AP7
TEXT Table AP8
UPDATED Table AP9
TEXT UPDATED Table AP10
UPDATED Table AP11
Table AP12a and b
TEXT Table AP12c
TEXT Table AP13
TEXT Table AP14
TEXT UPDATED Table AP15
TEXT UPDATED Table AP16
TEXT Table AP17
TEXT Table AP18
TEXT UPDATED Table AP19
Table AP20
UPDATED Table AP21
TO ACCESS THESE TABLES LISTED BELOW, USE THE LINKS AT THE TOP OF THIS PAGE
CROSSCUT WORKING GROUP TABLES
Environment, Safety, and HealthTEXT Table ESH1
UPDATED Table ESH2a
UPDATED Table ESH2b
Table ESH3
Table ESH4a
Table ESH4b
UPDATED Table ESH5
Table ESH6
Yield EnhancementTEXT Table YE1
TEXT Table YE2
UPDATED Table YE3
UPDATED Table YE4
UPDATED Table YE5
UPDATED Table YE6
UPDATED Table YE7
UPDATED Table YE8
UPDATED Table YE9
MetrologyTEXT Table MET1
Table MET2
Table MET3
Table MET4a and b
Table MET4c and d
UPDATED Table MET5a
Table MET5b
Table MET6
Modeling and SimulationTEXT UPDATED Table MS1
UPDATED Table MS2a
UPDATED Table MS2b
UPDATED Table MS3
Major Product Market Segments and Impact on System DriversSOC Consumer Driver Design Productivity TrendsProjected Mixed-Signal Figures of Merit for Four Circuit TypesEmbedded Memory Requirements
Overall Design Technology ChallengesDescription of ImprovementSystem-Level Design RequirementsCorrespondence Between System-Level Design Requirements and SolutionsLogical/Circuit/Physical Design Technology RequirementsCorrespondence Between Logical/Circuit/Physical Requirements and SolutionsDesign Verification RequirementsCorrespondence Between Design Verification Requirements and SolutionsDesign for Test Technology RequirementsDesign for Manufacturability Technology RequirementsCorrespondence Between Design for Manufacturability Requirements and SolutionsNear-term Breakthroughs in Design Technology for AMSDesign Technology Improvements and Impact on Designer Productivity
Test and Test EquipmentSummary of Key Test Drivers, Challenges, and OpportunitiesMulti-site Test for Product SegmentsSystem on Chip Test RequirementsLogic Test RequirementsVector MultipliersMemory Test RequirementsMixed-signal Test RequirementsRF Test RequirementsBurn-in RequirementsTest Handler and Prober Difficult ChallengesProber Requirements
2008 FOCUS ITWG TABLES:System Drivers, Design, Test & Test Equipment, RF and AMS for Wireless, and Process Integration, Devices, & Structures
Link to file for Emerging Research Devices (ERD), Emerging Research Materials (ERM), Front-end Processes (FEP), Lithography, Interconnect, Factory Integration, and Assembly & Packaging
Link to file for Environment, Safety, & Health (ESH), Yield Enhancement, Metrology, Modeling & Simulation
Link to the 2008 Update Overview
Link to the 2007 ITRS chapters
Handler RequirementsProbing Difficult ChallengesWafer Probe Technology RequirementsTest Socket Technology Requirements
RF and AMS for WirelessRF and Analog Mixed-Signal CMOS Technology RequirementsRF and Analog Mixed-Signal Bipolar Technology RequirementsOn-Chip Passives Technology RequirementsEmbedded Passives Technology RequirementsPower Amplifier Technology RequirementsBase Station Devices Technology Requirements
RF and Analog Mixed-Signal RFMEMS
Process Integration, Devices, and Structures (PIDS)Process Integration Difficult ChallengesHigh-performance Logic Technology RequirementsLow Standby Power Technology RequirementsLow Operating Power Technology RequirementsDRAM Technology RequirementsNon-volatile Memory Technology RequirementsReliability Difficult ChallengesReliability Technology Requirements
TO ACCESS THESE TABLES LISTED BELOW, USE THE LINKS AT THE TOP OF THIS PAGE
Emerging Research DevicesPlease refer to the ERD summary in the 2008 Update Overview
Emerging Research MaterialsPlease refer to the ERM summary in the 2008 Update Overview
Front End Processes Difficult ChallengesStarting Materials Technology Requirements—Near and Long-term YearsFront End Surface Preparation Technology Requirements—Near and Long-term YearsThermal, Thin Film, Doping and Etching Technology Requirements—Near-term YearsThermal, Thin Film, Doping and Etching Technology Requirements—Long-term YearsDRAM Stacked Capacitor Technology Requirements—Near and Long-term YearsDRAM Trench Capacitor Technology Requirements—Near and Long-term YearsFLASH Non-volatile Memory Technology RequirementsPhase Change Memory (PCM) Technology Requirements—Near and Long-term Years
Millimeter Wave 10 GHz–100 GHz Technology Requirements
FeRAM Technology Requirements—Near and Long-term Years
Various Techniques for Achieving Desired CD Control and Overlay with Optical Projection LithographyLithography Difficult ChallengesLithography Technology Requirements—Near and Long-term YearsResist Requirements—Near and Long-term YearsResist SensitivitiesOptical Mask Requirements—Near and Long-term YearsEUVL Mask Requirements—Near and Long-term YearsImprint Template Requirements—Near and Long-term YearsMaskless Technology Requirements—Near and Long-term Years
Interconnect Difficult Challenges MPU Interconnect Technology Requirements—Near and Long-term YearsDRAM Interconnect Technology Requirements—Near and Long-term YearsInterconnect Surface Preparation Technology Requirements—Near and Long-term YearsOptions for Interconnects Beyond the Metal/Dielectric SystemHigh Density Through Silicon via Draft Specification
Factory Integration Difficult Challenges—Near and Long-term Years Key Focus Areas and Issues for FI Functional Areas Beyond 2007 Factory Operations Technology Requirements—Near and Long-term Years Production Equipment Technology Requirements—Near and Long-term Years Material Handling Systems Technology Requirements—Near and Long-term Years Factory Information and Control Systems Technology Requirements—Near and Long-term Years Facilities Technology Requirements—Near and Long-term Years Crosscut Issues Relating to Factory Integration List of Next Wafer Size Challenges
Assembly and PackagingAssembly and Packaging Difficult Challenges Single-chip Packages Technology Requirements—Near and Long-term YearsChip-to-package Substrate Technology Requirements—Near and Long-term YearsSubstrate to Board Pitch—Near and Long-term YearsWarpage at Peak TemperaturePackage Substrates—Near and Long-term YearsPolymer Package Substrate Design Parameters—Near and Long-term YearsCost Performance Glass-ceramic Substrates (high end FCBGA)Wafer Level Packaging—Near and Long-term Years
Minimum Density of Metallic SWCNTs Needed to Exceed Minimum Cu Wire Conductivity
Key Technical Parameters for Stacked Architectures Using TSVComparison of SoC and SiP ArchitecturePackage Level System IntegrationProcesses for SiPSystem in Package Requirements—Near and Long-term YearsThinned Silicon Wafer Thickness 200 mm/300 mm—Near and Long-term YearsChallenges and Potential Solutions in Thinning Si WafersSiP Failure ModesSome Common Optoelectronic Packages and Their ApplicationsOptical Communications and InterconnectOptoelectronic Packaging Challenges and Potential SolutionsMEMS Packaging MethodsMEMS Packaging ExamplesMaterials ChallengesPackage Substrate Physical Properties Automotive Operating Environment Specifications
TO ACCESS THESE TABLES LISTED BELOW, USE THE LINKS AT THE TOP OF THIS PAGE
CROSSCUT WORKING GROUP TABLES
Environment, Safety, and HealthESH Difficult ChallengesESH Intrinsic Requirements—Near-term YearsESH Intrinsic Requirements—Long-term YearsChemicals and Materials Management Technology RequirementsProcess and Equipment Management Technology Requirements—Near-term YearsProcess and Equipment Management Technology Requirements—Long-term YearsFacilities Energy and Water Optimization Technology RequirementsSustainability and Product Stewardship Technology Requirements
Definitions for the Different Interface PointsYield Enhancement Difficult ChallengesDefect Budget Technology Requirement AssumptionsYield Model and Defect Budget MPU Technology RequirementsYield Model and Defect Budget DRAM/Flash Technology RequirementsDefect Inspection on Patterned Wafer Technology Requirements
Defect Inspection on Unpatterned Wafers: Macro, and Bevel Inspection Technology Requirements
Defect Review and Automated Defect Classification Technology RequirementsTechnology Requirements for Wafer Environmental Contamination Control
Metrology Difficult ChallengesMetrology Technology RequirementsLithography Metrology (Wafer) Technology RequirementsLithography Metrology (Mask) Technology Requirements: OpticalLithography Metrology (Mask) Technology Requirements: EUVFront End Processes Metrology Technology Requirements—Near-term YearsFront End Processes Metrology Technology Requirements—Long-term YearsInterconnect Metrology Technology Requirements—Near and Long-term Years
Modeling and SimulationModeling and Simulation Difficult ChallengesModeling and Simulation Technology Requirements: Capabilities—Near-term YearsModeling and Simulation Technology Requirements: Capabilities—Long-term YearsModeling and Simulation Technology Requirements: Accuracy—Near-term Years
http://www.itrs.net/Links/2008ITRS/Update/2008Tables_FOCUS_B.xls
http://www.itrs.net/Links/2008ITRS/Update/2008Tables_CROSSCUT.xls
http://www.itrs.net/Links/2008ITRS/Update/2008_Update.pdf
http://www.itrs.net/Links/2007ITRS/Home2007.htm
Various Techniques for Achieving Desired CD Control and Overlay with Optical Projection Lithography
Overall Roadmap Technology Characteristics [ORTC]
UPDATED
UPDATED
UPDATED
UPDATED
UPDATED
UPDATED
UPDATED
UPDATED
UPDATED
UPDATED
UPDATED
UPDATED
UPDATED
Trends GraphicTable 1a&b
Table 1c&d
Table 1e&f
Table 1g&h
Table 1i&j
Table 2a&b
Table 3a&b
Table 4a&b
Table 4c&d
Table 5a&b
Table 6a&b
Table 7a&b
2007 ITRS Chapters
2008INDEX
Overall Roadmap Technology Characteristics [ORTC]
Product Generations and Chip Size Model Technology Trend Targets
DRAM and Flash Production Product Generations and Chip Size Model
DRAM Introduction Product Generations and Chip Size Model
MPU (High-volume Microprocessor) Cost-Performance Product Generations and Chip Size Model
High-Performance MPU and ASIC Product Generations and Chip Size Model
Lithographic-Field and Wafer Size Trends
Performance of Packaged Chips: Number of Pads and Pins
Performance and Package Chips: Pads, Cost
Performance and Package Chips: Frequency On-chip Wiring Levels
Electrical Defects [**]
Power Supply and Power Dissipation
Cost
2008 Update Trend Graphic, including ITRS 7/15 meetings Final Litho Printed Gate Length ProposalORTCINDEX
2007 ITRS Chapters
2008INDEX
2008 Update Trend Graphic, including ITRS 7/15 meetings Final Litho Printed Gate Length Proposal
Table 1a&bProduct Generations and Chip Size Model Technology Trend Targets
Year of Production 2007 2008 2009 2010 2011 2012 2013 2014
Flash ½ Pitch (nm) (un-contacted Poly)(f) 54 45 40 36 32 28 25 23
DRAM ½ Pitch (nm) (contacted) 65 57 50 45 40 36 32 28
DRAM ½ Pitch (nm) (contacted) 68 59 52 45 40 36 32 28
MPU/ASIC Metal 1 (M1) ½ Pitch (nm) 68 59 52 45 40 36 32 28
MPU Printed Gate Length (nm) 42 38 34 30 27 24 21 19
MPU Printed Gate Length (GLpr) (nm) †† 54 47 41 35 31 28 25 22
MPU Physical Gate Length (GLph) (nm) 25 23 20 18 16 14 13 11
MPU Physical Gate Length (GLph) (nm) 32 29 27 24 22 20 18 17
ASIC/Low Operating Power Printed Gate Length (nm) †† 54 48 42 38 34 30 27 24
ASIC/Low Operating Power Printed Gate Length (nm) †† 64 54 47 41 35 31 25 22
ASIC/Low Operating Power Physical Gate Length (nm) 32 28 25 23 20 18 16 14
ASIC/Low Operating Power Physical Gate Length (nm) 38 32 29 27 24 22 18 17
2015
20
25
25
25
17
20
10
15
21
20
13
15
2016 2017 2018 2019 2020 2021 2022
17.9 15.9 14.2 12.6 11.3 10.0 8.9
23 20 18 16 14 13 11
22.5 20.0 17.9 15.9 14.2 12.6 11.3
22.5 20.0 17.9 15.9 14.2 12.6 11.3
15 13 12 11 9.5 8.4 7.5
17.7 15.7 14.0 12.5 11.1 9.9 8.8
8.9 8.0 7.1 6.3 5.6 5.0 4.5
14.0 12.8 11.7 10.7 9.7 8.9 8.1
19 17 15 13 12 11 9.5
17.7 15.7 14.0 12.5 11.1 9.9 8.8
11 10 8.9 8.0 7.1 6.3 5.6
14.0 12.8 11.7 10.7 9.7 8.9 8.1
Table 1c&d DRAM and Flash Production Product Generations and Chip Size Model
Year of Production 2007 2008 2009 2010 2011 2012 2013 2014 2015
DRAM ½ Pitch (nm) (contacted) 68 59 52 45 40 36 32 28 25
MPU/ASIC Metal 1 (M1) ½ Pitch (nm) 68 59 52 45 40 36 32 28 25
MPU Physical Gate Length (nm) 32 29 27 24 22 20 18 17 15
DRAM Product Table
Cell area factor [a] 6 6 6 6 6 6 6 6 6
0.028 0.021 0.016 0.012 0.0096 0.0077 0.0061 0.0048 0.0038
Cell array area at production (% of chip size) § 56.08% 56.08% 56.08% 56.08% 56.08% 56.08% 56.08% 56.08% 56.08%
Generation at production § 2G 2G 2G 4G 4G 4G 8G 8G 8G
Functions per chip (Gbits) 2.15 2.15 2.15 4.29 4.29 4.29 8.59 8.59 8.59
107 81 61 93 74 59 93 74 59
Gbits/cm2 at production § 2.01 2.65 3.50 4.62 5.82 7.33 9.23 11.63 14.65
Flash Product Table
Flash ½ Pitch (nm) (un-contacted Poly)(f) 53.5 45.0 40.1 35.7 31.8 28.3 25.3 22.5 20.0
Cell area factor [a] 4 4 4 4 4 4 4 4 4
0.0115 0.0081 0.0064 0.0051 0.0040 0.0032 0.0026 0.0020 0.0016
Cell array area at production (% of chip size) § 68.35% 68.35% 68.35% 68.35% 68.35% 68.35% 68.35% 68.35% 68.35%
Generation at production § SLC 8G 8G 8G 16G 16G 16G 32G 32G 32G
Generation at production § MLC [2 bits/cell] 16G 16G 16G 32G 32G 32G 64G 64G 64G
Generation at production § MLC [4 bits/cell] 32G 32G 32G 64G 64G 64G 128G 128G 128G
Functions per chip (Gbits) SLC 8.59 8.59 8.59 17.18 17.18 17.18 34.36 34.36 34.36
Functions per chip (Gbits) MLC [2 bits/cell] 17.18 17.18 17.18 34.36 34.36 34.36 68.72 68.72 68.72
Functions per chip (Gbits) MLC [4 bits/cell] 34.36 34.36 34.36 68.72 68.72 68.72 137.44 137.44 137.44
143.96 101.80 80.80 128.26 101.80 80.80 128.26 101.80 80.80
143.96 101.80 80.80 128.26 101.80 80.80 128.26 101.80 80.80
5.97E+09 8.44E+09 1.06E+10 1.34E+10 1.69E+10 2.13E+10 2.68E+10 3.38E+10 4.25E+10
Bits/cm2 at production § MLC [2 bits/cell] 1.19E+10 1.69E+10 2.13E+10 2.68E+10 3.38E+10 4.25E+10 5.36E+10 6.75E+10 8.51E+10
Cell area [Ca = af2] (um2)
Chip size at production (mm2)§
Cell area [Ca = af2] (um2)
Chip size at production (mm2)§ SLC
Chip size at production (mm2)§ MLC [2 bits/cell & 4 bits/cell]
Bits/cm2 at production § SLC
2016 2017 2018 2019 2020 2021 2022
22.5 20.0 17.9 15.9 14.2 12.6 11.3
22.5 20.0 17.9 15.9 14.2 12.6 11.3
14.0 12.8 11.7 10.7 9.7 8.9 8.1
6 6 6 6 6 6 6
0.0030 0.0024 0.0019 0.0015 0.0012 0.00096 0.00076
56.08% 56.08% 56.08% 56.08% 56.08% 56.08% 56.08%
16G 16G 16G 32G 32G 32G 64G
17.18 17.18 17.18 34.36 34.36 34.36 68.72
93 74 59 93 74 59 93
18.46 23.26 29.31 36.93 46.52 58.61 73.85
17.9 15.9 14.2 12.6 11.3 10.0 8.9
4 4 4 4 4 4 4
0.0013 0.0010 0.00080 0.00064 0.00051 0.00040 0.00032
68.35% 68.35% 68.35% 68.35% 68.35% 68.35% 68.35%
64G 64G 64G 128G 128G 128G 256G
128G 128G 128G 256G 256G 256G 512G
256G 256G 256G 512G 512G 512G 1024G
68.72 68.72 68.72 137.44 137.44 137.44 274.88
137.44 137.44 137.44 274.88 274.88 274.88 549.76
274.88 274.88 274.88 549.76 549.76 549.76 1099.51
128.26 101.80 80.80 128.26 101.80 80.80 128.26
128.26 101.80 80.80 128.26 101.80 80.80 128.26
5.36E+10 6.75E+10 8.51E+10 1.07E+11 1.35E+11 1.70E+11 2.14E+11
1.07E+11 1.35E+11 1.70E+11 2.14E+11 2.70E+11 3.40E+11 4.29E+11
Table 1e&f DRAM Introduction Product Generations and Chip Size Model
Year of Production 2007 2008 2009 2010 2011 2012 2013 2014 2015
DRAM ½ Pitch (nm) (contacted) 68 59 52 45 40 36 32 28 25
MPU/ASIC Metal 1 (M1) ½ Pitch (nm) 68 59 52 45 40 36 32 28 25
MPU Physical Gate Length (nm) 32 29 27 24 22 20 18 17 15
Cell area factor [a] 6 6 6 6 6 6 6 6 6
0.028 0.021 0.016 0.012 0.0096 0.0077 0.0061 0.0048 0.0038
Cell array area at introduction (% of chip size) § 73.52% 73.76% 73.97% 74.16% 74.30% 74.47% 74.61% 74.70% 74.83%
Generation at introduction § 16G 16G 16G 32G 32G 32G 64G 64G 64G
Functions per chip (Gbits) 17.18 17.18 34.36 34.36 34.36 68.72 68.72 68.72 68.72
652 493 745 563 446 706 560 444 351
Cell area [Ca = af2] (um2)
Chip size at introduction (mm2) §
2016 2017 2018 2019 2020 2021 2022
22.5 20.0 17.9 15.9 14.2 12.6 11.3
22.5 20.0 17.9 15.9 14.2 12.6 11.3
14.0 12.8 11.7 10.7 9.7 8.9 8.1
6 6 6 6 6 6 6
0.0030 0.0024 0.0019 0.0015 0.0012 0.00096 0.00076
74.93% 75.00% 75.09% 75.18% 75.27% 75.36% 75.45%
128G 128G 128G 256G 256G 256G 512G
137.44 137.44 137.44 274.88 274.88 274.88 549.76
557 442 350 555 440 349 553
Table 1g&h MPU (High-volume Microprocessor) Cost-Performance Product Generations and Chip Size Model
Year of Production 2007 2008 2009 2010 2011 2012 2013 2014 2015
DRAM ½ Pitch (nm) (contacted) 68 59 52 45 40 36 32 28 25
MPU/ASIC Metal 1 (M1) ½ Pitch (nm) 68 59 52 45 40 36 32 28 25
MPU Physical Gate Length (nm) 32 29 27 24 22 20 18 17 15
SRAM Cell (6-transistor) Area factor ++ 97.5 100.7 104.1 107.8 106.7 105.7 104.8 104.1 103.4
Logic Gate (4-transistor) Area factor ++ 279 292 306 320 320 320 320 320 320
SRAM Cell (6-transistor) Area efficiency ++ 0.63 0.63 0.63 0.63 0.63 0.63 0.63 0.63 0.63
Logic Gate (4-transistor) Area efficiency ++ 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5
SRAM Cell (6-transistor) Area (um2)++ 0.45 0.35 0.28 0.22 0.17 0.13 0.11 0.084 0.066
SRAM Cell (6-transistor) Area w/overhead (um2)++ 0.73 0.57 0.45 0.35 0.27 0.22 0.17 0.13 0.11
Logic Gate (4-transistor) Area (um2) ++ 1.3 1.0 0.82 0.65 0.51 0.41 0.32 0.26 0.20
Logic Gate (4-transistor) Area w/overhead (um2) ++ 2.6 2.1 1.6 1.3 1.0 0.82 0.65 0.51 0.41
827 1,057 1,348 1,718 2,187 2,781 3,532 4,484 5,687
154 194 245 309 389 490 617 778 980
Generation at introduction * p10c p10c p13c p13c p13c p16c p16c p16c p19c
773 773 1546 1546 1546 3092 3092 3092 6184
280 222 353 280 222 353 280 222 353
276 348 438 552 696 876 1104 1391 1753
Generation at production * p07c p07c p07c p10c p10c p10c p13c p13c p13c
386 386 386 773 773 773 1546 1546 1546
140 111 88 140 111 88 140 111 88
276 348 438 552 696 876 1104 1391 1753
Transistor density SRAM (Mtransistors/cm2)
Transistor density logic (Mtransistors/cm2)
Functions per chip at introduction (million transistors [Mtransistors])
Chip size at introduction (mm2) ‡
Cost performance MPU (Mtransistors/cm2 at introduction) (including on-chip SRAM) ‡
Functions per chip at production (million transistors [Mtransistors])
Chip size at production (mm2) §§
Cost performance MPU (Mtransistors/cm2 at production, including on-chip SRAM) ‡
2016 2017 2018 2019 2020 2021 2022
22.5 20.0 17.9 15.9 14.2 12.6 11.3
22.5 20.0 17.9 15.9 14.2 12.6 11.3
14.0 12.8 11.7 10.7 9.7 8.9 8.1
102.8 102.2 101.7 101.3 100.9 100.5 100.1
320 320 320 320 320 320 320
0.63 0.63 0.63 0.63 0.63 0.63 0.63
0.5 0.5 0.5 0.5 0.5 0.5 0.5
0.052 0.041 0.032 0.026 0.020 0.016 0.01
0.083 0.066 0.052 0.041 0.032 0.026 0.020
0.16 0.13 0.10 0.081 0.064 0.051 0.040
0.32 0.26 0.20 0.16 0.13 0.10 0.08
7,208 9,130 11,558 14,625 18,497 23,394 29,588
1,235 1,555 1,960 2,469 3,111 3,920 4,938
p19c p19c p22c p22c p22c p25c p25c
6184 6184 12368 12368 12368 24736 24736
280 222 353 280 222 353 280
2209 2783 3506 4417 5565 7012 8834
p16c p16c p16c p19c p19c p19c p22c
3092 3092 3092 6184 6184 6184 12368
140 111 88 140 111 88 140
2209 2783 3506 4417 5565 7012 8834
Table 1i&j High-Performance MPU and ASIC Product Generations and Chip Size Model
Year of Production 2007 2008 2009 2010 2011 2012 2013 2014 2015
DRAM ½ Pitch (nm) (contacted) 68 59 52 45 40 36 32 28 25
MPU/ASIC Metal 1 (M1) ½ Pitch (nm) 68 59 52 45 40 36 32 28 25
MPU Physical Gate Length (nm) 32 29 27 24 22 20 18 17 15
Logic (Low-volume Microprocessor) High-performance ‡
Generation at Introduction p10h p10h p13h p13h p13h p16h p16h p16h p19h
Functions per chip at introduction (million transistors) 2212 2212 4424 4424 4424 8848 8848 8848 17696
620 492 391 620 492 391 620 492 391
Generation at production ** p07h p07h p07h p10h p10h p10h p13h p13h p13h
Functions per chip at production (million transistors) 1106 1106 1106 2212 2212 2212 4424 4424 4424
310 246 195 310 246 195 310 246 195
357 449 566 714 899 1133 1427 1798 2265
ASIC
357 449 566 714 899 1133 1427 1798 2265
858 858 858 858 858 858 858 858 858
3,061 3,857 4,859 6,122 7,713 9,718 12,244 15,427 19,436
Chip size at introduction (mm2)
Chip size at production (mm2) §§
High-performance MPU Mtransistors/cm2 at introduction and production (including on-chip SRAM) ‡
ASIC usable Mtransistors/cm2 (auto layout)
ASIC max chip size at production (mm2) (maximum lithographic field size)
ASIC maximum functions per chip at production (Mtransistors/chip) (fit in maximum lithographic field size)
2016 2017 2018 2019 2020 2021 2022
22.5 20.0 17.9 15.9 14.2 12.6 11.3
22.5 20.0 17.9 15.9 14.2 12.6 11.3
14.0 12.8 11.7 10.7 9.7 8.9 8.1
p19h p19h p22h p22h p22h p25h p25h
17696 17696 35391 35391 35391 70782 70782
620 492 391 620 492 391 620
p16h p16h p16h p19h p19h p19h p22h
8848 8848 8848 17696 17696 17696 35391
310 246 195 310 246 195 310
2854 3596 4531 5708 7192 9061 11416
2854 3596 4531 5708 7192 9061 11416
858 858 858 858 858 858 1716
24,488 30,853 38,873 48,977 61,707 77,746 195,906
Table 2a&b Lithographic-Field and Wafer Size Trends
Year of Production 2007 2008 2009 2010 2011 2012 2013 2014 2015
DRAM ½ Pitch (nm) (contacted) 68 59 52 45 40 36 32 28 25
Flash ½ Pitch (nm) (un-contacted Poly)(f) 54 45 40 36 32 28 25 22 20
MPU/ASIC Metal 1 (M1) ½ Pitch (nm) 68 59 52 45 40 36 32 28 25
MPU Physical Gate Length (nm) 32 29 27 24 22 20 18 17 15
Lithography Field Size
858 858 858 858 858 858 858 858 858
Maximum Lithography Field Size—length (mm) 33 33 33 33 33 33 33 33 33
Maximum Lithography Field Size—width (mm) 26 26 26 26 26 26 26 26 26
Bulk or epitaxial or SOI wafer 300 300 300 300 300 450 450 450 450
Maximum Lithography Field Size—area (mm2)
Maximum Substrate Diameter (mm)—High-volume Production (>20K parts wafer starts per month)
2016 2017 2018 2019 2020 2021 2022
22.5 20.0 17.9 15.9 14.2 12.6 11.3
18 16 14 13 11 10 9
22.5 20.0 17.9 15.9 14.2 12.6 11.3
14.0 12.8 11.7 10.7 9.7 8.9 8.1
858 858 858 858 858 858 858
33 33 33 33 33 33 33
26 26 26 26 26 26 26
450 450 450 450 450 450 450
Table 3a&bPerformance of Packaged Chips: Number of Pads and Pins
Year of Production 2007 2008 2009 2010 2011 2012 2013 2014 2015
DRAM ½ Pitch (nm) (contacted) 68 59 52 45 40 36 32 28 25
Flash ½ Pitch (nm) (un-contacted Poly)(f) 54 45 40 36 32 28 25 22 20
MPU/ASIC Metal 1 (M1) ½ Pitch (nm) 68 59 52 45 40 36 32 28 25
MPU Physical Gate Length (nm) 32 29 27 24 22 20 18 17 15
Total pads—MPU unchanged 3,072 3,072 3,072 3,072 3,072 3,072 3,072 3,072 3,072
Signal I/O—MPU (% of total pads) 33.3% 33.3% 33.3% 33.3% 33.3% 33.3% 33.3% 33.3% 33.3%
Power and ground pads—MPU (% of total pads) 66.7% 66.7% 66.7% 66.7% 66.7% 66.7% 66.7% 66.7% 66.7%
Total pads—ASIC High Performance unchanged 4,400 4,400 4,600 4,800 4,800 5,000 5,400 5,400 5,600
50.0% 50.0% 50.0% 50.0% 50.0% 50.0% 50.0% 50.0% 50.0%
50.0% 50.0% 50.0% 50.0% 50.0% 50.0% 50.0% 50.0% 50.0%
Number of Total Package Pins—Maximum [1]
Microprocessor/controller, cost-performance 600–2140 600–2400 660–2801 660–2783 720- 3061 720–3367 800–3704 800-4075 880–4482
Microprocessor/controller, high-performance 4000 4400 4620 4851 5094 5348 5616 5896 6191
ASIC (high-performance) 4000 4400 4620 4851 5094 5348 5616 5896 6191
Number of Chip I/Os (Number of Total Chip Pads)—Maximum
Signal I/O pads—ASIC high-performance (% of total pads)
Power and ground pads—ASIC high-performance (% of total pads)
2016 2017 2018 2019 2020 2021 2022
22.5 20.0 17.9 15.9 14.2 12.6 11.3
18 16 14 13 11 10 9
22.5 20.0 17.9 15.9 14.2 12.6 11.3
14.0 12.8 11.7 10.7 9.7 8.9 8.1
3,072 3,072 3,072 3,072 3,072 3,072 3,072
33.3% 33.3% 33.3% 33.3% 33.3% 33.3% 33.3%
66.7% 66.7% 66.7% 66.7% 66.7% 66.7% 66.7%
6,000 6,000 6,200 6,200 6,200 6,840 6,840
50.0% 50.0% 50.0% 50.0% 50.0% 50.0% 50.0%
50.0% 50.0% 50.0% 50.0% 50.0% 50.0% 50.0%
880–4930 960-5423 960–5966 1050-6562 1050 - 7218 1155-7940 1155-8337
6501 6826 7167 7525 7902 8297 8712
6501 6826 7167 7525 7902 8297 8712
Table 4a&b Performance and Package Chips: Pads, Cost
Year of Production 2007 2008 2009 2010 2011 2012 2013 2014
DRAM ½ Pitch (nm) (contacted) 68 59 52 45 40 36 32 28
Flash ½ Pitch (nm) (un-contacted Poly)(f) 54 45 40 36 32 28 25 22
MPU/ASIC Metal 1 (M1) ½ Pitch (nm) 68 59 52 45 40 36 32 28
MPU Physical Gate Length (nm) 32 29 27 24 22 20 18 17
Pad pitch—ball bond 30 30 25 25 25 20 20 20
25 25 20 20 20 20 20 20
130 130 130 130 120 110 110 100
55 50 45 45 45 40 40 40
60 60 60 55 55 50 45 45
Cost-Per-Pin
.69-1.19 .66-1.13 .63-1.70 .60-1.20 .57-.97 .54-.92 .51-.87 .48 - .83
.27-.50 .25-.48 .24-.46 .23-.44 .22-.42 .21-.40 .20-.38 .20-.36
Chip Pad Pitch (micron)
Pad pitch—Wedge bond
Pad Pitch—Area array flip-chip (cost-performance, high-performance)
Pad Pitch—2-row staggered-pitch (micron)
Pad Pitch—Three-tier-pitch pitch (micron)
Package cost (cents/pin) (Cost per Pin Minimum for Contract Assembly – Cost-performance) — minimum–maximum
Package cost (cents/pin) (Low-cost, hand-held and memory) — minimum–maximum
2015
25
20
25
15
20
20
100
40
45
.46 - .79
.20 -.34
2016 2017 2018 2019 2020 2021 2022
22.5 20.0 17.9 15.9 14.2 12.6 11.3
18 16 14 13 11 10 9
22.5 20.0 17.9 15.9 14.2 12.6 11.3
14.0 12.8 11.7 10.7 9.7 8.9 8.1
20 20 20 20 20 20 20
20 20 20 20 20 20 20
95 95 90 90 85 85 80
35 35 35 35 35 35 35
45 45 45 45 45 45 45
.44 - .75 .42 - .71 .39 - .68 .37 - .64 .35 - .61 .33-.58 0.32-0.55
.20-.32 .20-.30 .20-.29 .20-.27 .20-.26 .19-.25 .19-.25
Table 4c&d Performance and Package Chips: Frequency On-chip Wiring Levels
Year of Production 2007 2008 2009 2010 2011 2012 2013 2014 2015
DRAM ½ Pitch (nm) (contacted) 68 59 52 45 40 36 32 28 25
Flash ½ Pitch (nm) (un-contacted Poly)(f) 54 45 40 36 32 28 25 22 20
MPU/ASIC Metal 1 (M1) ½ Pitch (nm) 68 59 52 45 40 36 32 28 25
MPU Physical Gate Length (nm) 32 29 27 24 22 20 18 17 15
On-chip local clock [1] 4.700 5.063 5.454 5.875 6.329 6.817 7.344 7.911 8.522
11 12 12 12 12 12 13 13 13
Chip Frequency (MHz)
Maximum number wiring levels [3] [**]
[**] [Note ** : The Interconnect TWG has deleted their "optional levels" from table 80a&b, therefore the ORTC "Maximum number wiring levels - maximum" line is deleted; also the "Maximum number wiring levels - minimum" is now just "Maximum number of wiring levels."
2016 2017 2018 2019 2020 2021 2022
22.5 20.0 17.9 15.9 14.2 12.6 11.3
18 16 14 13 11 10 9
22.5 20.0 17.9 15.9 14.2 12.6 11.3
14.0 12.8 11.7 10.7 9.7 8.9 8.1
9.180 9.889 10.652 11.475 12.361 13.315 14.343
13 14 14 14 14 15 15
[**] [Note ** : The Interconnect TWG has deleted their "optional levels" from table 80a&b, therefore the ORTC "Maximum number wiring levels - maximum" line is deleted; also the "Maximum number wiring levels - minimum" is now just "Maximum number of wiring levels."
Year of Production 2007 2008 2009 2010 2011 2012 2013 2014 2015
DRAM ½ Pitch (nm) (contacted) 68 59 52 45 40 36 32 28 25
Flash ½ Pitch (nm) (un-contacted Poly)(f) 54 45 40 36 32 28 25 22 20
MPU/ASIC Metal 1 (M1) ½ Pitch (nm) 68 59 52 45 40 36 32 28 25
MPU Physical Gate Length (nm) 32 29 27 24 22 20 18 17 15
2503 2503 2503 2503 2503 2503 2503 2503 2503
2503 2503 2503 2503 2503 2503 2503 2503 2503
3517 2957 2957 2957 2957 2957 2957 2957 2957
2430 2430 2430 2430 2430 2430 2430 2430 2430
1395 1395 1395 1395 1395 1395 1395 1395 1395
# Mask Levels—MPU 33 35 35 35 35 35 37 37 37
# Mask Levels—DRAM 24 24 24 26 26 26 26 26 26
# Mask Levels—Flash [to be added in 2009] ?? ?? ?? ?? ?? ?? ?? ?? ??
Table 5a&b Electrical Defects [**]
Flash Random Defect D 0 at production chip size and 89.5% yield (faults/m 2) §
Flash Random Defect D 0 at production chip size and 89.5% yield (faults/m 2) §
DRAM Random Defect D0 at production chip size and 89.5% yield (faults/m2) §
DRAM Random Defect D0 at production chip size and 89.5% yield (faults/m2) §
MPU Random Defect D0 at production chip size and 83% yield (faults/ m2) §§
2016 2017 2018 2019 2020 2021 2022
22.5 20.0 17.9 15.9 14.2 12.6 11.3
18 16 14 13 11 10 9
22.5 20.0 17.9 15.9 14.2 12.6 11.3
14.0 12.8 11.7 10.7 9.7 8.9 8.1
2503 2503 2503 2503 2503 2503 2503
2503 2503 2503 2503 2503 2503 2503
2957 2957 2957 2957 2957 2957 2957
2430 2430 2430 2430 2430 2430 2430
1395 1395 1395 1395 1395 1395 1395
37 39 39 39 39 39 39
26 26 26 26 26 26 26
?? ?? ?? ?? ?? ?? ??
Year of Production 2007 2008 2009 2010 2011 2012 2013 2014
DRAM ½ Pitch (nm) (contacted) 68 59 52 45 40 36 32 28
Flash ½ Pitch (nm) (un-contacted Poly)(f) 54 45 40 36 32 28 25 22
MPU/ASIC Metal 1 (M1) ½ Pitch (nm) 68 59 52 45 40 36 32 28
MPU Physical Gate Length (nm) 32 29 27 24 22 20 18 17
Power Supply Voltage (V)
Vdd (high-performance) 1.1 1.0 1.0 1.0 0.95 0.90 0.90 0.90
Vdd (high-performance) 1.1 1.1 1.1 1.1 1.0 1.0 1.0 1.0
0.80 0.80 0.80 0.70 0.70 0.70 0.60 0.60
0.90 0.80 0.80 0.80 0.77 0.70 0.70 0.65
Allowable Maximum Power [1]
High-performance with heatsink (W) 102 146 143 146 161 158 149 152
310 310 310 310 310 310 310 310
0.33 0.47 0.46 0.47 0.52 0.51 0.48 0.49
Cost-performance (W) 102 146 143 146 161 158 149 152
140 140 140 140 140 140 140 140
0,57 0.86 0.9 0.96 1.13 1.11 1.1 1.17
Battery (W)—(low-cost/hand-held) 3 3 3 3 3 3 3 3
Table 6a&b Power Supply and Power Dissipation
Vdd (Low Operating Power, high Vdd transistors)[WAS]
Vdd (Low Operating Power, high Vdd transistors)
Maximum Affordable Chip Size Target for High-performance MPU Maximum Power Calculation
Maximum High-performance MPU Maximum Power Density for Maximum Power Calculation
Maximum Affordable Chip Size Target for Cost-performance MPU Maximum Power Calculation
Maximum Cost-performance MPU Maximum Power Density for Maximum Power Calculation
2015
25
20
25
15
0.80
1.0
0.60
0.60
143
310
0.46
143
140
1.19
3
2016 2017 2018 2019 2020 2021 2022
22.5 20.0 17.9 15.9 14.2 12.6 11.3
18 16 14 13 11 10 9
22.5 20.0 17.9 15.9 14.2 12.6 11.3
14.0 12.8 11.7 10.7 9.7 8.9 8.1
0.80 0.70 0.70 0.70 0.65 0.65 0.65
0.90 0.90 0.90 0.90 0.80 0.80 0.80
0.50 0.50 0.50 0.50 0.50 0.45 0.45
0.60 0.60 0.60 0.57 0.50 0.50 0.50
130 130 136 133 130 130 130
310 310 310 310 310 310 310
0.42 0.42 0.44 0.43 0.42 0.42 0.42
130 130 136 133 130 130 130
140 140 140 140 140 140 140
1.07 1.12 1.19 1.27 1.24 1.63 1.73
3 3 3 3 3 3 3
Table 7a&b Cost
Year of Production 2007 2008 2009 2010 2011 2012 2013 2014 2015
DRAM ½ Pitch (nm) (contacted) 68 59 52 45 40 36 32 28 25
Flash ½ Pitch (nm) (un-contacted Poly)(f) 54 45 40 36 32 28 25 22 20
MPU/ASIC Metal 1 (M1) ½ Pitch (nm) 68 59 52 45 40 36 32 28 25
MPU Physical Gate Length (nm) 32 29 27 24 22 20 18 17 15
Affordable Cost per Function ++
DRAM cost/bit at (packaged microcents) at samples/introduction 2.6 1.9 1.3 0.9 0.7 0.5 0.3 0.2 0.2
DRAM cost/bit at (packaged microcents) at production § 0.96 0.68 0.48 0.34 0.24 0.17 0.12 0.08 0.06
22.0 15.6 11.0 7.8 5.5 3.9 2.8 1.9 1.4
13.3 9.4 6.7 4.7 3.3 2.4 1.7 1.2 0.83
12.2 8.6 6.1 4.3 3.0 2.2 1.5 1.1 0.76
Cost-performance MPU (microcents/transistor)(including on-chip SRAM) at introduction §§
Cost-performance MPU (microcents/transistor)(including on-chip SRAM) at production §§
High-performance MPU (microcents/transistor)(including on-chip SRAM) at production §§
2016 2017 2018 2019 2020 2021 2022
22.5 20.0 17.9 15.9 14.2 12.6 11.3
18 16 14 13 11 10 9
22.5 20.0 17.9 15.9 14.2 12.6 11.3
14.0 12.8 11.7 10.7 9.7 8.9 8.1
0.1 0.1 0.1 0.0 0.0 0.0 0.0
0.04 0.03 0.02 0.01 0.01 0.01 0.01
0.97 0.69 0.49 0.34 0.24 0.17 0.12
0.59 0.42 0.29 0.21 0.15 0.10 0.07
0.54 0.38 0.27 0.19 0.13 0.10 0.07
Table SYSD1 Major Product Market Segments and Impact on System DriversMarket Drivers SOC Analog/MS MPU
I. Portable/consumer
1. Size/weight ratio: peak in 2004 Low power paramount Migrating on-chip for voice processing, A/D sampling, and even for some RF transceiver function
Specialized cores to optimize processing per microwatt
2. Battery life: peak in 2004
3. Function: 2×/2 years Need SOC integration (DSP, MPU, I/O cores, etc.)
4. Time-to-market: ASAP
II. Medical
1. Cost: slight downward pressure High-end products only. Reprogrammability possible. Mainly ASSP, especially for patient data storage and telemedicine; more SOC for high-end digital with cores for imaging, real-time diagnostics, etc.
Absolutely necessary for physical measurement and response but may not be integrated on chip
Often used for programmability especially when real-time performance is not important
(~1/2 every 5 years)
2. Time-to-market: >12 months Recent advances in multicore processors have made programmability and real-time performance possible
3. Function: new on-chip functions
4. Form factor often not important
5. Durability/safety
6. Conservation/ ecology
III. Networking and communications
1. Bandwidth: 4/3–4 years Large gate counts Migrating on-chip for MUX/DEMUX circuitry
MPU cores, FPGA cores and some specialized functions
2. Reliability High reliability
3. Time-to-market: ASAP More reprogrammability to accommodate custom functions
MEMS for optical switching.
4. Power: W/m3 of system
IV. Defense
1. Cost: not prime concern Most case leverage existing processors but some requirements may drive towards single-chip designs with programmability
Absolutely necessary for physical measurement and response but may not be integrated on chip
Often used for programmability especially when real-time performance is not important
2. Time-to-market: >12 months
3. Function: mostly on SW to ride Recent advances in multicore processors have made programmability and real-time performance possible
technology curve
4. Form factor may be important
5. High durability/safety
V. Office
1. Speed: 2/2 years Large gate counts; high speed Minimal on-chip analog; simple A/D and D/A
MPU cores and some specialized functions
2. Memory density: 2/2 years
3. Power: flat to decreasing, Drives demand for digital functionality Video i/f for automated camera monitoring, video conferencing
Increased industry partnerships on common designs to reduce development costs (requires data sharing and reuse across multiple design systems)
driven by cost and W/m3
4. Form factor: shrinking size Primarily SOC integration of custom off-the-shelf MPU and I/O cores
Integrated high-speed A/D, D/A for monitoring, instrumentation, and range-speed-position resolution
5. Reliability
VI. Automotive
1. Functionality Mainly entertainment systems Cost-driven on-chip A/D and D/A for sensor and actuators
2. Ruggedness (external
environment, noise) Mainly ASSP, but increasing SOC for high end using standard HW platforms with RTOS kernel, embedded software
Signal processing shifting to DSP for voice, visual
3. Reliability and safety
4. Cost Physical measurement (“communicating sensors” for proximity, motion, positioning); MEMS for sensors
A/D—analog to digital ASSP—application-specific standard product D/A—digital to analog DEMUX—demultiplexer
DSP—digital signal processing FPGA—field programmable gate array i/f—interface I/O—input/output HW—hardware
MEMS—microelectromechanical systems MUX—multiplexer RTOS—real-time operating system
ORTCINDEX
2007 ITRS Chapters
2008INDEX
Table SYSD1 Major Product Market Segments and Impact on System DriversMarket Drivers SOC Analog/MS MPU
I. Portable/consumer
1. Size/weight ratio: peak in 2004 Low power paramount Migrating on-chip for voice processing, A/D sampling, and even for some RF transceiver function
Specialized cores to optimize processing per microwatt
2. Battery life: peak in 2004
3. Function: 2×/2 years Need SOC integration (DSP, MPU, I/O cores, etc.)
4. Time-to-market: ASAP
II. Medical
1. Cost: slight downward pressure High-end products only. Reprogrammability possible. Mainly ASSP, especially for patient data storage and telemedicine; more SOC for high-end digital with cores for imaging, real-time diagnostics, etc.
Absolutely necessary for physical measurement and response but may not be integrated on chip
Often used for programmability especially when real-time performance is not important
(~1/2 every 5 years)
2. Time-to-market: >12 months Recent advances in multicore processors have made programmability and real-time performance possible
3. Function: new on-chip functions
4. Form factor often not important
5. Durability/safety
6. Conservation/ ecology
III. Networking and communications
1. Bandwidth: 4/3–4 years Large gate counts Migrating on-chip for MUX/DEMUX circuitry
MPU cores, FPGA cores and some specialized functions
2. Reliability High reliability
3. Time-to-market: ASAP More reprogrammability to accommodate custom functions
MEMS for optical switching.
4. Power: W/m3 of system
IV. Defense
1. Cost: not prime concern Most case leverage existing processors but some requirements may drive towards single-chip designs with programmability
Absolutely necessary for physical measurement and response but may not be integrated on chip
Often used for programmability especially when real-time performance is not important
2. Time-to-market: >12 months
3. Function: mostly on SW to ride Recent advances in multicore processors have made programmability and real-time performance possible
technology curve
4. Form factor may be important
5. High durability/safety
V. Office
1. Speed: 2/2 years Large gate counts; high speed Minimal on-chip analog; simple A/D and D/A
MPU cores and some specialized functions
2. Memory density: 2/2 years
3. Power: flat to decreasing, Drives demand for digital functionality Video i/f for automated camera monitoring, video conferencing
Increased industry partnerships on common designs to reduce development costs (requires data sharing and reuse across multiple design systems)
driven by cost and W/m3
4. Form factor: shrinking size Primarily SOC integration of custom off-the-shelf MPU and I/O cores
Integrated high-speed A/D, D/A for monitoring, instrumentation, and range-speed-position resolution
5. Reliability
VI. Automotive
1. Functionality Mainly entertainment systems Cost-driven on-chip A/D and D/A for sensor and actuators
2. Ruggedness (external
environment, noise) Mainly ASSP, but increasing SOC for high end using standard HW platforms with RTOS kernel, embedded software
Signal processing shifting to DSP for voice, visual
3. Reliability and safety
4. Cost Physical measurement (“communicating sensors” for proximity, motion, positioning); MEMS for sensors
A/D—analog to digital ASSP—application-specific standard product D/A—digital to analog DEMUX—demultiplexer
DSP—digital signal processing FPGA—field programmable gate array i/f—interface I/O—input/output HW—hardware
MEMS—microelectromechanical systems MUX—multiplexer RTOS—real-time operating system
Table SYSD1 Major Product Market Segments and Impact on System DriversMarket Drivers SOC Analog/MS MPU
I. Portable/consumer
1. Size/weight ratio: peak in 2004 Low power paramount Migrating on-chip for voice processing, A/D sampling, and even for some RF transceiver function
Specialized cores to optimize processing per microwatt
2. Battery life: peak in 2004
3. Function: 2×/2 years Need SOC integration (DSP, MPU, I/O cores, etc.)
4. Time-to-market: ASAP
II. Medical
1. Cost: slight downward pressure High-end products only. Reprogrammability possible. Mainly ASSP, especially for patient data storage and telemedicine; more SOC for high-end digital with cores for imaging, real-time diagnostics, etc.
Absolutely necessary for physical measurement and response but may not be integrated on chip
Often used for programmability especially when real-time performance is not important
(~1/2 every 5 years)
2. Time-to-market: >12 months Recent advances in multicore processors have made programmability and real-time performance possible
3. Function: new on-chip functions
4. Form factor often not important
5. Durability/safety
6. Conservation/ ecology
III. Networking and communications
1. Bandwidth: 4/3–4 years Large gate counts Migrating on-chip for MUX/DEMUX circuitry
MPU cores, FPGA cores and some specialized functions
2. Reliability High reliability
3. Time-to-market: ASAP More reprogrammability to accommodate custom functions
MEMS for optical switching.
4. Power: W/m3 of system
IV. Defense
1. Cost: not prime concern Most case leverage existing processors but some requirements may drive towards single-chip designs with programmability
Absolutely necessary for physical measurement and response but may not be integrated on chip
Often used for programmability especially when real-time performance is not important
2. Time-to-market: >12 months
3. Function: mostly on SW to ride Recent advances in multicore processors have made programmability and real-time performance possible
technology curve
4. Form factor may be important
5. High durability/safety
V. Office
1. Speed: 2/2 years Large gate counts; high speed Minimal on-chip analog; simple A/D and D/A
MPU cores and some specialized functions
2. Memory density: 2/2 years
3. Power: flat to decreasing, Drives demand for digital functionality Video i/f for automated camera monitoring, video conferencing
Increased industry partnerships on common designs to reduce development costs (requires data sharing and reuse across multiple design systems)
driven by cost and W/m3
4. Form factor: shrinking size Primarily SOC integration of custom off-the-shelf MPU and I/O cores
Integrated high-speed A/D, D/A for monitoring, instrumentation, and range-speed-position resolution
5. Reliability
VI. Automotive
1. Functionality Mainly entertainment systems Cost-driven on-chip A/D and D/A for sensor and actuators
2. Ruggedness (external
environment, noise) Mainly ASSP, but increasing SOC for high end using standard HW platforms with RTOS kernel, embedded software
Signal processing shifting to DSP for voice, visual
3. Reliability and safety
4. Cost Physical measurement (“communicating sensors” for proximity, motion, positioning); MEMS for sensors
A/D—analog to digital ASSP—application-specific standard product D/A—digital to analog DEMUX—demultiplexer
DSP—digital signal processing FPGA—field programmable gate array i/f—interface I/O—input/output HW—hardware
MEMS—microelectromechanical systems MUX—multiplexer RTOS—real-time operating system
Table SYSD1 Major Product Market Segments and Impact on System DriversMarket Drivers SOC Analog/MS MPU
I. Portable/consumer
1. Size/weight ratio: peak in 2004 Low power paramount Migrating on-chip for voice processing, A/D sampling, and even for some RF transceiver function
Specialized cores to optimize processing per microwatt
2. Battery life: peak in 2004
3. Function: 2×/2 years Need SOC integration (DSP, MPU, I/O cores, etc.)
4. Time-to-market: ASAP
II. Medical
1. Cost: slight downward pressure High-end products only. Reprogrammability possible. Mainly ASSP, especially for patient data storage and telemedicine; more SOC for high-end digital with cores for imaging, real-time diagnostics, etc.
Absolutely necessary for physical measurement and response but may not be integrated on chip
Often used for programmability especially when real-time performance is not important
(~1/2 every 5 years)
2. Time-to-market: >12 months Recent advances in multicore processors have made programmability and real-time performance possible
3. Function: new on-chip functions
4. Form factor often not important
5. Durability/safety
6. Conservation/ ecology
III. Networking and communications
1. Bandwidth: 4/3–4 years Large gate counts Migrating on-chip for MUX/DEMUX circuitry
MPU cores, FPGA cores and some specialized functions
2. Reliability High reliability
3. Time-to-market: ASAP More reprogrammability to accommodate custom functions
MEMS for optical switching.
4. Power: W/m3 of system
IV. Defense
1. Cost: not prime concern Most case leverage existing processors but some requirements may drive towards single-chip designs with programmability
Absolutely necessary for physical measurement and response but may not be integrated on chip
Often used for programmability especially when real-time performance is not important
2. Time-to-market: >12 months
3. Function: mostly on SW to ride Recent advances in multicore processors have made programmability and real-time performance possible
technology curve
4. Form factor may be important
5. High durability/safety
V. Office
1. Speed: 2/2 years Large gate counts; high speed Minimal on-chip analog; simple A/D and D/A
MPU cores and some specialized functions
2. Memory density: 2/2 years
3. Power: flat to decreasing, Drives demand for digital functionality Video i/f for automated camera monitoring, video conferencing
Increased industry partnerships on common designs to reduce development costs (requires data sharing and reuse across multiple design systems)
driven by cost and W/m3
4. Form factor: shrinking size Primarily SOC integration of custom off-the-shelf MPU and I/O cores
Integrated high-speed A/D, D/A for monitoring, instrumentation, and range-speed-position resolution
5. Reliability
VI. Automotive
1. Functionality Mainly entertainment systems Cost-driven on-chip A/D and D/A for sensor and actuators
2. Ruggedness (external
environment, noise) Mainly ASSP, but increasing SOC for high end using standard HW platforms with RTOS kernel, embedded software
Signal processing shifting to DSP for voice, visual
3. Reliability and safety
4. Cost Physical measurement (“communicating sensors” for proximity, motion, positioning); MEMS for sensors
A/D—analog to digital ASSP—application-specific standard product D/A—digital to analog DEMUX—demultiplexer
DSP—digital signal processing FPGA—field programmable gate array i/f—interface I/O—input/output HW—hardware
MEMS—microelectromechanical systems MUX—multiplexer RTOS—real-time operating system
Table SYSD2 SOC Consumer Driver Design Productivity Trends
2007 2008 2009
WAS Trend: SoC total Logic Size 1 1.29 1.62
IS (Normalized to 2007) 1 1.24 1.52
Requirement: % of reused design 38% 42% 46%
WAS Requirement: Productivity for new designs 1 1.25 1.54
IS (Normalized to 2007) 1 1.21 1.45
WAS 2 2.51 3.08
IS 2 2.42 2.89
Requirement: Productivity for reused designs (Normalized to productivity for new designs at 2008)
ORTCINDEX
2007 ITRS Chapters
2008INDEX
SOC Consumer Driver Design Productivity Trends
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022
2.12 2.64 3.24 4.07 5.29 6.62 8.52 10.33 12.76 16.17 21.14 24.6 34.4
1.93 2.41 2.98 3.74 4.74 5.96 7.54 9.41 11.78 14.87 18.91 23.35 30.26
50% 54% 58% 62% 66% 70% 74% 78% 82% 86% 90% 92% 94%
1.96 2.38 2.84 3.47 4.37 5.31 6.63 7.78 9.3 11.38 14.36 16.4 22.51
1.79 2.17 2.61 3.19 3.92 4.78 5.87 7.09 8.58 10.46 12.84 15.57 19.8
3.92 4.76 5.68 6.94 8.74 10.62 13.26 15.56 18.59 22.75 28.71 32.79 45.02
3.58 4.35 5.22 6.37 7.84 9.56 11.73 14.18 17.16 20.93 25.68 31.13 39.6
Table SYSD3 Projected Mixed-Signal Figures of Merit for Four Circuit Types
Year of Production 2007 2010 2013 2016 2019
RF-CMOS ½ Pitch 65 45 32 22 18
20 28–32 40–50 50–80 60–90
1.4 1.5–1.7 1.8–2 2–2.4 2.4–3
15 30 50–70 90–100 110–130
1.5 2–2.5 2.5–3.5 3–5 4–6
[1] Lower bound is for "high-resolution/thermal noise limited" A/D converters; upper bound is for "low-resolution/speed limited" A/D converters.
FoMLNA (GHz)
FoMVCO (1/J) ×1022
FoMPA (W×GHz2) ×104
FoMADC (GHz/W) ×103 [1]
ORTCINDEX
2007 ITRS Chapters
2008INDEX
2022 Driver
13
70-100 Refer to the RF and AMS
2.7–3.5
120-140
6–10
[1] Lower bound is for "high-resolution/thermal noise limited" A/D converters; upper bound is for "low-resolution/speed limited" A/D converters.
Technologies for Wireless chapter
Table SYSD4a Embedded Memory Requirements
Year of Production 2007 2008 2009
DRAM ½ Pitch (nm) 65 55 50
65 65 65
140F² 140F² 140F²
Array efficiency [2] 0.7 0.7 0.7
Process overhead versus standard CMOS – #added mask layers 2 2 2
1.1 1/1.1 1/1.1
Static power dissipation (mW/Cell) [5] 3E-4/1E-6 3E-4/1E-6 3E-4/1E-6
Dynamic power consumption per cell (mW/MHz) [6] 4.5E-7/7E-7 4E-7/6.5E-7 4E-7/6E-7
Read cycle time (ns) [7] 0.3/1.5 0.3/1.5 0.3/1.5
Write cycle time (ns) [7] 0.3/1.5 0.3/1.5 0.3/1.5
Percentage of MBU on total SER 16% 16% 16%
Soft error rate (FIT/Mb) [8] 1150 1150 1150
Embedded Non-Volatile Memory (code/data), DRAM ½ pitch (nm) 90 90 90
Array efficiency – NOR FLOTOX/ NAND FLOTOX [10] 0.6/0.8 0.6/0.8 0.6/0.8
Process overhead versus standard CMOS – #added mask layers [3] 6–8 6–8 6–8
Read operating voltage (V) 2V 2V 2V
12V/15V 12V/15V 12V/15V
Static power dissipation (mW/cell) [5] 1.00E-06 1.00E-06 1.00E-06
Dynamic power consumption per cell (mW/MHz) [6] 6.00E-09 6.00E-09 6.00E-09
Read cycle time (ns) – NOR FLOTOX / NAND FLOTOX [7] Oct-50 Oct-50 Oct-50
Program time per cell (µs) – NOR FLOTOX / NAND FLOTOX [12] 1.0/1000.0 1.0/1000.0 1.0/1000.0
Erase time per cell (ms) – NOR FLOTOX / NAND FLOTOX [12] 10.0/0.1 10.0/0.1 10.0/0.1
Data retention requirement (years) [12] 10 10 10
Endurance requirement [12] 100000 100000 100000
Embedded DRAM, ½ pitch (nm) 90 90 65
12–30 12–30 12–30
Array efficiency [2] 0.6 0.6 0.6
Process overhead versus standard CMOS – #added mask layers [3] 3–5 3–5 3–5
Read operating voltage (V) 2 2 1.8
Static power dissipation (mW/Cell) [5] 1.00E-11 1.00E-11 1.00E-11
Dynamic power consumption per cell (mW/MHz) [6] 1.00E-07 1.00E-07 1.00E-07
DRAM retention time (ms) [12] 64 64 64
Read/Write cycle time (ns) [7] 0.7 0.7 0.5
Soft error rate (FIT/Mb) [8] 60 60 60
FIT—failures in time FLOTOX—floating gate tunnel oxide MBU—multiple bit upsets NAND—“not AND” logic operation
NOR—“not OR” logic operation
Definitions of Terms for Tables SYSD4a and SYSD4b:
CMOS SRAM High-performance, low standby power (HP/LSTP) DRAM ½ pitch (nm), Feature Size – F
6T bit cell size (F2) [1]
Operating voltage – Vdd (V) [4]
Cell size (F2) – NOR FLOTOX / NAND FLOTOX [9] 10F2/5F2 10F2/5F2 10F2/5F2
Write (program/erase) on chip maximum voltage (V) – NOR/NAND [11]
1T1C bit cell size (F2) [13]
ORTCINDEX
2007 ITRS Chapters
2008INDEX
[1] Size of the standard 6T CMOS SRAM cell as a function of minimum feature size.
[2] Typical array efficiency defined as (core area / memory instance area).
[8] A FIT is a failure in 1 billion hours. This data is presented as FIT per megabit.
[11] Maximum voltage required for operation, typically used in WRITE operation. Data refer to the NVM device requirements table in the PIDS chapter.
[13] Size of the standard cell for embedded trench DRAM cell. Data refers to the DRAM requirements table in the PIDS chapter.
[3] Typical number of extra masks needed over standard CMOS logic process in equivalent technology. This is typically zero; however for some high-performance or highly reliable (noise immune) SRAMs special process options are sometimes applied like additional high—V th pMOS cell transistors and using higher Vdd for better noise margin or zero-Vth access transistors for fast read-out.
[4] Nominal operating voltage refers to the HP and LSTP devices in the logic device requirements table in the PIDS chapter.
[5] Static power dissipation per cell in standby mode. This is measured at I_standby × Vdd. (off-current and Vdd are taken from the HP and LSTP devices in the logic device requirements table in the PIDS Chapter.
[6] This parameter is a strong function of array architecture. However, a parameter for technology can be determined per cell level. Assume full V dd swing on the Wordline (WL) and 0.8 Vdd swing on the Bitline (BL). Determine the WL capacitance per cell (CWL) and BL capacitance per cell (CBL). Then: dynamic power consumption per MHz per cell = V dd × CWL (per cell) × (Vdd) + Vdd × CBL (per cell) × (Vdd) ×106.
[7] Read cycle time is the typical time it takes to complete a READ operation from an address. Write cycle time is the typical time it takes to complete a WRITE operation to an address. Both cycle times depend on memory size and architecture.
[9] Size of the standard 1T FLOTOX cell/size of the standard 2T select gate (SG) cell/size of the standard NAND cell. Cell size is somewhat enhanced compared to stand-alone NVM due to integration issues.
[10] Array efficiency of the standard stacked gate NOR architecture/standard split gate NOR architecture/standard NAND architecture. Data refer to the NVM device requirements table in the PIDS chapter.
[12] Program time per cell is typically the time needed to program data to a cell. Erase time per cell is typically the time needed to erase a cell. Data retention requirement is the duration for which the data must remain non-volatile even under worst-case conditions. Endurance requirement specifies the number of times the cell can be programmed and erased.
2010 2013 2016 2019 2022
45 35 25 18 13
45 35 25 18 13
140F²
0.7 0.7 0.7 0.7 0.7
2 2 2 2 2
1 0.9/1 0.8/0.9 0.7/0.8 0.7/0.8
5E-4/1.2E-6 1E-3/1.5E-6 2E-3/2E-6 3E-3/2.5E-6 5E-3/3E-6
3E-7/5E-7 2.5E-7/4.5E-7 2E-7/4E-7 1.5E-7/3E-7 1E-7/2E-7
0.2/1.2 0.15/0.8 0.1/0.5 0.07/0.3 0.07/0.3
0.2/1.2 0.15/0.8 0.1/0.5 0.07/0.3 0.07/0.3
32% 64% 100% 100% 100%
1200 1250 1300 1350 1400
65 45 35 25 18
0.6/0.8 0.6/0.8 0.6/0.8 0.6/0.8 0.6/0.8
6–8 6–8 6–8 6–8 6–8
1.8V 1.5V 1.3V 1.2V 1.1V
12V/15V 12V/15V 12V/15V 12V/15V 12V/15V
1.00E-06 1.00E-06 1.00E-06 1.00E-06 1.00E-06
6.00E-09 4.00E-09 3.50E-09 3.00E-09 3.00E-09
Jul-35 25-May 3.5/18 2.5/12 10-Feb
1.0/1000.0 1.0/1000.0 1.0/1000.0 1.0/1000.0 1.0/1000.0
10.0/0.1 10.0/0.1 10.0/0.1 10.0/0.1 10.0/0.1
10 10 10 10 10
100000 100000 100000 100000 100000
65 45 35 25 25
12–30 12–30 12–30 12–30 12–30
0.6 0.6 0.6 0.6 0.6
3–5 3–6 3–6 3–6 3–6
1.7 1.6 1.5 1.5 1.5
1.00E-11 1.00E-11 1.00E-11 1.00E-11 1.00E-11
1.50E-07 1.60E-07 1.70E-07 1.70E-07 1.70E-07
64 64 64 64 64
0.4 0.3 0.25 0.2 0.2
60 60 60 60 60
140F2 140F2 140F2 140F2
10F2/5F2 10F2/5F2 10F2/5F2 10F2/5F2 10F2/5F2
[1] Size of the standard 6T CMOS SRAM cell as a function of minimum feature size.
[2] Typical array efficiency defined as (core area / memory instance area).
[8] A FIT is a failure in 1 billion hours. This data is presented as FIT per megabit.
[11] Maximum voltage required for operation, typically used in WRITE operation. Data refer to the NVM device requirements table in the PIDS chapter.
[13] Size of the standard cell for embedded trench DRAM cell. Data refers to the DRAM requirements table in the PIDS chapter.
[3] Typical number of extra masks needed over standard CMOS logic process in equivalent technology. This is typically zero; however for some high-performance or highly reliable (noise pMOS cell transistors and using higher Vdd for better noise margin or zero-Vth access transistors for
PIDS chapter.
are taken from the HP and LSTP devices in the logic device requirements table
[6] This parameter is a strong function of array architecture. However, a parameter for technology can be determined per cell level. Assume full V dd swing on the Wordline (WL) and 0.8 Vdd swing on the Bitline (BL). Determine the WL capacitance per cell (CWL) and BL capacitance per cell (CBL). Then: dynamic power consumption per MHz per cell = V dd × CWL (per cell) ×
[7] Read cycle time is the typical time it takes to complete a READ operation from an address. Write cycle time is the typical time it takes to complete a WRITE operation to an address. Both
[9] Size of the standard 1T FLOTOX cell/size of the standard 2T select gate (SG) cell/size of the standard NAND cell. Cell size is somewhat enhanced compared to stand-alone NVM due to
[10] Array efficiency of the standard stacked gate NOR architecture/standard split gate NOR architecture/standard NAND architecture. Data refer to the NVM device requirements table in
[12] Program time per cell is typically the time needed to program data to a cell. Erase time per cell is typically the time needed to erase a cell. Data retention requirement is the duration for which the data must remain non-volatile even under worst-case conditions. Endurance requirement specifies the number of times the cell can be programmed and erased.
Table DESN1 Overall Design Technology ChallengesChallenges ≥ 32 nm Summary of Issues
Design productivity System level: high level of abstraction (HW/SW) functionality spec, platform based design, multi-processor programmability, system integration, AMS co-design and automation
Verification: executable specification, ESL formal verification, intelligent test bench, coverage-based verification
Logic/circuit/layout: analog circuit synthesis, multi-objective optimization
Power consumption Logic/circuit/layout: dynamic and static (leakage), system and circuit, power optimization
Manufacturability Performance/power variability, device parameter variability, lithography limitations impact on design, mask cost, quality of (process) models
ATE interface test (multi-Gb/s), mixed-signal test, delay BIST, test-volume-reducing DFT
Reliability Logic/circuit/layout: MTTF-aware design, BISR, soft-error correction
Interference Logic/circuit/layout: signal integrity analysis, EMI analysis, thermal analysis
Challenges <32 nm Summary of Issues
Design productivity Complete formal verification of designs, complete verification code reuse, complete deployment of functional coverage
Tools specific for SOI and non-static logic, and emerging devices
Cost-driven design flow
Heterogeneous component integration (optical, mechanical, chemical, bio, etc.)
Power consumption SOI power management
Manufacturability Uncontrollable threshold voltage variability
Advanced analog/mixed signal DFT (digital, structural, radio), “statistical” and yield-improvement DFT
Thermal BIST, system-level BIST
Reliability Autonomic computing, robust design, SW reliability
Interference Interactions between heterogeneous components (optical, mechanical, chemical, bio, etc.)
ATE—automatic test equipment BISR—built-in self repair BIST—built-in self test DFT—design for test
EMI—electromagnetic interference ESL—Electronic System-Level HW/SW—hardware/software MTTF—mean time to failureSOI—silicon on insulator
ORTCINDEX
2007 ITRS Chapters
2008INDEX
Table DESN1 Overall Design Technology ChallengesChallenges ≥ 32 nm Summary of Issues
Design productivity System level: high level of abstraction (HW/SW) functionality spec, platform based design, multi-processor programmability, system integration, AMS co-design and automation
Verification: executable specification, ESL formal verification, intelligent test bench, coverage-based verification
Logic/circuit/layout: analog circuit synthesis, multi-objective optimization
Power consumption Logic/circuit/layout: dynamic and static (leakage), system and circuit, power optimization
Manufacturability Performance/power variability, device parameter variability, lithography limitations impact on design, mask cost, quality of (process) models
ATE interface test (multi-Gb/s), mixed-signal test, delay BIST, test-volume-reducing DFT
Reliability Logic/circuit/layout: MTTF-aware design, BISR, soft-error correction
Interference Logic/circuit/layout: signal integrity analysis, EMI analysis, thermal analysis
Challenges <32 nm Summary of Issues
Design productivity Complete formal verification of designs, complete verification code reuse, complete deployment of functional coverage
Tools specific for SOI and non-static logic, and emerging devices
Cost-driven design flow
Heterogeneous component integration (optical, mechanical, chemical, bio, etc.)
Power consumption SOI power management
Manufacturability Uncontrollable threshold voltage variability
Advanced analog/mixed signal DFT (digital, structural, radio), “statistical” and yield-improvement DFT
Thermal BIST, system-level BIST
Reliability Autonomic computing, robust design, SW reliability
Interference Interactions between heterogeneous components (optical, mechanical, chemical, bio, etc.)
ATE—automatic test equipment BISR—built-in self repair BIST—built-in self test DFT—design for test
EMI—electromagnetic interference ESL—Electronic System-Level HW/SW—hardware/software MTTF—mean time to failureSOI—silicon on insulator
Description of Improvement
In-House place and Route The transfer of the IC Place and Route function from the semiconductor to the design team.
IC Implementation Tool Suite Tightly integrated tool set that goes from RTL Synthesis to GDS II through IC Place and Route.
RTL Functional verification tool suite
Tightly integrated RTL Verification tool suite including all simulators and formal tools needed to complete the verification process.
Transactional Modeling The development of standard SystemC models at the Transaction Level of abstraction.
Very Large Block reuse Blocks that exceed 1M gates
Homogeneous Parallel Processing
Many identical processor cores which allows for performance, power efficiency and high reuse. (SMP)
Concurrent Software Infrastructure
A set of tools that allow concurrent software development and debug.
Heterogeneous parallel processing
Parallel Processing using different application specific processors for each of the separate functions in the system.
Transactional Memory A concurrency control mechanism analogous to database transactions for controlling access to shared memory in concurrent computing . It functions as an alternative to lock-based synchroniza tion.
System Design Automation True System Level design including Electronic hardware and software, Mechanical, Bio, Opto, Chemical and fluids domains.
Executable Specification A design flow that has no manual processes from the specification to completed system and that can be completely validated at each step.
ReBlocks from 75.000 – 1M gates
InA verification tool (cockpit) that takes in an ES-Level description and partitions it into verification blocks, then executes the proper verification tools on the blocks;
TaThe presence in the Design Team of at least one senior engineer who had experience in all phases of the design process.
ReBlocks from 2,500 – 74,999 gates
ORTCINDEX
2007 ITRS Chapters
2008INDEX
Description of Improvement
In-House place and Route The transfer of the IC Place and Route function from the semiconductor to the design team.
IC Implementation Tool Suite Tightly integrated tool set that goes from RTL Synthesis to GDS II through IC Place and Route.
RTL Functional verification tool suite
Tightly integrated RTL Verification tool suite including all simulators and formal tools needed to complete the verification process.
Transactional Modeling The development of standard SystemC models at the Transaction Level of abstraction.
Very Large Block reuse Blocks that exceed 1M gates
Homogeneous Parallel Processing
Many identical processor cores which allows for performance, power efficiency and high reuse. (SMP)
Concurrent Software Infrastructure
A set of tools that allow concurrent software development and debug.
Heterogeneous parallel processing
Parallel Processing using different application specific processors for each of the separate functions in the system.
Transactional Memory A concurrency control mechanism analogous to database transactions for controlling access to shared memory in concurrent computing . It functions as an alternative to lock-based synchroniza tion.
System Design Automation True System Level design including Electronic hardware and software, Mechanical, Bio, Opto, Chemical and fluids domains.
Executable Specification A design flow that has no manual processes from the specification to completed system and that can be completely validated at each step.
ReBlocks from 75.000 – 1M gates
InA verification tool (cockpit) that takes in an ES-Level description and partitions it into verification blocks, then executes the proper verification tools on the blocks;
TaThe presence in the Design Team of at least one senior engineer who had experience in all phases of the design process.
ReBlocks from 2,500 – 74,999 gates
Table DESN2 System-Level Design Requirements
Year of Production 2007 2008 2009
Design Reuse
Design block reuse [1] % of all logic 35% 36%
Platform Based Design
Available platforms [2] Normalized to 100% in the start year [3] 87% 83% 75%
Platforms supported [4] % of platforms fully supported by tools [5]10% 35%
High Level Synthesis
60% 66%
Reconfigurability
SOC reconfigurability [7] % of SOC functionality that is reconfigurable28% 30%
Analog/Mixed Signal
Analog automation [8] % versus digital automation [9] 17% 17% 24%
58% 62%
ADD Embedded Software
ADD Hardware productivity 100% 200.0% 200.0%
ADD Software productivity 100% 100.0% 100.0%
ADD100% 200% 400%
38%
25%
Accuracy of high level estimates (performance, area, power, costs) [6] % versus measurements 63%
28%
Modeling methodology, description languages, simulation environments [10] % vs. digital methodology 60%
SW productivity needed according to forecast from figure DESN3: (2x every year), normalized to 2007
ORTCINDEX
2007 ITRS Chapters
2008INDEX
2010 2011 2012 2013 2014 2015 2016 2017 2018
40% 41% 42% 44% 48% 49% 51% 52%
70% 60% 55% 52% 48% 45% 43% 40% 37%
50% 57% 64% 75% 85% 90% 92% 94%
70% 73% 76% 80% 86% 90% 92% 94%
35% 38% 40% 42% 48% 50% 53% 56%
24% 27% 30% 32% 38% 40% 43% 46%
65% 67% 70% 73% 78% 80% 83% 86%
275.0% 275.0% 275.0% 275.0% 550% 550% 1100% 1100% 1760%
100.0% 100.0% 100% 100% 100% 100% 100% 100% 100%
800% 1600% 3200% 6400% 12800% 25600% 51200% 102400% 204800%
46%
80%
83%
45%
35%
76%
2019 2020 2021 2022
54% 55% 57% 58%
35% 32% 29% 27%
95% 97% 99% 100%
95% 97% 99% 100%
60% 62% 65% 68%
50% 52% 55% 58%
90% 92% 95% 98%
1760% 5280% 5280% 5280%
100% 100% 100% 100%
409600% 819200% 1638400% 3276800%
Table DESN3 Correspondence Between System-Level Design Requirements and SolutionsRequirement Solution Explanation of the Correspondence
Design block reuse System-level component reuse The larger and more complex the components that can be reused, the greater the expected overall design reuse
On-chip network design methods Standardized communication structures and interfaces support reuse: IPs with standardized interfaces can be easily integrated and exchanged, and communication structures reused
Available platforms Multi-fabric implementation planning (AMS, RF, MEMS, …)
Enables integration of different fabrics on same die or in same package (SIP); hence, enables reduced number of platforms
Platforms supported Automated interface synthesis Automated interface synthesis is one building block to an integrated synthesis flow for whole platforms.
Automated HW-SW co-design and verification
Required for integrated, platform-based system development
Accuracy of high level estimates Improved system-level power estimation techniques
System-level power estimation needs to match progress in high-level area and performance estimation
Chip-package co-design methods Packaging effects, e.g., on timing, must be accounted for in higher-level estimations
SOC reconfigurability On-chip network design methods To provide flexible, reconfigurable communication structures
Analog automation Multi-fabric implementation planning (AMS, RF, MEMS, …)
Multi-fabric implementation planning for AMS and RF components are a building block to analog automation
Modeling methodology, description languages, simulation environments
Mixed-Signal/RF verification As in digital design, verification is an increasingly critical and time-consuming activity in the design flow
ADD HW offers multi-cores that have to be exploited by SW
Parallel Processing Due to thermal and power limitations further performance increases have to be realized with multi-core systems.
ADD Reduce SW verification effort Intelligent Testbench SW simulation, formal verification and automated testbenches for SW will reduce the verification effort for embedded software and enhance quality.
ADD Productivity increase required for SW since SW cost >> 50% of total system cost
Concurrent Software Infrastructure A set of tools that allow concurrent software development and debug
ADD Increase SW execution performance Heterogeneous Parallel Processing Parallel Processing using different application specific processors for each of the separate functions in the system
ADD SW Productivity increase required Transactional Memory A concurrency control mechanism analogous to database transactions for controlling access to shared memory in concurrent computing. It functions as an alternative to lock-based synchronization
ADD Productivity increase required for HW/SW co-design
System Design Automation (SDA) True System Level design including Electronic hardware and software, Mechanical, Bio, Opto, Chemical and fluids domains
ADD Reduce verification effort Executable Specification Specifications written in a formal language allow automated verification process starting early in the design process and at high abstraction levels without the need to code several new verification models. This enables a design flow that has no manual processes from the specification to completed system and that can be completely validated at each step.
ORTCINDEX
2007 ITRS Chapters
2008INDEX
Table DESN3 Correspondence Between System-Level Design Requirements and SolutionsRequirement Solution Explanation of the Correspondence
Design block reuse System-level component reuse The larger and more complex the components that can be reused, the greater the expected overall design reuse
On-chip network design methods Standardized communication structures and interfaces support reuse: IPs with standardized interfaces can be easily integrated and exchanged, and communication structures reused
Available platforms Multi-fabric implementation planning (AMS, RF, MEMS, …)
Enables integration of different fabrics on same die or in same package (SIP); hence, enables reduced number of platforms
Platforms supported Automated interface synthesis Automated interface synthesis is one building block to an integrated synthesis flow for whole platforms.
Automated HW-SW co-design and verification
Required for integrated, platform-based system development
Accuracy of high level estimates Improved system-level power estimation techniques
System-level power estimation needs to match progress in high-level area and performance estimation
Chip-package co-design methods Packaging effects, e.g., on timing, must be accounted for in higher-level estimations
SOC reconfigurability On-chip network design methods To provide flexible, reconfigurable communication structures
Analog automation Multi-fabric implementation planning (AMS, RF, MEMS, …)
Multi-fabric implementation planning for AMS and RF components are a building block to analog automation
Modeling methodology, description languages, simulation environments
Mixed-Signal/RF verification As in digital design, verification is an increasingly critical and time-consuming activity in the design flow
ADD HW offers multi-cores that have to be exploited by SW
Parallel Processing Due to thermal and power limitations further performance increases have to be realized with multi-core systems.
ADD Reduce SW verification effort Intelligent Testbench SW simulation, formal verification and automated testbenches for SW will reduce the verification effort for embedded software and enhance quality.
ADD Productivity increase required for SW since SW cost >> 50% of total system cost
Concurrent Software Infrastructure A set of tools that allow concurrent software development and debug
ADD Increase SW execution performance Heterogeneous Parallel Processing Parallel Processing using different application specific processors for each of the separate functions in the system
ADD SW Productivity increase required Transactional Memory A concurrency control mechanism analogous to database transactions for controlling access to shared memory in concurrent computing. It functions as an alternative to lock-based synchronization
ADD Productivity increase required for HW/SW co-design
System Design Automation (SDA) True System Level design including Electronic hardware and software, Mechanical, Bio, Opto, Chemical and fluids domains
ADD Reduce verification effort Executable Specification Specifications written in a formal language allow automated verification process starting early in the design process and at high abstraction levels without the need to code several new verification models. This enables a design flow that has no manual processes from the specification to completed system and that can be completely validated at each step.
Table DESN4 Logical/Circuit/Physical Design Technology Requirements
Year of Production 2007 2008 2009 2010
7% 11% 15% 17%
Parameter uncertainty:%-effect (on signoff delay) 6% 8% 10% 11%
4 5 6 6
Circuit families: # of families in a single design 3 3 4 4
Synthesized analog content: % of total design analog content 15% 16% 17% 18%
Full-chip leakage (normalized to full-chip leakage power dissipation in 2007) 1 1.5 2 2.5
Asynchronous global signaling:% of a design driven by handshake clocking
Simultaneous analysis objectives:# of objectives during optimization
ORTCINDEX
2007 ITRS Chapters
2008INDEX
2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022
19% 20% 22% 23% 25% 30% 30% 30% 35% 40% 43% 45%
11% 12% 14% 15% 18% 20% 20% 20% 22% 25% 26% 28%
6 6 7 8 8 8 8 8 8 8 8 8
4 4 4 4 4 4 4 4 4 4 4 4
19% 20% 23% 25% 28% 30% 35% 40% 45% 50% 55% 60%
2.75 3 3.5 4 6 8 8 8 8 8 8 8
Table DESN5 Correspondence Between Logical/Circuit/Physical Requirements and SolutionsRequirement Solution Explanation of the Correspondence
Asynchronous global signaling Automated handshake logic/circuit tools Departure from fully synchronous design paradigm needed for power reduction, latency insensitivity, variation-tolerance
% of a design (SOC)
Parameter uncertainty Synthesis and timing analysis accounting for variability Tools that account for process uncertainty, and resulting parametric uncertainty, will reduce guardbanding and increase chip yields
%-effect (on signoff delay)
Simultaneous analysis objectives Circuit/layout enhancement accounting for variability Optimizations which consider parametric uncertainty
Simultaneous analysis objectives Power management analysis and logic insertion SOI SOC tools Requires budgeting of area/power/timing constraints
Simultaneous analysis objectives Cost-driven implementation flow Cost is an engineering parameter that affects turnaround times. Silicon cost no longer dominant; test and manufacturing costs increase emphasis on adaptive, self-repairing circuits
Circuit families Non-static logic implementation Non-static implementations help improving different chip parameters
# of circuit families in a single design
Synthesized analog content Analog synthesis (circuit/layout) Allows for larger portions of a chip to be analog
Full-chip leakage Macro block/chip leakage analysis Enables accurate leakage predictions
ORTCINDEX
2007 ITRS Chapters
2008INDEX
Table DESN5 Correspondence Between Logical/Circuit/Physical Requirements and SolutionsRequirement Solution Explanation of the Correspondence
Asynchronous global signaling Automated handshake logic/circuit tools Departure from fully synchronous design paradigm needed for power reduction, latency insensitivity, variation-tolerance
% of a design (SOC)
Parameter uncertainty Synthesis and timing analysis accounting for variability Tools that account for process uncertainty, and resulting parametric uncertainty, will reduce guardbanding and increase chip yields
%-effect (on signoff delay)
Simultaneous analysis objectives Circuit/layout enhancement accounting for variability Optimizations which consider parametric uncertainty
Simultaneous analysis objectives Power management analysis and logic insertion SOI SOC tools Requires budgeting of area/power/timing constraints
Simultaneous analysis objectives Cost-driven implementation flow Cost is an engineering parameter that affects turnaround times. Silicon cost no longer dominant; test and manufacturing costs increase emphasis on adaptive, self-repairing circuits
Circuit families Non-static logic implementation Non-static implementations help improving different chip parameters
# of circuit families in a single design
Synthesized analog content Analog synthesis (circuit/layout) Allows for larger portions of a chip to be analog
Full-chip leakage Macro block/chip leakage analysis Enables accurate leakage predictions
Table DESN6 Design Verification Requirements
Year of Production 2007 2008 2009 2010
Productivity
7.9 10.3 13.5 17.6
Methodology
4.7 7.1 9.4 11.8
11.6 13.1 14.7 16.3
Portion of the design specification formalized for verifiability (%) 13.8 17.5 21.3 25
Bugs
8 7 7 7
62 68 74 79
Reuse
73.9 70.8 67.8 64.7
15.5 18.3 21.1 23.8
Functional coverage
46.5 49.7 52.9 56.2
1294 1608 1922 2235
Design size verifiable by 1 engineer-year (in millions of transistors - based on an SOC design and a 10-person engineering team) [1]
Design errors exposed using formal or semi-formal verification (%, versus simulation)
Effort spent on system-level verification: software, hardware and electrical effects (%)
Escape rate: bugs found after first tapeout (per each 100K lines of design code)
Bugs found after system integration until tapeout (per each 100K lines of design code)
Portion of the verification infrastructure (e.g., test beds, coverage, checkers) which is newly developed (versus reused components and acquired IP) (%) [2]
Portion of the verification infrastructure which is acquired from third parties (i.e., verification IP) (%) [2]
Portion of design for which verification quality is evaluated through functional coverage (%)
Coverage goal density (expressed as number of coverage goals for each million transistors of the design) [3]
ORTCINDEX
2007 ITRS Chapters
2008INDEX
2011 2012 2013 2014 2015 2016 2017 2018 2019
23.1 30.3 39.8 52.3 69.6 91.8 121 159.7 210.9
14.1 16.5 18.8 21.2 23.5 25.9 28.2 30.6 32.9
17.8 19.4 20.9 22.5 24.1 25.6 27.2 28.8 30.3
28.8 32.5 36.3 40 43.8 47.5 51.3 55 58.8
6 6 6 6 5 5 5 4 4
85 91 97 103 109 115 121 126 132
61.6 58.6 55.5 52.5 49.4 46.4 43.3 40.2 37.2
26.6 29.4 32.1 34.9 37.6 40.4 43.2 45.9 48.7
59.4 62.6 65.9 69.1 72.4 75.6 78.8 82.1 85.3
2549 2863 3176 3490 3804 4118 4431 4745 5059
2020 2021 2022
278.6 368.5 487.6
35.3 37.6 40
31.9 33.4 35
62.5 66.25 70
4 3 3
138 144 150
34.1 31.1 28
51.5 54.2 57
88.5 91.8 95
5373 5686 6000
Table DESN7 Correspondence Between Design Verification Requirements and SolutionsRequirement Solution Explanation of the Correspondence
Productivity of verification tasks Verification methodology centered on verification IPs and reuse
Verification IPs and reuse reduce the amount of new verification development required in a project
Hierarchical hardware verification Structured methodologies improve design team productivity
Reusable methodologies for functional coverage development
Functional coverage is time-consuming, and specific for each distinct design; development of reusability techniques is critical to boosting productivity
Concurrent verification of hardware and software components during development
Advancing the verification of hardware in parallel with that of software components can significantly shorten time-to-market of a product, in contrast to methodologies that begin software verification only after the first hardware prototype
Formal and semi-formal verification centered methodology
Hierarchical hardware verification methodology Enables the decomposition of the system into smaller blocks which are suitable for formal verification
Design development and structure taking into account verifiability
Design for verifiability organizes a design so as to simplify verification; additional verification-specific hardware structures further simplify design-time verification tasks
Methodologies for system-level verification
Verification methodology centered on verification IPs and reuse
Verification IP components enable an early start on system-level verification
Integrated verification of hardware and embedded software and their interface
Directly provides solutions for effective system-level verification
Portion of design specification formalized for verifiability
Design specification formalized for verifiability Formal languages and methodologies to support the formal specification of a design
Escape rate after tapeout Design structure taking into account verifiability Development of hardware structures (checker-like) which can be used to detect and correct a system entering an escaped erroneous configuration after customer shipment
System integration bug rate Analog and mixed-signal verification Limits the bug rate due to analog effects
Simulation and verification solutions for the detection and correction of soft failures and manufacturing faults
Manufacturing faults occurring in post-silicon are detected at system-level integration; techniques to detect and correct electrical and transient defects reduce the effort required to expose and correct these problems.
Hierarchical verification methodology Supports management of complexity through decomposition
Functional coverage Reusable methodologies for functional coverage development
Reusable functional coverage solutions leverage the coverage development effort and boost quality of results
ORTCINDEX
2007 ITRS Chapters
2008INDEX
Table DESN7 Correspondence Between Design Verification Requirements and SolutionsRequirement Solution Explanation of the Correspondence
Productivity of verification tasks Verification methodology centered on verification IPs and reuse
Verification IPs and reuse reduce the amount of new verification development required in a project
Hierarchical hardware verification Structured methodologies improve design team productivity
Reusable methodologies for functional coverage development
Functional coverage is time-consuming, and specific for each distinct design; development of reusability techniques is critical to boosting productivity
Concurrent verification of hardware and software components during development
Advancing the verification of hardware in parallel with that of software components can significantly shorten time-to-market of a product, in contrast to methodologies that begin software verification only after the first hardware prototype
Formal and semi-formal verification centered methodology
Hierarchical hardware verification methodology Enables the decomposition of the system into smaller blocks which are suitable for formal verification
Design development and structure taking into account verifiability
Design for verifiability organizes a design so as to simplify verification; additional verification-specific hardware structures further simplify design-time verification tasks
Methodologies for system-level verification
Verification methodology centered on verification IPs and reuse
Verification IP components enable an early start on system-level verification
Integrated verification of hardware and embedded software and their interface
Directly provides solutions for effective system-level verification
Portion of design specification formalized for verifiability
Design specification formalized for verifiability Formal languages and methodologies to support the formal specification of a design
Escape rate after tapeout Design structure taking into account verifiability Development of hardware structures (checker-like) which can be used to detect and correct a system entering an escaped erroneous configuration after customer shipment
System integration bug rate Analog and mixed-signal verification Limits the bug rate due to analog effects
Simulation and verification solutions for the detection and correction of soft failures and manufacturing faults
Manufacturing faults occurring in post-silicon are detected at system-level integration; techniques to detect and correct electrical and transient defects reduce the effort required to expose and correct these problems.
Hierarchical verification methodology Supports management of complexity through decomposition
Functional coverage Reusable methodologies for functional coverage development
Reusable functional coverage solutions leverage the coverage development effort and boost quality of results
Table DESN7 Correspondence Between Design Verification Requirements and SolutionsRequirement Solution Explanation of the Correspondence
Productivity of verification tasks Verification methodology centered on verification IPs and reuse
Verification IPs and reuse reduce the amount of new verification development required in a project
Hierarchical hardware verification Structured methodologies improve design team productivity
Reusable methodologies for functional coverage development
Functional coverage is time-consuming, and specific for each distinct design; development of reusability techniques is critical to boosting productivity
Concurrent verification of hardware and software components during development
Advancing the verification of hardware in parallel with that of software components can significantly shorten time-to-market of a product, in contrast to methodologies that begin software verification only after the first hardware prototype
Formal and semi-formal verification centered methodology
Hierarchical hardware verification methodology Enables the decomposition of the system into smaller blocks which are suitable for formal verification
Design development and structure taking into account verifiability
Design for verifiability organizes a design so as to simplify verification; additional verification-specific hardware structures further simplify design-time verification tasks
Methodologies for system-level verification
Verification methodology centered on verification IPs and reuse
Verification IP components enable an early start on system-level verification
Integrated verification of hardware and embedded software and their interface
Directly provides solutions for effective system-level verification
Portion of design specification formalized for verifiability
Design specification formalized for verifiability Formal languages and methodologies to support the formal specification of a design
Escape rate after tapeout Design structure taking into account verifiability Development of hardware structures (checker-like) which can be used to detect and correct a system entering an escaped erroneous configuration after customer shipment
System integration bug rate Analog and mixed-signal verification Limits the bug rate due to analog effects
Simulation and verification solutions for the detection and correction of soft failures and manufacturing faults
Manufacturing faults occurring in post-silicon are detected at system-level integration; techniques to detect and correct electrical and transient defects reduce the effort required to expose and correct these problems.
Hierarchical verification methodology Supports management of complexity through decomposition
Functional coverage Reusable methodologies for functional coverage development
Reusable functional coverage solutions leverage the coverage development effort and boost quality of results
Table DESN7 Correspondence Between Design Verification Requirements and SolutionsRequirement Solution Explanation of the Correspondence
Productivity of verification tasks Verification methodology centered on verification IPs and reuse
Verification IPs and reuse reduce the amount of new verification development required in a project
Hierarchical hardware verification Structured methodologies improve design team productivity
Reusable methodologies for functional coverage development
Functional coverage is time-consuming, and specific for each distinct design; development of reusability techniques is critical to boosting productivity
Concurrent verification of hardware and software components during development
Advancing the verification of hardware in parallel with that of software components can significantly shorten time-to-market of a product, in contrast to methodologies that begin software verification only after the first hardware prototype
Formal and semi-formal verification centered methodology
Hierarchical hardware verification methodology Enables the decomposition of the system into smaller blocks which are suitable for formal verification
Design development and structure taking into account verifiability
Design for verifiability organizes a design so as to simplify verification; additional verification-specific hardware structures further simplify design-time verification tasks
Methodologies for system-level verification
Verification methodology centered on verification IPs and reuse
Verification IP components enable an early start on system-level verification
Integrated verification of hardware and embedded software and their interface
Directly provides solutions for effective system-level verification
Portion of design specification formalized for verifiability
Design specification formalized for verifiability Formal languages and methodologies to support the formal specification of a design
Escape rate after tapeout Design structure taking into account verifiability Development of hardware structures (checker-like) which can be used to detect and correct a system entering an escaped erroneous configuration after customer shipment
System integration bug rate Analog and mixed-signal verification Limits the bug rate due to analog effects
Simulation and verification solutions for the detection and correction of soft failures and manufacturing faults
Manufacturing faults occurring in post-silicon are detected at system-level integration; techniques to detect and correct electrical and transient defects reduce the effort required to expose and correct these problems.
Hierarchical verification methodology Supports management of complexity through decomposition
Functional coverage Reusable methodologies for functional coverage development
Reusable functional coverage solutions leverage the coverage development effort and boost quality of results
Table DESN8 Design for Test Technology Requirements
Year of Production 2007 2008 2009 2010
System Driver: Analog/Mixed-signal/RF
All-digital DFT for analog/mixed-signal/RF circuits and systems. 40 45 50 55
% digital circuits in DFT implementations
40 45 50 55
25 30 35 40
System Drivers: MPU/PE/DSP
DFT coverage of digital blocks or subsystems. % blocks with DFT 70 70 70 75
DFT for delay test of critical paths. % paths covered 55 55 60 60
DFT for fault tolerance in logic blocks. 40 40 45 45
% blocks with fault tolerance
System Drivers: Memories
DFT for yield improvement. 85 90 90 90
General SOC/SIP requirements
DFT support for logic and other non-memory circuit repair. 50 60 60 60
% blocks with repair
35 35 40 40
15 15 10 10
DFT efficacy in test volume reduction. Reduction factor 5× 5× 5× 10×
45 45 50 50
Correlation of DFT results with existing specification-based test methods. % results correlated
Availability of fault/defect models for DFT-oriented test methods. % AMS/RF blocks with accepted fault models
DFT reuse for performance calibration, and measurement purposes. % DFT circuits reused
DFT impact on system performance (noise, power, sensitivity, bandwidth, etc.). % performance impact (aggregate figure of merit)
DFT / ATE interface standard, including DFT control via standard test access protocols. % of test interface standardized
ORTCINDEX
2007 ITRS Chapters
2008INDEX
2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022
60 60 60 60 80 85 90 90 100 100 100 100
60 60 60 60 80 85 90 90 100 100 100 100
45 50 55 60 65 70 75 80 85 90 95 100
75 75 80 80 85
60 60 70 70 70 85 90 90 95 95 97.5 100
50 50 55 55 60 80 80 90 90 100 100 100
65 70 80 90 100 100 100
90 95 95 95 95 98 98 98 100 100 100 100
70 70 70 80 80 80 90 90 100 100 105 110
60 60 70 70 70 72.5 75
40 45 45 50 50
10 10 10 10 10 5 5 5 5 5 5 5
10× 10× 20× 20× 20× 20× 50× 50× 50× 50× 50× 50×
60 60 70 70 75 90 80 90 100 100 100 100
Table DESN9 Design for Manufacturability Technology Requirements
Year of Production 2007 2008 2009
Normalized mask cost from public and IDM data 1 1.3 1.7
10% 10% 10%
31% 35% 40%
33% 37% 42%
16% 18% 20%
% CD variability 12% 12% 12%
46% 48% 49%
56% 57% 63%
124% 143% 186%
% Vdd variability: % variability seen in on-chip circuits
% Vth variability: doping variability impact on Vth, (minimum size devices, memory)
% Vth variability: includes all sources
% Vth variability: typical size logic devices, all sources
% circuit performance variability circuit comprising gates and wires
% circuit total power variability circuit comprising gates and wires
% circuit leakage power variability circuit comprising gates and wires
ORTCINDEX
2007 ITRS Chapters
2008INDEX
2010 2011 2012 2013 2014 2015 2016 2017 2018
2.3 3 3.9 5.1 6.6 8.7 11.4 14.9 19.6
10% 10% 10% 10% 10% 10% 10% 10% 10%
40% 40% 58% 58% 81% 81% 81% 81% 112%
42% 42% 58% 58% 81% 81% 81% 81% 112%
20% 20% 26% 26% 36% 36% 36% 36% 50%
12% 12% 12% 12% 12% 12% 12% 12% 12%
51% 60% 63% 63% 63% 63% 63% 65% 66%
68% 72% 76% 80% 84% 88% 92% 96% 102%
229% 255% 281% 287% 294% 331% 368% 381% 395%
2019 2020 2021 2022
25.6 33.6 44.2 57.7
10% 10% 10% 10%
112% 112% 112% 112%
112% 112% 112% 112%
50% 50% 50% 50%
12% 12% 12% 12%
69% 69% 71% 73%
110% 121% 130% 140%
360% 325% 477% 628%
Table DESN10 Correspondence Between Design for Manufacturability Requirements and Solutions Requirement Solution Explanation of the Correspondence
Mask cost Tools that account for mask cost in their algorithms Obvious
RDRs (grid-like layouts, no diagonals, etc.) Better manufacturability and yield, less mask complexity
RET tools aware of circuit metrics (timing, power) More effective optimization, fewer design iterations
Statistical leakage analysis and optimization tools Estimation and control of soaring leakage variability
Post-tapeout RET interacting with synthesis, timing, P&R
By interacting with earlier-in-the-flow EDA tools, can more effectively address litho issues
Model-based physical verification Can address litho issues with precision
Model-based physical synthesis Explicit litho model-based approach moves into the physical synthesis toolset
Manufacturing-friendly design rules (hard rules) Reduces mask, manufacturing cost; addresses printability
% Vdd variability seen at on-chip circuits Tools that account for mask cost in their algorithms Obvious
RDRs (grid-like layouts, no diagonals, etc.) Better manufacturability and yield, less mask complexity
RET tools aware of circuit metrics (timing, power) More effective optimization, fewer design iterations
Statistical leakage analysis and optimization tools Estimation and control of soaring leakage variability
Post-tapeout RET interacting with synthesis, timing, P&R
By interacting with earlier-in-the-flow EDA tools, can more effectively address litho issues
Model-based physical verification Can address litho issues with precision
Model-based physical synthesis Explicit litho model-based approach moves into the physical synthesis toolset
Manufacturing-friendly design rules (hard rules) Reduces mask, manufacturing cost; addresses printability
% Vth variability (doping variability impact) Statistical analysis, opt tools and flows (Vdd, T, Vth) Better estimate of variability impact reduces overdesign
% Vth variability Includes all sources Statistical analysis, opt tools and flows (Vdd, T, Vth) Better estimate of variability impact reduces overdesign
Adaptable and redundant circuits Inherent circuit robustness to variability
Statistical leakage analysis and optimization tools Estimation and control of soaring leakage variability.
% CD variability RET tools aware of circuit metrics (timing, power) More effective optimization, fewer design iterations
RDRs (grid-like layouts, no diagonals, etc.) Better manufacturability and yield, less mask complexity
Adaptable and redundant circuits Inherent circuit robustness to variability
Statistical leakage analysis and optimization tools Leakage power variability will soar. Statistical leakage tools are critical to estimate and control it.
Post-tapeout RET interacting with synthesis, timing, P&R
By interacting with earlier-in-the-flow EDA tools, can more effectively address litho issues
Model-based physical verification Can address litho issues with precision
Model-based physical synthesis Explicit litho model-based approach moves into the physical synthesis toolset
Manufacturing-friendly design rules (hard rules) Reduces mask, manufacturing cost; addresses printability Circuit performance variability (gates and wires) Router-friendly standard cells
Routing-friendly rules reduce design, mask, and manufacturing complexity
Adaptable and redundant circuits Inherent circuit robustness to variability Circuit power variability (gates and wires) Adaptable and redundant circuits Inherent circuit robustness to variability
Statistical leakage analysis and optimization tools Estimation and control of soaring leakage variability.
Post-tapeout RET interacting with synthesis, timing, P&R
By interacting with earlier-in-the-flow EDA tools, can more effectively address litho issues
Model-based physical verification Can address litho issues with precision
Model-based physical synthesis Explicit litho model-based approach moves into the physical synthesis toolset
Manufacturing-friendly design rules (hard rules) Reduces mask, manufacturing cost; addresses printability
Router-friendly standard cells Routing-friendly rules reduce design, mask, and manufacturing complexity
ORTCINDEX
2007 ITRS Chapters
2008INDEX
Table DESN10 Correspondence Between Design for Manufacturability Requirements and Solutions Requirement Solution Explanation of the Correspondence
Mask cost Tools that account for mask cost in their algorithms Obvious
RDRs (grid-like layouts, no diagonals, etc.) Better manufacturability and yield, less mask complexity
RET tools aware of circuit metrics (timing, power) More effective optimization, fewer design iterations
Statistical leakage analysis and optimization tools Estimation and control of soaring leakage variability
Post-tapeout RET interacting with synthesis, timing, P&R
By interacting with earlier-in-the-flow EDA tools, can more effectively address litho issues
Model-based physical verification Can address litho issues with precision
Model-based physical synthesis Explicit litho model-based approach moves into the physical synthesis toolset
Manufacturing-friendly design rules (hard rules) Reduces mask, manufacturing cost; addresses printability
% Vdd variability seen at on-chip circuits Tools that account for mask cost in their algorithms Obvious
RDRs (grid-like layouts, no diagonals, etc.) Better manufacturability and yield, less mask complexity
RET tools aware of circuit metrics (timing, power) More effective optimization, fewer design iterations
Statistical leakage analysis and optimization tools Estimation and control of soaring leakage variability
Post-tapeout RET interacting with synthesis, timing, P&R
By interacting with earlier-in-the-flow EDA tools, can more effectively address litho issues
Model-based physical verification Can address litho issues with precision
Model-based physical synthesis Explicit litho model-based approach moves into the physical synthesis toolset
Manufacturing-friendly design rules (hard rules) Reduces mask, manufacturing cost; addresses printability
% Vth variability (doping variability impact) Statistical analysis, opt tools and flows (Vdd, T, Vth) Better estimate of variability impact reduces overdesign
% Vth variability Includes all sources Statistical analysis, opt tools and flows (Vdd, T, Vth) Better estimate of variability impact reduces overdesign
Adaptable and redundant circuits Inherent circuit robustness to variability
Statistical leakage analysis and optimization tools Estimation and control of soaring leakage variability.
% CD variability RET tools aware of circuit metrics (timing, power) More effective optimization, fewer design iterations
RDRs (grid-like layouts, no diagonals, etc.) Better manufacturability and yield, less mask complexity
Adaptable and redundant circuits Inherent circuit robustness to variability
Statistical leakage analysis and optimization tools Leakage power variability will soar. Statistical leakage tools are critical to estimate and control it.
Post-tapeout RET interacting with synthesis, timing, P&R
By interacting with earlier-in-the-flow EDA tools, can more effectively address litho issues
Model-based physical verification Can address litho issues with precision
Model-based physical synthesis Explicit litho model-based approach moves into the physical synthesis toolset
Manufacturing-friendly design rules (hard rules) Reduces mask, manufacturing cost; addresses printability Circuit performance variability (gates and wires) Router-friendly standard cells
Routing-friendly rules reduce design, mask, and manufacturing complexity
Adaptable and redundant circuits Inherent circuit robustness to variability Circuit power variability (gates and wires) Adaptable and redundant circuits Inherent circuit robustness to variability
Statistical leakage analysis and optimization tools Estimation and control of soaring leakage variability.
Post-tapeout RET interacting with synthesis, timing, P&R
By interacting with earlier-in-the-flow EDA tools, can more effectively address litho issues
Model-based physical verification Can address litho issues with precision
Model-based physical synthesis Explicit litho model-based approach moves into the physical synthesis toolset
Manufacturing-friendly design rules (hard rules) Reduces mask, manufacturing cost; addresses printability
Router-friendly standard cells Routing-friendly rules reduce design, mask, and manufacturing complexity
Table DESN10 Correspondence Between Design for Manufacturability Requirements and Solutions Requirement Solution Explanation of the Correspondence
Mask cost Tools that account for mask cost in their algorithms Obvious
RDRs (grid-like layouts, no diagonals, etc.) Better manufacturability and yield, less mask complexity
RET tools aware of circuit metrics (timing, power) More effective optimization, fewer design iterations
Statistical leakage analysis and optimization tools Estimation and control of soaring leakage variability
Post-tapeout RET interacting with synthesis, timing, P&R
By interacting with earlier-in-the-flow EDA tools, can more effectively address litho issues
Model-based physical verification Can address litho issues with precision
Model-based physical synthesis Explicit litho model-based approach moves into the physical synthesis toolset
Manufacturing-friendly design rules (hard rules) Reduces mask, manufacturing cost; addresses printability
% Vdd variability seen at on-chip circuits Tools that account for mask cost in their algorithms Obvious
RDRs (grid-like layouts, no diagonals, etc.) Better manufacturability and yield, less mask complexity
RET tools aware of circuit metrics (timing, power) More effective optimization, fewer design iterations
Statistical leakage analysis and optimization tools Estimation and control of soaring leakage variability
Post-tapeout RET interacting with synthesis, timing, P&R
By interacting with earlier-in-the-flow EDA tools, can more effectively address litho issues
Model-based physical verification Can address litho issues with precision
Model-based physical synthesis Explicit litho model-based approach moves into the physical synthesis toolset
Manufacturing-friendly design rules (hard rules) Reduces mask, manufacturing cost; addresses printability
% Vth variability (doping variability impact) Statistical analysis, opt tools and flows (Vdd, T, Vth) Better estimate of variability impact reduces overdesign
% Vth variability Includes all sources Statistical analysis, opt tools and flows (Vdd, T, Vth) Better estimate of variability impact reduces overdesign
Adaptable and redundant circuits Inherent circuit robustness to variability
Statistical leakage analysis and optimization tools Estimation and control of soaring leakage variability.
% CD variability RET tools aware of circuit metrics (timing, power) More effective optimization, fewer design iterations
RDRs (grid-like layouts, no diagonals, etc.) Better manufacturability and yield, less mask complexity
Adaptable and redundant circuits Inherent circuit robustness to variability
Statistical leakage analysis and optimization tools Leakage power variability will soar. Statistical leakage tools are critical to estimate and control it.
Post-tapeout RET interacting with synthesis, timing, P&R
By interacting with earlier-in-the-flow EDA tools, can more effectively address litho issues
Model-based physical verification Can address litho issues with precision
Model-based physical synthesis Explicit litho model-based approach moves into the physical synthesis toolset
Manufacturing-friendly design rules (hard rules) Reduces mask, manufacturing cost; addresses printability Circuit performance variability (gates and wires) Router-friendly standard cells
Routing-friendly rules reduce design, mask, and manufacturing complexity
Adaptable and redundant circuits Inherent circuit robustness to variability Circuit power variability (gates and wires) Adaptable and redundant circuits Inherent circuit robustness to variability
Statistical leakage analysis and optimization tools Estimation and control of soaring leakage variability.
Post-tapeout RET interacting with synthesis, timing, P&R
By interacting with earlier-in-the-flow EDA tools, can more effectively address litho issues
Model-based physical verification Can address litho issues with precision
Model-based physical synthesis Explicit litho model-based approach moves into the physical synthesis toolset
Manufacturing-friendly design rules (hard rules) Reduces mask, manufacturing cost; addresses printability
Router-friendly standard cells Routing-friendly rules reduce design, mask, and manufacturing complexity
Table DESN10 Correspondence Between Design for Manufacturability Requirements and Solutions Requirement Solution Explanation of the Correspondence
Mask cost Tools that account for mask cost in their algorithms Obvious
RDRs (grid-like layouts, no diagonals, etc.) Better manufacturability and yield, less mask complexity
RET tools aware of circuit metrics (timing, power) More effective optimization, fewer design iterations
Statistical leakage analysis and optimization tools Estimation and control of soaring leakage variability
Post-tapeout RET interacting with synthesis, timing, P&R
By interacting with earlier-in-the-flow EDA tools, can more effectively address litho issues
Model-based physical verification Can address litho issues with precision
Model-based physical synthesis Explicit litho model-based approach moves into the physical synthesis toolset
Manufacturing-friendly design rules (hard rules) Reduces mask, manufacturing cost; addresses printability
% Vdd variability seen at on-chip circuits Tools that account for mask cost in their algorithms Obvious
RDRs (grid-like layouts, no diagonals, etc.) Better manufacturability and yield, less mask complexity
RET tools aware of circuit metrics (timing, power) More effective optimization, fewer design iterations
Statistical leakage analysis and optimization tools Estimation and control of soaring leakage variability
Post-tapeout RET interacting with synthesis, timing, P&R
By interacting with earlier-in-the-flow EDA tools, can more effectively address litho issues
Model-based physical verification Can address litho issues with precision
Model-based physical synthesis Explicit litho model-based approach moves into the physical synthesis toolset
Manufacturing-friendly design rules (hard rules) Reduces mask, manufacturing cost; addresses printability
% Vth variability (doping variability impact) Statistical analysis, opt tools and flows (Vdd, T, Vth) Better estimate of variability impact reduces overdesign
% Vth variability Includes all sources Statistical analysis, opt tools and flows (Vdd, T, Vth) Better estimate of variability impact reduces overdesign
Adaptable and redundant circuits Inherent circuit robustness to variability
Statistical leakage analysis and optimization tools Estimation and control of soaring leakage variability.
% CD variability RET tools aware of circuit metrics (timing, power) More effective optimization, fewer design iterations
RDRs (grid-like layouts, no diagonals, etc.) Better manufacturability and yield, less mask complexity
Adaptable and redundant circuits Inherent circuit robustness to variability
Statistical leakage analysis and optimization tools Leakage power variability will soar. Statistical leakage tools are critical to estimate and control it.
Post-tapeout RET interacting with synthesis, timing, P&R
By interacting with earlier-in-the-flow EDA tools, can more effectively address litho issues
Model-based physical verification Can address litho issues with precision
Model-based physical synthesis Explicit litho model-based approach moves into the physical synthesis toolset
Manufacturing-friendly design rules (hard rules) Reduces mask, manufacturing cost; addresses printability Circuit performance variability (gates and wires) Router-friendly standard cells
Routing-friendly rules reduce design, mask, and manufacturing complexity
Adaptable and redundant circuits Inherent circuit robustness to variability Circuit power variability (gates and wires) Adaptable and redundant circuits Inherent circuit robustness to variability
Statistical leakage analysis and optimization tools Estimation and control of soaring leakage variability.
Post-tapeout RET interacting with synthesis, timing, P&R
By interacting with earlier-in-the-flow EDA tools, can more effectively address litho issues
Model-based physical verification Can address litho issues with precision
Model-based physical synthesis Explicit litho model-based approach moves into the physical synthesis toolset
Manufacturing-friendly design rules (hard rules) Reduces mask, manufacturing cost; addresses printability
Router-friendly standard cells Routing-friendly rules reduce design, mask, and manufacturing complexity
Table DESN11 Near-term Breakthroughs in Design Technology for AMS
Field of Breakthrough 2007 State-of-the-Art 2008/09 2010/11 Specification, validation, verification
Established AMS Hardware Description Languages
Multi-language support, AMS extension of HW/SW description languages for full system simulation
Complete specification-driven design flow; some specialized formal verification methods
Architectural design Algorithm-oriented design (e.g., with Matlab/Simulink)
Language-based performance evaluation; closer coupling of architectural, block, and circuit level
Synthesizeable AMS description; power-aware HW/SW partitioning extended to AMS systems
Physical mixed A/D and RF design
Procedural layout generation, module generators for a few block types
Module generators for often re-used blocks, design centering, performance estimation
Synthesis: behavior to layout (at least for the most important building blocks)
Parasitics extraction, automated modeling, accelerated simulation
Electromagnetic immunity simulation works but is too complicated for broad usage
2D/3D model-based order reduction for interconnect systems and substrate effects on chip, thermal package modeling
New fault-tolerant circuit architectures, robustness against technology parameter variations; order reduction for all kinds of parasitics and antennas
ORTCINDEX
2007 ITRS Chapters
2008INDEX
Table DESN11 Near-term Breakthroughs in Design Technology for AMS
Field of Breakthrough 2007 State-of-the-Art 2008/09 2010/11 Specification, validation, verification
Established AMS Hardware Description Languages
Multi-language support, AMS extension of HW/SW description languages for full system simulation
Complete specification-driven design flow; some specialized formal verification methods
Architectural design Algorithm-oriented design (e.g., with Matlab/Simulink)
Language-based performance evaluation; closer coupling of architectural, block, and circuit level
Synthesizeable AMS description; power-aware HW/SW partitioning extended to AMS systems
Physical mixed A/D and RF design
Procedural layout generation, module generators for a few block types
Module generators for often re-used blocks, design centering, performance estimation
Synthesis: behavior to layout (at least for the most important building blocks)
Parasitics extraction, automated modeling, accelerated simulation
Electromagnetic immunity simulation works but is too complicated for broad usage
2D/3D model-based order reduction for interconnect systems and substrate effects on chip, thermal package modeling
New fault-tolerant circuit architectures, robustness against technology parameter variations; order reduction for all kinds of parasitics and antennas
Table DESN12 Design Technology Improvements and Impact on Designer Productivity
DT Improvement Year Productivity
Delta
Productivity (Gates/Design-
Year)
Cost of Component
Affected Description of Improvement
None 1990 4K
In-house place and route 1993 +38.9% 5.55K PD
Integration Automated block placement and routing
Engineer 1995 +63.6% 9.09K Chip/circuit/PD
Verification
Engineer can pursue all required tasks to complete a design block, from RTL to GDSII
Reuse—small blocks 1997 +340% 40K Circuit/PD
Verification Blocks from 2,500–74,999 gates
Reuse—large blocks 1999 +38.9% 56K Chip/circuit/PD
Integration Verification
Blocks from 75,000–1M gates
IC implementation suite 2001 +63.6% 91K Chip/circuit/PD
Integration EDA support
Tightly integrated tool set that goes from RTL synthesis to GDSII through IC place and route
RTL functional verification tool suite
2003 +37.5% 125K SW development
Verification
RTL verification tool (“cockpit”) that takes an ES-level description and partitions it into verifiable blocks, then executes verification tools on the blocks, while tracking and reporting code coverage
Transactional Modeling 2005 +60% 200K SW development
Verification
Level above RTL, including both HW and SW design; it consists of a behavioral (where the system function has not been partitioned) and an architectural level (where HW and SW are identified and handed off to design teams)
Very large block reuse 2007 +200% 600K Chip/circuit/PD
Verification Blocks >1M gates; intellectual-property cores
Homogeneous parallel processing
2009 +100% HW
+100% SW 1200K
Chip/circuit/PD Design and Verification
Many identical cores provide specialized processing around a main processor, which allows for performance, power efficiency, and high reuse
Intelligent test bench 2011 37.5% 1650K Chip/circuit/PD
Verification
Like RTL verification tool suite, but also with automation of the Verification Partitioning step
Concurrent software compiler
2013 200% SW 1650K
Chip and Electronic System
Design and Verification
Enables compilation and SW development in highly parallel processing SOCs
Heterogeneous massive parallel processing
2015 +100% HW +100% SW
3300K System Electronic
Design and Verification
Each of the specialized cores around the main processor is not identical from the programming and implementation standpoint
Transactional Memory 2017 +100% HW +100% SW
6600K System Electronic
Design and Verification
Automates true electronic system design on- and off-chip for the first time, including heterogeneous technologies (Phase 1)
System-level DA 2019 60% HW 38% SW
10557K System Electronic
Design and Verification
Automates true electronic system design on- and off-chip for the first time, including heterogeneous technologies (Phase 2)
Executable specification 2021 200% HW +200% SW
31671K System Electronic
Design and Verification
Automates true electronic system design on- and off-chip for the first time, including heterogeneous technologies (Phase 3)
Total +264,000%
ORTCINDEX
2007 ITRS Chapters
2008INDEX
Table DESN12 Design Technology Improvements and Impact on Designer Productivity
DT Improvement Year Productivity
Delta
Productivity (Gates/Design-
Year)
Cost of Component
Affected Description of Improvement
None 1990 4K
In-house place and route 1993 +38.9% 5.55K PD
Integration Automated block placement and routing
Engineer 1995 +63.6% 9.09K Chip/circuit/PD
Verification
Engineer can pursue all required tasks to complete a design block, from RTL to GDSII
Reuse—small blocks 1997 +340% 40K Circuit/PD
Verification Blocks from 2,500–74,999 gates
Reuse—large blocks 1999 +38.9% 56K Chip/circuit/PD
Integration Verification
Blocks from 75,000–1M gates
IC implementation suite 2001 +63.6% 91K Chip/circuit/PD
Integration EDA support
Tightly integrated tool set that goes from RTL synthesis to GDSII through IC place and route
RTL functional verification tool suite
2003 +37.5% 125K SW development
Verification
RTL verification tool (“cockpit”) that takes an ES-level description and partitions it into verifiable blocks, then executes verification tools on the blocks, while tracking and reporting code coverage
Transactional Modeling 2005 +60% 200K SW development
Verification
Level above RTL, including both HW and SW design; it consists of a behavioral (where the system function has not been partitioned) and an architectural level (where HW and SW are identified and handed off to design teams)
Very large block reuse 2007 +200% 600K Chip/circuit/PD
Verification Blocks >1M gates; intellectual-property cores
Homogeneous parallel processing
2009 +100% HW
+100% SW 1200K
Chip/circuit/PD Design and Verification
Many identical cores provide specialized processing around a main processor, which allows for performance, power efficiency, and high reuse
Intelligent test bench 2011 37.5% 1650K Chip/circuit/PD
Verification
Like RTL verification tool suite, but also with automation of the Verification Partitioning step
Concurrent software compiler
2013 200% SW 1650K
Chip and Electronic System
Design and Verification
Enables compilation and SW development in highly parallel processing SOCs
Heterogeneous massive parallel processing
2015 +100% HW +100% SW
3300K System Electronic
Design and Verification
Each of the specialized cores around the main processor is not identical from the programming and implementation standpoint
Transactional Memory 2017 +100% HW +100% SW
6600K System Electronic
Design and Verification
Automates true electronic system design on- and off-chip for the first time, including heterogeneous technologies (Phase 1)
System-level DA 2019 60% HW 38% SW
10557K System Electronic
Design and Verification
Automates true electronic system design on- and off-chip for the first time, including heterogeneous technologies (Phase 2)
Executable specification 2021 200% HW +200% SW
31671K System Electronic
Design and Verification
Automates true electronic system design on- and off-chip for the first time, including heterogeneous technologies (Phase 3)
Total +264,000%
Table DESN12 Design Technology Improvements and Impact on Designer Productivity
DT Improvement Year Productivity
Delta
Productivity (Gates/Design-
Year)
Cost of Component
Affected Description of Improvement
None 1990 4K
In-house place and route 1993 +38.9% 5.55K PD
Integration Automated block placement and routing
Engineer 1995 +63.6% 9.09K Chip/circuit/PD
Verification
Engineer can pursue all required tasks to complete a design block, from RTL to GDSII
Reuse—small blocks 1997 +340% 40K Circuit/PD
Verification Blocks from 2,500–74,999 gates
Reuse—large blocks 1999 +38.9% 56K Chip/circuit/PD
Integration Verification
Blocks from 75,000–1M gates
IC implementation suite 2001 +63.6% 91K Chip/circuit/PD
Integration EDA support
Tightly integrated tool set that goes from RTL synthesis to GDSII through IC place and route
RTL functional verification tool suite
2003 +37.5% 125K SW development
Verification
RTL verification tool (“cockpit”) that takes an ES-level description and partitions it into verifiable blocks, then executes verification tools on the blocks, while tracking and reporting code coverage
Transactional Modeling 2005 +60% 200K SW development
Verification
Level above RTL, including both HW and SW design; it consists of a behavioral (where the system function has not been partitioned) and an architectural level (where HW and SW are identified and handed off to design teams)
Very large block reuse 2007 +200% 600K Chip/circuit/PD
Verification Blocks >1M gates; intellectual-property cores
Homogeneous parallel processing
2009 +100% HW
+100% SW 1200K
Chip/circuit/PD Design and Verification
Many identical cores provide specialized processing around a main processor, which allows for performance, power efficiency, and high reuse
Intelligent test bench 2011 37.5% 1650K Chip/circuit/PD
Verification
Like RTL verification tool suite, but also with automation of the Verification Partitioning step
Concurrent software compiler
2013 200% SW 1650K
Chip and Electronic System
Design and Verification
Enables compilation and SW development in highly parallel processing SOCs
Heterogeneous massive parallel processing
2015 +100% HW +100% SW
3300K System Electronic
Design and Verification
Each of the specialized cores around the main processor is not identical from the programming and implementation standpoint
Transactional Memory 2017 +100% HW +100% SW
6600K System Electronic
Design and Verification
Automates true electronic system design on- and off-chip for the first time, including heterogeneous technologies (Phase 1)
System-level DA 2019 60% HW 38% SW
10557K System Electronic
Design and Verification
Automates true electronic system design on- and off-chip for the first time, including heterogeneous technologies (Phase 2)
Executable specification 2021 200% HW +200% SW
31671K System Electronic
Design and Verification
Automates true electronic system design on- and off-chip for the first time, including heterogeneous technologies (Phase 3)
Total +264,000%
Table DESN12 Design Technology Improvements and Impact on Designer Productivity
DT Improvement Year Productivity
Delta
Productivity (Gates/Design-
Year)
Cost of Component
Affected Description of Improvement
None 1990 4K
In-house place and route 1993 +38.9% 5.55K PD
Integration Automated block placement and routing
Engineer 1995 +63.6% 9.09K Chip/circuit/PD
Verification
Engineer can pursue all required tasks to complete a design block, from RTL to GDSII
Reuse—small blocks 1997 +340% 40K Circuit/PD
Verification Blocks from 2,500–74,999 gates
Reuse—large blocks 1999 +38.9% 56K Chip/circuit/PD
Integration Verification
Blocks from 75,000–1M gates
IC implementation suite 2001 +63.6% 91K Chip/circuit/PD
Integration EDA support
Tightly integrated tool set that goes from RTL synthesis to GDSII through IC place and route
RTL functional verification tool suite
2003 +37.5% 125K SW development
Verification
RTL verification tool (“cockpit”) that takes an ES-level description and partitions it into verifiable blocks, then executes verification tools on the blocks, while tracking and reporting code coverage
Transactional Modeling 2005 +60% 200K SW development
Verification
Level above RTL, including both HW and SW design; it consists of a behavioral (where the system function has not been partitioned) and an architectural level (where HW and SW are identified and handed off to design teams)
Very large block reuse 2007 +200% 600K Chip/circuit/PD
Verification Blocks >1M gates; intellectual-property cores
Homogeneous parallel processing
2009 +100% HW
+100% SW 1200K
Chip/circuit/PD Design and Verification
Many identical cores provide specialized processing around a main processor, which allows for performance, power efficiency, and high reuse
Intelligent test bench 2011 37.5% 1650K Chip/circuit/PD
Verification
Like RTL verification tool suite, but also with automation of the Verification Partitioning step
Concurrent software compiler
2013 200% SW 1650K
Chip and Electronic System
Design and Verification
Enables compilation and SW development in highly parallel processing SOCs
Heterogeneous massive parallel processing
2015 +100% HW +100% SW
3300K System Electronic
Design and Verification
Each of the specialized cores around the main processor is not identical from the programming and implementation standpoint
Transactional Memory 2017 +100% HW +100% SW
6600K System Electronic
Design and Verification
Automates true electronic system design on- and off-chip for the first time, including heterogeneous technologies (Phase 1)
System-level DA 2019 60% HW 38% SW
10557K System Electronic
Design and Verification
Automates true electronic system design on- and off-chip for the first time, including heterogeneous technologies (Phase 2)
Executable specification 2021 200% HW +200% SW
31671K System Electronic
Design and Verification
Automates true electronic system design on- and off-chip for the first time, including heterogeneous technologies (Phase 3)
Total +264,000%
Table TST1 Summary of Key Test Drivers, Challenges, and Opportunities
Key Drivers (not in any particular order)
Device trends
Increasing device interface bandwidth (# of signals and data rates)
Increasing device integration (SoC, SiP, MCP, 3D packaging)
Integration of emerging and non-digital CMOS technologies
Complex package electrical and mechanical characteristics
Device characteristics beyond one sided stimulus/response model
Multiple I/O types and power supplies on same device
Multiple digital I/O types on same device
Increasing test process complexity
Device customization during the test process
“Distributed test” to maintain cost scaling
Feedback data for tuning manufacturing
Dynamic test flows via “Adaptive Test”
Higher order dimensionality of test conditions
Continued economic scaling of test
Physical limits of test parallelism
Managing (logic) test data and feedback data volume
Defining an effective limit for performance difference for HVM ATE versus DUT
Managing interface hardware and (test) socket costs
Trade-off between the cost of test and the cost of quality
Multiple insertions due to system test and BIST
Difficult Challenges (in order of priority)
Test for yield learning Critically essential for fab process and device learning below optical device dimensions
Detecting Systemic Defects Testing for local non-uniformities, not just hard defects
Detecting symptoms and effects of line width variations, finite dopant distributions, systemic process defects
Screening for reliability Implementation challenges and efficacies of burn-in, IDDQ, and Vstress
Erratic, non deterministic, and intermittent device behavior
Potential yield losses
Tester inaccuracies (timing, voltage, current, temperature control, etc)
Over testing (e.g., delay faults on non-functional paths)
Mechanical damage during the testing process
Defects in test-only circuitry or spec failures in a test mode e.g., BIST, power, noise
Some IDDQ-only failures
Faulty repairs of normally repairable circuits
Decisions made on overly aggressive statistical post-processing
Future Opportunities (not in any order)
Test program automation (not ATPG) Automation of generation of entire test programs for ATE
Simulation and modeling Seamless Integration of simulation and modeling of test interface hardware and instrumentation into the device design process
Convergence of test and system reliability solutions
Re-use and fungibility of solutions between test (DFT), device, and system reliability (error detection, reporting, correction)
ATE—automatic test equipment ATPG—automatic test pattern generation BIST—built-in self test HVM—high volume manufacturing MCP—multi-chip packaging MEMs—micro-electromechanical systems
ORTCINDEX
2007 ITRS Chapters
2008INDEX
Table TST1 Summary of Key Test Drivers, Challenges, and Opportunities
Key Drivers (not in any particular order)
Device trends
Increasing device interface bandwidth (# of signals and data rates)
Increasing device integration (SoC, SiP, MCP, 3D packaging)
Integration of emerging and non-digital CMOS technologies
Complex package electrical and mechanical characteristics
Device characteristics beyond one sided stimulus/response model
Multiple I/O types and power supplies on same device
Multiple digital I/O types on same device
Increasing test process complexity
Device customization during the test process
“Distributed test” to maintain cost scaling
Feedback data for tuning manufacturing
Dynamic test flows via “Adaptive Test”
Higher order dimensionality of test conditions
Continued economic scaling of test
Physical limits of test parallelism
Managing (logic) test data and feedback data volume
Defining an effective limit for performance difference for HVM ATE versus DUT
Managing interface hardware and (test) socket costs
Trade-off between the cost of test and the cost of quality
Multiple insertions due to system test and BIST
Difficult Challenges (in order of priority)
Test for yield learning Critically essential for fab process and device learning below optical device dimensions
Detecting Systemic Defects Testing for local non-uniformities, not just hard defects
Detecting symptoms and effects of line width variations, finite dopant distributions, systemic process defects
Screening for reliability Implementation challenges and efficacies of burn-in, IDDQ, and Vstress
Erratic, non deterministic, and intermittent device behavior
Potential yield losses
Tester inaccuracies (timing, voltage, current, temperature control, etc)
Over testing (e.g., delay faults on non-functional paths)
Mechanical damage during the testing process
Defects in test-only circuitry or spec failures in a test mode e.g., BIST, power, noise
Some IDDQ-only failures
Faulty repairs of normally repairable circuits
Decisions made on overly aggressive statistical post-processing
Future Opportunities (not in any order)
Test program automation (not ATPG) Automation of generation of entire test programs for ATE
Simulation and modeling Seamless Integration of simulation and modeling of test interface hardware and instrumentation into the device design process
Convergence of test and system reliability solutions
Re-use and fungibility of solutions between test (DFT), device, and system reliability (error detection, reporting, correction)
ATE—automatic test equipment ATPG—automatic test pattern generation BIST—built-in self test HVM—high volume manufacturing MCP—multi-chip packaging MEMs—micro-electromechanical systems
document.xls
2008_TST2
2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022
High Performance MPU, ASIC
Wafer test Number of sites 8 16 16 16 16 32 32 32 64 64 64 64 64 64 64 64
Wafer test Number of sites 1 1 1 1 1 2 2 2 4 4 4 4 8 8 8 8
Package test Number of sites 4 8 8 8 8 16 16 16 32 32 32 32 32 32 32 32
Package test Number of sites 1 2 2 2 2 4 4 4 4 4 4 8 8 8 16 16
SoC
Wafer test Number of sites 4 4 4 4 4 8 8 8 16 16 16 16 16 16 16 16
Package test Number of sites 4 4 4 4 4 8 8 8 16 16 32 32 32 32 32 32Low Performance - MCU, MPU, ASIC
Wafer test Number of sites 32 64 64 64 64 64 64 64 128 128 128 128 128 128 128 128
Wafer test Number of sites 16 32 32 32 32 64 64 64 128 128 128 128 128 128 128 128
Package test Number of sites 32 64 64 64 64 64 64 64 128 128 128 128 128 128 128 128
Package test Number of sites 8 16 16 16 16 32 32 32 64 64 64 64 64 64 64 64
Mixed-signal & Communications
Wafer test Number of sites 4 8 8 16 16 16 16 16 32 32 64 128 128 256 256 256
wafer test Number of sites 8 16 16 32 32 32 32 32 64 64 64 64 64 128 128 128
Package test Number of sites 8 16 16 16 16 16 64 64 128 128 256 256 256 512 512 512
Packaged Test Number of sites 4 8 8 16 16 16 16 16 32 32 32 32 32 32 32 32Commodity DRAM Memory
Wafer test Number of sites 256 512 512 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000
Wafer test Number of sites 384 384 384 512 512 512 768 768 768 1536 1536 1536 2048 2048 2048 2048
Package test Number of sites 256 512 512 512 512 512 512 512 1024 1024 1024 1024 1024 1024 1024 1024
Package test Number of sites 256 256 256 512 512 512 1024 1024 1024 1024 1024 1024 1024 1024 1024 1024Commodity Flash Memory
Wafer test Number of sites256 512 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000
Wafer test Number of sites768 768 768 1024 1024 1024 1536 1536 1536 2048 2048 2048 2048 2048 2048 2048
Package test Number of sites512 512 512 1024 1024 1024 1024 1024 2048 2048 2048 2048 2048 2048 2048 2048
Package test Number of sites700 700 700 700 1024 1024 1024 1024 1024 2048 2048 2048 2048 2048 2048 2048
RF
Wafer test Number of sites 4 4 8 8 16 16 16 32 32 32 64 128 128 256 256 256
Wafer test Number of sites 4 4 8 16 16 16 32 32 32 32 64 64 64 128 128 128
Package test Number of sites 8 16 32 48 64 64 64 128 128 128 256 256 256 512 512 512
document.xls
2007_TST3
Table TST3 System on Chip Test Requirements
Year of Production 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022
Embedded Cores: Logic
Random Pattern Logic BIST
Area Investment beyond Scan (%) [1] 3.1 3.1 3.1 3.1 3.1 3.1 3.1 3.1 3.1 3.1 3.1 3.1 3.1 3.1 3.1 3.1
Compressed Deterministic Pattern Test
Area Investment beyond Scan (%) [2] 1.1 1.1 1.2 1.3 1.4 1.5 1.6 1.6 1.7 1.8 1.9 2 2.1 2.1 2.1 2.1
SA+T SA+T SA+T +SD +SD +SD +SDX +SDX +SDX +NDF +NDF +NDF +NDF +NDF +NDF +NDF
Ratio of Overall Pattern Count per Gate to Stuck-at Fault Pattern [3] X 5 X 5 X 5 X 15 X 15 X 15 X 30 X 30 X 30 X 60 X 60 X 60 X 120 X 120 X 120 X 240
Estimation & Requirements for SoC Test Pattern (without Core-Parallel Test)
SAF Pattern Count per Chip (k) 11 14 17 22 28 34 43 56 70 90 109 134 170 222 258 361
Overall Pattern Count per Chip (k) 53 68 85 334 416 510 1,283 1,665 2,085 5,370 6,510 8,040 20,370 26,640 30,990 86,700
Ratio of Test Application Time per Chip to 2007's [4] 1.00 1.29 1.62 4.24 5.29 6.48 10.81 14.03 17.57 30.26 36.69 45.31 76.68 100.28 116.66 217.29
1.00 0.96 0.93 2.00 2.00 1.90 2.65 2.65 2.65 3.55 3.55 3.55 4.74 4.74 3.67 5.04
Required Test Data Volume Compression Ratio [6] 30 51 84 202 314 496 746 1,257 1,972 3,269 4,805 7,329 11,761 20,116 35,180 66,721
DFT Methodology for SoC Level Design
DRC DRC DRC +TA +TA +TA +TA +TA +TA +SYN +SYN +SYN +SYN +SYN +SYN +SYN
AH AH AH PA PA PA LA LA LA GA GA GA GA GA GA GA
PA PA PA +NA +NA +NA +NAX +NAX +NAX +NAX +NAX +NAX +NAX +NAX +NAX +NAX
Embedded cores: Memory (SRAM)
RC RC RC RCM RCM RCM M M M M M M M M M M
Area Investment of BIST/BISR/BISD [8] (Kgates/Mbits) 35 35 35 35 35 35 35 35 35 35 35 35 35
Standardized High-Speed Memory Test I/F [9] (S: Some, P: Partially, F: Fully) S S S P P P P P P F F F F F F F
Core Integration
Standardization of I/F for Reusable IP Cores [10] (P: Partial Use, F: Full Use) P P P P P P F F F F F F F F F F
LP LP LP LF LF LF F F F F F F F F F F
AH AH AH L+M L+M L+M +IO +IO +IO +A +A +A +A +A +A +A
F F F PA PA PA PA PA PA FA FA FA FA FA FA FA
SoC Manufacturing
Systematic Hierarchical Diagnosis (L: Logic, M: Memory, I: Interface) L L L +M +M +M +I +I +I +I +I +I +I +I +I +I
C C C +D +D +D +CT +CT +CT +CT +TRF +TRF +TRF +TRF +TRF +TRF
ATE ATE ATE +DFT +DFT +DFT +PFA +PFA +PFA +PFA +PFA +PFA +PFA +PFA +PFA +PFA
SI(B) SI(B) SI(B) +SI(G) +SI(G) +SI(G) +SI(G) +SI(G) +SI(G) +AD +AD +AD +AD +AD +AD +AD
Manufacturable solutions exist, and are being optimized
Manufacturable solutions are known
Interim solutions are known uManufacturable solutions are NOT known
Definitions for System on Chip Test Requirements Table: [1] Area investment of random pattern logic BIST consists of BIST controller and test points.[2] Area investment of compressed deterministic pattern test consists of controller and test points.[3] This shows the number of pattern count (number of captures), which corresponds to various fault models. [4] This is proportional to the overall pattern count and inversely proportional to internal scan data rate.[5] We set the requirement that test application time per gate should be stable.[6] The size of ATE vector memory is assumed to increase as fast as DRAM bit size increases.[7] Growing number of row & column spares, and both divided and shared spares for segments in the future.[8] The current BISR for two dimensional repair is limited to a few row and column spares.[9] Common interface of test logic embedded in memory hard macro for high-speed testing.[10] IEEE1500 is an example. Standardization of I/F for re-usable IP Cores.
[12] A method to obtain overall test quality measure of SoC considering all cores; logic, memory and analog.
Supported Fault Models by ATPG for Overall Test (SA+T: Stuck-at & Transition, SD: Small Delay, SDX: Extended Small Delay, NDF: New Defect-based Fault Model)
Ratio of Test Application Time per Gate to 2007's = Required Reduction Ratio [5]
DFT method in High Level Design Phase (DRC: DFT Design Rule Check, TA: Testability Analysis and Fault Coverage Estimation, SYN: Test Synthesis)
Application of BISR for Logic Cores (AH: Ad hoc Method, PA: Partially Automated Method, LA: Limited Use of Automated Method, GA: General Use of Automated Method)
DFT/ATPG Approach to Reduce Yield-Loss (PA: Power-Aware DFT/ATPG, NA: Noise Aware DFT/ATPG, NAX: Extended Noise-Aware DFT/ATPG)
Repairing Mechanism of Memory Cells to improve Yield [7] ( RC: BISR/BISD for a few Row & Col R/D, RCM: for more Row & Col R/D, M: for More Sophisticated R/D)
u35 u35 u35
Standardization of DFT-ATE I/F [11] (LP: Limited Use of Partial Information, LF: Limited Use of Full Information, F: General Use of Full Information)
SoC Level Fault Coverage [12] (AH: Adhoc, L: Logic, M: Memory, IO: I/O, A: Analog)
Inter-Core/Core-Interface Test (F: Complemental Functional Test; PA: Partially Automated ; FA: Fully Automated)
Supported Defect Type for Fault Diagnosis (C: Conventional (SAF, TF, BF), D: Delay Fault Model Considering Defective Delay Size, CT: Cross-talk, TRF: Transient Response Fault)
Standardized Diagnosis Interface/Data in the diagnosis flow (ATE: Tester Log, DFT: DFT Method, PFA: Physical Failure Analysis)
Volume Diagnosis Data Base (SI: Collection and Storing Defect Information (B: Bad sample, G: Good sample), AD: Automated SoC Diagnosis)
[11] STIL (Test Interface Language, IEEE1450.x) is an example. I/F should include not only test vectors, but also parametric factors.
document.xls
2007_TST4
2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022
Device Characteristics
# of Transistors (M) - CPU 386 486 613 772 973 1,226 1,545 1,946 2,452 3,090 3,893 4,905 6,181 7,788 9,812 12,364# of Transistors (M) - Consumer 254 344 450 608 773 926 1,225 1,609 2,031 2,633 3,205 3,973 5,049 6,623 7,714 10,816Chip size at production mm^2 - CPU 140 140 140 140 140 140 140 140 140 140 140 140 140 140 140 140Chip size at production mm^2 - Consumer 64 64 64 64 64 64 64 64 64 64 64 64 64 64 64 64non differential data rate (GT/s) 2 2 2 2 3.2 3.2 3.2 3.2 3.2 3.2 3.2 3.2 3.2 3.2 3.2 3.2Internal Scan data rate (MHz) 50 50 50 75 75 75 113 113 113 169 169 169 253 253 253 380Single ended External Scan Data rate (Mb/s) 400 400 800 800 800 800 1200 1200 1200 1200 1200 1200 1200 1200 1200 1200Differential External Scan Data rate (Gb/s) 3 3 3 3 3 5 5 5 5 5 5 5 5 5 5 5Vdd 0.8–1.1 0.8–1.0 0.8–1.0 0.7–1.0 0.7–1.0 0.7–0.9 0.6–0.9 0.6–0.9 0.6–0.8 0.5–0.8 0.5–0.7 0.5–0.7 0.4–0.7 0.4–0.6 0.4–0.6 0.4–0.6
CPU
CPU total cores 4 4 5 6 6 7 8 9 10 11 12 14 16 17 20 22CPU Unique cores 1 1 2 2 2 2 2 4 4 4 4 4 6 6 6 6Percentage of Memory transistors 65% 65% 70% 70% 70% 70% 70% 75% 75% 75% 75% 75% 80% 80% 80% 80%Percentage of random logic transistors 5% 5% 5% 5% 5% 5% 5% 5% 5% 5% 5% 5% 5% 5% 5% 5%Percentage of core transistors 30% 30% 25% 25% 25% 25% 25% 20% 20% 20% 20% 20% 15% 15% 15% 15%Transistors per Flip Flop 26 26 26 26 26 26 26 26 26 26 26 26 26 26 26 26Supplies per DUT 1–6 1–6 1–6 1–4 1–4 1–3 1–3 1–3 1–3 1–3 1–3 1–3 1–3 1–3 1–3 1–3Number of patterns for high coverage SAF only 6572 7003 8744 9524 10456 11571 12911 17828 20193 23025 26426 30524 37458 43706 51278 60479Total # of bits scanned in (same as scan out) Gb 4 6 11 14 19 24 32 69 92 123 167 228 353 490 685 965
Maximum power consumption at test (W) 200 200 300 300 300 300 300 300 300 300 300 300 300 300 300 300
Maximum power consumption at test (W) - Server 300 300 300 300 300 300 300 400 400 400 400 400 400 400 400 400
Number of logic gates (M) 34 43 46 58 73 92 116 122 153 193 243 307 309 389 491 618
Consumer
Consumer total cores 32 44 58 79 101 126 161 212 268 348 424 526 669 878 1023 1435Consumer Unique cores 4 6 7 10 13 16 20 27 34 44 53 66 84 110 128 179Percentage of Memory transistors 83% 84% 85% 85% 86% 86% 86% 86% 86% 86% 86% 87% 87% 87% 87% 87%Percentage of random logic transistors 2% 2% 1% 1% 1% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0%Percentage of core transistors 15% 14% 14% 14% 13% 14% 14% 14% 14% 14% 14% 13% 13% 13% 13% 13%Transistors per Flip Flop 26 26 26 26 26 26 26 26 26 26 26 26 26 26 26 26Supplies per DUT 1–6 1–6 1–6 1–4 1–4 1–3 1–3 1–3 1–3 1–3 1–3 1–3 1–3 1–3 1–3 1–3Number of patterns for high coverage SAF only 10,795 13,760 16,875 22,800 27,055 32,410 42,875 56,315 71,085 92,155 112,175 129,123 164,093 215,248 250,705 351,520Total # of bits scanned in (same as scan out) Gb 2 3 3 5 8 7 13 22 36 60 89 118 191 328 445 875
Number of logic gates (M) 11 14 17 23 27 32 43 56 71 92 112 129 164 215 251 352
Table TST5 Vector Multipliers
Vector Multiplier
Fault Type Min MaxBF (Bridging Fault) 1.3 1.3
TF (Transition Fault) 3 5SD (Small Delay) 2 40
ORTCINDEX
2007 ITRS Chapters
2008INDEX
document.xls
2008_TST6
2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022
DRAM Characteristics
Capacity (Gbits)
R&D 8 8 16 16 16 32 32 32 64 64 64 128 128 128 256 256
Mass production 2 2 4 4 4 8 8 8 16 16 16 32 32 32 64 64
Mass production I/O data rate (Gb/s) 1.1 1.3 1.3 1.6 1.6 2.1 2.7 2.7 3.2 3.2 4.3 5.3 5.4 6.4 6.4 8.5
Performance I/O data rate (Gb/s) 1.9 2.4 2.4 2.9 2.9 3.8 4.8 4.8 5.8 5.8 7.7 9.6 9.6 11.5 11.5 15.4
Mass production I/O width 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16
Mass Production CLK rate (GHz) 0.5 0.7 0.7 0.8 0.8 1.1 1.3 1.3 1.6 1.6 2.1 2.7 2.7 3.2 3.2 4.3
NAND Characteristics
Capacity (Gbits)
R&D 64 64 128 128 128 256 256 256 512 512 512 1024 1024 1024 2048 2048
Mass production 16 16 32 32 32 64 64 64 128 128 128 256 256 256 512 512
Maximum I/O data rate (Gb/s) 0.05 0.05 0.05 0.066 0.066 0.1 0.1 0.1 0.1 0.133 0.133 0.133 0.133 0.266 0.266 0.266
Maximum I/O data rate (Gb/s) 0.05 0.1 0.133 0.2 0.2 0.266 0.266 0.333 0.333 0.4 0.4 0.4 0.533 0.533 0.533 0.533
Data width (bits) 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16
Power supply voltage range 1.5–5.5 1.5–3.5 1.5–3.5 1.5–3.5 1.5–3.5 1.5–3.5 1.5–3.5 1.0–3.5 1.0–3.5 1.0–3.5 1.0–3.5 1.0–3.5 1.0–3.5 1.0–3.5 1.0–3.5 1.0–3.5
Power supplies per device 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
Maximum current (MA) 35 35 35 35 35 35 35 35 50 50 50 50 50 50 50 50
Tester channels per device 24 24 24 24 24 24 24 24 24 24 24 24 24 24 24 24
NOR Characteristics
Capacity (Gbits)
R&D 4 4 8 8 8 16 16 16 32 32 32 64 64 64 128 128
Mass production 1 1 2 2 2 4 4 4 8 8 8 16 16 16 32 32
Maximum I/O data rate (Gb/s) 0.2 0.2 0.266 0.266 0.266 0.333 0.333 0.333 0.4 0.4 0.4 0.533 0.533 0.533 0.533 0.533
Data width (bits) 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16
Power supply voltage range 1.0–5.5 1.0–5.5 0.9–3.5 0.9–3.5 0.9–3.5 0.9–3.5 0.9–3.5 0.9–3.5 0.9–3.5 0.9–3.5 0.9–3.5 0.9–3.5 0.9–3.5 0.9–3.5 0.9–3.5 0.9–3.5
Power supplies per device 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
Maximum current (MA) 150 150 150 150 150 150 150 150 150 150 150 150 150 150 150 150
Tester channels per test site 72 72 72 72 72 72 72 72 72 72 72 72 72 72
Embedded DRAM
Capacity (Mbits) 256 512 512 512 1024 1024 1024 2048 2048 2048 4096 4096 4096 4096 4096 8192DFT BIST/BISR
Embedded Flash
Capacity (Mbits) 64 128 128 128 256 256 256 512 512 512 1023 1024 1024 1024 2048 2048DFT BIST/BIST/DAT
Embedded SRAM
Capacity (Mbits) 0.5 1 1 1 2 2 2 4 4 4 8 8 8 16 16 16
DFT BIST/BISR
document.xls
2008_TST7
2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022
Low Frequency Waveform [Note 1]
BW (MHz) 50 75 75 75 100 100 100 100 100 100 100 100 100 100 100 100
BW-max (MHz) 120 120 120 320 320 320 320 320 320 320 320 320 320 320 320 320
BW-center (MHz) [Note 2] 45 45 45 80 80 80 80 80 80 80 80 80 80 80 80 80
Sample rate (MS/s) Moving from Nyquist sample rates to over/under sampling sources/digitizers
Sample rate (MS/s) Nyquist sample rates or higher for sources/digitizers
Resolution (bits) DSP computation to 24 bits, effective number of bits limited by noise floor
Noise floor (dB/RT Hz) -155 -160 -160 -160 -165 -165 -165 -165 -165 -165 -165 -165 -165 -165 -165 -165
Noise floor-Max (dB/RT Hz) [Note 3] -160 -160 -160 -160 -165 -165 -165 -165 -165 -165 -165 -165 -165 -165 -165 -165
Noise floor-mid (dB/RT Hz) [Note 4] -145 -145 -155 -155 -155 -155 -155 -155 -155 -155 -155 -155 -155 -155 -155 -155
Very High Frequency Waveform Source [Note 5]
Level V (pk–pk) 4 4 <4 <4 <4 <4 <4 <4 <4 <4 <4 <4 <4 <4 <4 <4
Accuracy (±) 0.50% 0.50% 0.50% 0.50% 0.50% 0.50% 0.50% 0.50% 0.50% 0.50% 0.50% 0.50% 0.50% 0.50% 0.50% 0.50%
BW (GHz) 1.6 1.9 2.25 2.7 2.7 3 3 3.75 3.75 3.75 3.75 3.75 3.75 3.75 3.75 3.75
BW (GHz) 2 2 4 4 4 4 5 5 5 5 5 5 5 5 5 5
Sample rate (GS/s) 6.4 7.6 9 10.8 11 12 12 15 15 15 15 15 15 15 15 15
Sample rate (GS/s) 6.4 6.6 10 10 10 10 12 12 12 12 12 12 12 12 12 12
Resolution (bits) AWG/Sine† 8-10 8-10 8-10 8-10 8-10 10-12 10-12 10-12 10-12 10-12 10-12 10-12 10-12 10-12 10-12 10-12
Resolution (bits) AWG/Sine† 8-10 8-10 8-10 8-10 8-10 8-10 8-10 8-10 8-10 8-10 8-10 8-10 8-10 8-10 8-10 8-10
Noise floor (dB/RT Hz) -140 -140 -140 -140 -140 -145 -145 -145 -145 -145 -145 -145 -145 -145 -145 -145
Very High Frequency Waveform Digitizer [Note 6]
Level V (pk–pk) 4 4 <4 <4 <4 <4 <4 <4 <4 <4 <4 <4 <4 <4 <4 <4
Accuracy (±) 0.50% 0.50% 0.50% 0.50% 0.50% 0.50% 0.50% 0.50% 0.50% 0.50% 0.50% 0.50% 0.50% 0.50% 0.50% 0.50%
BW (GHz) (under sampled) 9.2 10.8 10.8 12.5 12.5 15 15 15 15 15 15 15 15 15 15 15
Sample rate (GS/s) 0.4 0.4 0.4 0.4 0.4 0.6 0.6 0.6 0.6 0.6 0.6 0.6 0.6 0.6 0.6 0.6
Min resolution (bits) 12 12 12 12 12 14 14 14 14 14 14 14 14 14 14 14
Noise floor (dB/RT Hz) -145 -145 -145 -145 -145 -150 -150 -150 -150 -150 -150 -150 -150 -150 -150 -150
Time Measurement
Jitter measurement (ps RMS) Will be driven by high-speed serial communication ports
Frequency measurement (MHz) Will be driven by high-performance ASIC clock rates
Single shot time capability (ps) Will be driven by high-speed serial communication ports
document.xls
2008_TST8
Year of Production 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022
Leading Edge
Carrier Frequency (GHz) [1] 18 18 22 22 60 77 77 95 95 95 95 95 95 95 95 95
Modulation RF BW (MHz) [2] 80 528 528 528 528 528 528 528 528 528 528 528 1000 1000 1000 1000
High Volume
Carrier Frequency (GHz) 6 8 12 12 22 22 36 36 36 36 36 36 36 36 36 36
Carrier Frequency (GHz) [1] 6 8 12 12 22 22 36 36 36 36 36 36 36 36 36 36
Modulation RF BW (MHz) 20 40 80 528 528 528 528 528 528 528 528 528 528 528 1000 1000
Amplitude Accuracy (dB) <0.8 <0.6 <0.5 <0.5 <0.5 <0.25 <0.25 <0.25 <0.25 <0.25 <0.25 <0.25 <0.125 <0.125 <0.125 <0.125
Amplitude Accuracy (dB) <0.8 <0.6 <0.5 <0.5 <0.5 <0.25 <0.25 <0.25 <0.25 <0.25 <0.25 <0.25 <0.125 <0.125 <0.125 <0.125
ACLR (dB) 65 65 70 72 72 72 75 75 80 80 80 85 85 85 85 85
Number of RF Ports per Device <9 <12 <16 <20 <24 <20 <18 <16 <16 <16 <16 <16 <16 <16 <16 <16
Phase Noise (dBc/Hz @ 100k offset) -125 -130 -135 -140 -142 -145 -148 -150 -150 -150 -152 -152 -152 -152 -152 -152
Phase Noise (dBc/Hz @ 100k offset) -125 -130 -135 -140 -142 -145 -148 -150 -150 -150 -152 -152 -152 -152 -152 -152
Error Vector Magnitude 3G/4G [3] 1-2% 1-2% 0.5% 0.5% 0.5% 0.5% 0.5% 0.5% 0.5% 0.5% 0.5% 0.5% 0.5% 0.5% 0.5% 0.5%
OIP3 (dBm) [4] 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30
IIP3 (dBm) [4] 40 50 60 60 60 60 60 60 60 60 60 60 60 60 60 60
Manufacturable solutions exist, and are being optimized
Manufacturable solutions are known
Interim solutions are known uManufacturable solutions are NOT known
document.xls
2007_TST9
Year of Production 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022
Clock input frequency (MHz) 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400Off-chip data frequency (MHz) 75 75 75 75 75 75 75 75 75 75 75 75 75 75 75 75Power dissipation (W per DUT) 600 600 600 600 600 600 600 600 600 600 600 600 600 600 600 600Power Supply Voltage Range (V)
0.5–2.5 0.5–2.5 0.5–2.5 0.5–2.5 0.5-2.5 0.5–2.5 0.5–2.5 0.5–2.5 0.5–2.5 0.5–2.5 0.5–2.5 0.4–2.5 0.4–2.5 0.4–2.5 0.4–2.5 0.4–2.5 Low-end microcontroller 0.7–10.0 0.7–10.0 0.7–10.0 0.5–10 0.5–10 0.5–10 0.5–10 0.5–10 0.5–10 0.5–10 0.5–10 0.5–10 0.5–10 0.5–10 0.5–10 0.5–10
Mixed-signal 0.5–500 0.5–500 0.5–500 0.5–500 0.5–500 0.5–500 0.5–500 0.5–500 0.5–1000 0.5–1000 0.5–1000 0.5–1000 0.5–1000 0.5–1000 0.5–1000 0.5–1000Maximum Number of Signal I/O
High-performance ASIC 384 384 384 384 384 384 384 384 384 384 384 384 384 384 384 384
128 128 128 128 128 128 128 128 128 128 128 128 128 128 128 128 Commodity memory 72 72 72 72 72 72 72 72 72 72 72 72 72 72 72 72Maximum Current (A)
High-performance microprocessor 450 450 450 450 450 450 450 450 450 450 450 450 450 450 450 450 High-performance graphics processor 100 150 200 200 200 200 200 200 200 200 200 200 200 200 200 200 Mixed-signal 20 20 20 30 30 30 30 30 30 30 30 30 30 30 30 30Burn-in Socket
Pin count 3000 3000 3000 3000 3000 3000 3000 3000 3000 3000 3000 3000 3000 3000 3000 3000 Pitch (mm) 0.3 0.3 0.3 0.2 0.2 0.2 0.2 0.2 0.1 0.1 0.1 0.08 0.08 0.08 0.08 0.08 Power consumption (A/Pin) 3 4 4 5 5 5 5 5 5 6 6 6 6 6 6 6Wafer Level Burn-In
Maximum burn-in temperature (ºC) 175±3 175±3 175±3 175±3 175±3 175±3 175±3 175±3 175±3 175±3 175±3 175±3 175±3 175±3 175±3 175±3Pad Layout – Linear
Minimum pad pitch (μm) 65 65 65 65 65 65 65 65 50 50 50 50 50 50 50 50 Minimum pad size (μm) 50 50 50 50 50 50 50 50 40 40 40 40 40 40 40 40 Maximum number of probes 70k 70k 70k 70k 70k 70k 70k 70k 140k 140k 140k 140k 140k 140k 140k 140kPad Layout – Periphery, Area Array
Minimum pad pitch (μm) *1 100 80 80 80 80 80 80 80 60 60 60 60 60 60 60 60 Minimum pad size (μm) 40 35 35 35 35 30 30 30 25 25 25 25 25 25 25 25 Maximum number of probes 150k 150k 150k 150k 150k 150k 150k 150k 300k 300k 300k 300k 300k 300k 300k 300k
10 10 10 20 20 20 20 20 20 20 20 20 20 20 20 20
32 64 64 64 64 64 64 64 128 128 128 256 256 256 256 256
High-performance ASIC / microprocessor / graphics processor
High-performance microprocessor / graphics processor / mixed-signal
Power consumption (W/DUT – Low-end microcontroller, DFT/BIST SOC *2)
Vector memory depth (M vectors – DFT/BIST SOC *2)
Table TST10 Test Handler and Prober Difficult Challenges
High Power Handler
Temperature control and temperature rise control due to high power densities during test
Continuous lot processing (lot cascading), auto-retest, asynchronous device socketing with low-conversion times Better ESD control as products are more sensitive to ESD and on-die protection circuitry increases cost.
Lower stress socketing, low-cost change kits, higher I/O count for new package technologies
Package heat lids change thermal characteristics of device and hander
Multi-site handling capability for short test time devices (1–7 seconds)
Medium Power Hander
Support for stacked die packaging and thin die packaging
Wide range tri-temperature soak requirements (-45ºC to 150ºC) increases system complexity
Continuous lot processing (lot cascading), auto-retest, low conversion times, asynchronous operation
Shielding issues associated with high frequency testing (>10 GHz)
Low Power Handler
A wide variety of package sizes, thicknesses, and ball pitches requires kitless handlers with thin-die handling capability
Package ball-to-package edge gap decreases from 0.6 mm to 0 mm require new handling and socketing methods
Parallelism at greater than x128 drives thermal control and alignment challenges
Prober
Consistent and low thermal resistance across chuck is required to improve temperature control of device under test
Heat dissipation of >100 Watts at > 85ºC is a configuration gap in the prober industry
Advances in probe card technology require a new optical alignment methodology
ORTCINDEX
2007 ITRS Chapters
2008INDEX
Table TST10 Test Handler and Prober Difficult Challenges
High Power Handler
Temperature control and temperature rise control due to high power densities during test
Continuous lot processing (lot cascading), auto-retest, asynchronous device socketing with low-conversion times Better ESD control as products are more sensitive to ESD and on-die protection circuitry increases cost.
Lower stress socketing, low-cost change kits, higher I/O count for new package technologies
Package heat lids change thermal characteristics of device and hander
Multi-site handling capability for short test time devices (1–7 seconds)
Medium Power Hander
Support for stacked die packaging and thin die packaging
Wide range tri-temperature soak requirements (-45ºC to 150ºC) increases system complexity
Continuous lot processing (lot cascading), auto-retest, low conversion times, asynchronous operation
Shielding issues associated with high frequency testing (>10 GHz)
Low Power Handler
A wide variety of package sizes, thicknesses, and ball pitches requires kitless handlers with thin-die handling capability
Package ball-to-package edge gap decreases from 0.6 mm to 0 mm require new handling and socketing methods
Parallelism at greater than x128 drives thermal control and alignment challenges
Prober
Consistent and low thermal resistance across chuck is required to improve temperature control of device under test
Heat dissipation of >100 Watts at > 85ºC is a configuration gap in the prober industry
Advances in probe card technology require a new optical alignment methodology
document.xls
2007_TST11
Year of Production 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022
Wafer diameter (mm) 300 300 300 300 300 300 300 450 450 450 450 450 450 450 450 450
Wafer thickness (um) 80–775 80–775 80–775 80–775 80–775 80–775 80–775 50–1000 50–1000 50–1000 50–1000 50–1000 50–1000 50–1000 50–1000 50–1000
Maximum I/O pads 3000 4000 4000 5300 5300 5300 5300 5300 5300 5300 5300 5300 5300 5300 5300 5300
Chuck X & Y positioning accuracy (um) 2 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1
Chuck Z positioning accuracy (um) 1 1 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5
Probe-to-pad alignment (µm) 4.5 4.5 4.5 3.5 3.5 3.5 3.5 3.5 3.5 3.5 3.5 3.5 3.5 3.5 3.5 3.5
Maximum chuck force (kg) 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100
Set point range (ºC)-30 to +85 -30 to +85 -30 to +85
-45 to -45 to -45 to -45 to -45 to -45 to -45 to -45 to -45 to -45 to -45 to -45 to -45 to
+125 +125 +125 +125 +125 +125 +125 +125 +125 +125 +125 +125 +125
Total power (Watts) 130 130 250 250 250 250 250 250 250 250 250 250 250 250 250 250
60 60 120 120 120 120 120 120 120 120 120 120 120 120 120 120Power density (Watt/cm2)
document.xls
2007_TST12
Year of Production 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022
High, Medium and Low Power
Temperature set point range (ºC) -55 to 175 -55 to 175 -55 to 175 -55 to 175 -55 to 175 -55 to 175 -55 to 175 -55 to 175 -55 to 175 -55 to 175 -55 to 175 -55 to 175 -55 to 175 -55 to 175 -55 to 175 -55 to 175High Power - >10W per DUT
Temperature accuracy at DUT (ºC) ± 2 ± 2 ± 2 ± 2 ± 2 ± 2 ± 2 ± 2 ± 2 ± 2 ± 2 ± 2 ± 2 ± 2 ± 2 ± 2Number of pins/device 750 750 800 800 850 850 850 850 850 850 900 900 900 1000 1000 1000Parallel testing: 1-2 1-2 1-2 1-2 1-2 1-2 1-2 1-2 1-2 1-2 1-2 1-2 1-2 1-2 1-2 1-2Throughput (devices per hour) 1.5–2K 1.5–2K 1.5–2K 2–3.5K 2–3.5K 2–3.5K 2–3.5K 2–3.5K 2–3.5K 2–3.5K 2–3.5K 2–3.5K 2–3.5K 2–3.5K 2–3.5K 2–3.5KIndex time (S) 0.3 0.3 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25Sorting Categories 3–6 3–6 3–6 3–6 3–6 3–6 3–6 3–6 3–6 3–6 3–6 3–6 3–6 3–6 3–6 3–6Allowable device temperature rise (ºC) 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20Maximum socket load per unit (kg) 24 27 30 30 35 35 35 35 35 35 35 35 35 35 35 35Asynchronous capability Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes YesPin/land pitch (mm) 1.1 1.1 1 1 0.8 0.6 0.6 0.6 0.6 0.6 0.4 0.4 0.4 0.4 0.4 0.4
Medium Power - 0.5 to 10W per DUT
Temperature accuracy at DUT (ºC) ± 2 ± 2 ± 2 ± 2 ± 2 ± 2 ± 2 ± 2 ± 2 ± 2 ± 2 ± 2 ± 2 ± 2 ± 2 ± 2Number of pins/device 800 800 850 850 850 850 850 850 900 900 900 1000 1000 1000 1000 1000Parallel testing: 8-16 8-16 8-16 8-16 8-16 8-16 8-16 8-16 8-16 8-16 8-16 8-16 8-16 8-16 8-16 8-16Throughput (devices per hour) 4–6K 4–6K 4–6K 6–10K 6–10K 6–10K 6–10K 6–10K 6–10K 6–10K 6–10K 6–10K 6–10K 6–10K 6–10K 6–10KIndex time (S) 0.3 0.3 0.3 0.3 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25Sorting Categories 3–6 3–6 3–6 3–6 3–6 3–6 3–6 3–6 3–6 3–6 3–6 3–6 3-6 3–6 3–6 3–6Allowable device temperature rise (ºC) 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5Maximum socket load per unit (kg) 50 50 35 60 35 60 60 60 65 65 65 75 75 75 75 75Asynchronous capability Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes YesPin/land pitch (mm) 0.3 0.3 0.3 0.3 0.3 0.3 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2
Low Power - < 0.5W per DUT
Temperature accuracy at DUT (ºC) ±2 ±2 ±1.5 ±1.5 ±1.5 ±1.5 ±1.5 ±1.5 ±1.5 ±1.5 ±1.5 ±1.5 ±1.5 ±1.5 ±1.5 ±1.5
Number of pins/device 6–250 6–250 6–250 6–250 6–250 6–250 6–250 6–250 6–250 6–250 6–250 6–250 6–250 6–250 6–250 6–250
Parallel testing: 128-512 128-1024 128-1024 128-1024 128-1024 128-1024 128-1024 128-1024 128-2048 128-2048 128-2048 128-2048 128-2048 128-2048 128-2048 128-2048
Throughput (devices per hour) 8–10K 12–20K 12–20K 12–20K 12–20K 12–20K 12–20K 12–20K 12–20K 12–20K 12–20K 12–20K 12–20K 12–20K 12–20K 12–20K
Index time (S) 2–5 2–5 2–4 2–4 2–4 2–4 2–4 2–4 2–4 2–4 2–4 2–4 2–4 2–4 2–4 2–4Sorting Categories 5–9 5–9 5–9 5–9 5–9 5–9 5–9 5–9 5–9 5–9 5–9 5–9 5–9 5–9 5–9 5–9
4x6 3x5 3x5 3x5 3x5 3x5 3x5 2x3 2x3 2x3 2x3 2x3 2x3 2x3 2x3 2x3
Pin pitch (mm) 0.4–1.0 0.25–1.0 0.2–1.0 0.2–1.0 0.2–1.0 0.2–1.0 0.2–1.0 0.2–1.0 0.2–1.0 0.2–1.0 0.2–1.0 0.2–1.0 0.2–1.0 0.2–1.0 0.2–1.0 0.2–1.0
Ball edge to package edge clearance (mm) 0.25 0.25 0.25 0.25 0.25 0 0 0 0 0 0 0 0 0 0 0
Minimum package thickness (mm) 0.4–1.8 0.3–1.8 0.2–1.8 0.2–1.8 0.2–1.8 0.2–1.8 0.2–1.8 0.2–1.8 0.2–1.8 0.2–1.8 0.2–1.8 0.2–1.8 0.2–1.8 0.2–1.8 0.2–1.8 0.2–1.8
Min. Pkg. Size(mm2)
Table TST13 Probing Difficult Challenges
Geometry
Probe technologies to support peripheral fine pitch probe of 23 µm peripheral staggered pad probes at effective pitches of 20/40, and fine pitch (45 µm) for dual row, non-staggered probing on all four die sides.
Fine pitch vertical probe technologies to support 130 µm pitch area array solder bump and 50 µm pitch staggered pad devices.
Multi-site pad probing technologies with corner pitch capability below 125 µm.
Reduction of pad damage at probe commensurate with pad size reductions (or better).
Alternative probe technology for 75 µm on 150 µm pitch dense array (vertical probe; bumped device).
Increasing probe array planarity requirements in combination with increasing array size.
Parallel test Need a probe technology to handle the complexity of SoC devices while probing more than one
device.
Current probe technologies have I/O limitations for bumped device probes.
Probing at temperature Reduce effects on probes for non-ambient testing -50°C to 150°C; especially for fine-pitch devices.
For effects on Handlers and Probers, see that section.
Product
Probe technologies to direct probe on copper bond pads including various oxidation considerations.
Probe technologies for probing over active circuitry (including flip-chip).
Probe force
Reduce per pin force required for good contact resistance to lower total load for high pin count and multi DUT probe applications. Evaluation and reduction of probe force requirements to eliminate die damage, including interlayer dielectric damage with lo
A chuck motion model is required to minimize probe damage
Probe cleaning
Development of high temperature (85°C–150°C) in situ cleaning mediums/methods, particularly for fine pitch, multi-DUT, and non-traditional probes.
Reduction of cleaning requirements while maintaining electrical performance to increase lifetime.
A self cleaning probe card is required for fine pitch bumped pad devices
Cost and delivery
Fine pitch or high pin count probe cards are too expensive and take too long to build.
Time and cost to repair fine pitch or high pin count probe cards is very high.
The time between chip design completion (“tape-out”) and the availability of wafers to be probed is less than the time required to design and build a probe card in almost every probe technology except traditional cantilever.
Space transformer lead times are too long, thus causing some vertical probe technologies to have lengthy lead-times.
Probe metrology Tools are required that support fine pitch probe characterization and pad damage measurements.
Metrology correlation is needed for post repair test versus on-floor usage.
High power devices Probe technologies will need to incorporate thermal management features capable of handling
device power dissipations approaching 1000 Watts and the higher currents (≥ 1.5 amp) flowing through individual probe points.
Contact resistance Probe technologies that achieve contact resistance <.5 Ohms initially and throughout use are needed.
A method to measure contact resistance is needed. The traditional continuity test is insufficient to monitor contact resistance.
High frequency probing Traditional probe technologies do not have the necessary electrical bandwidth for higher frequency
devices. At the top end are RF devices, requiring up to 40 GHz.
ORTCINDEX
2007 ITRS Chapters
2008INDEX
Table TST13 Probing Difficult Challenges
Geometry
Probe technologies to support peripheral fine pitch probe of 23 µm peripheral staggered pad probes at effective pitches of 20/40, and fine pitch (45 µm) for dual row, non-staggered probing on all four die sides.
Fine pitch vertical probe technologies to support 130 µm pitch area array solder bump and 50 µm pitch staggered pad devices.
Multi-site pad probing technologies with corner pitch capability below 125 µm.
Reduction of pad damage at probe commensurate with pad size reductions (or better).
Alternative probe technology for 75 µm on 150 µm pitch dense array (vertical probe; bumped device).
Increasing probe array planarity requirements in combination with increasing array size.
Parallel test Need a probe technology to handle the complexity of SoC devices while probing more than one
device.
Current probe technologies have I/O limitations for bumped device probes.
Probing at temperature Reduce effects on probes for non-ambient testing -50°C to 150°C; especially for fine-pitch devices.
For effects on Handlers and Probers, see that section.
Product
Probe technologies to direct probe on copper bond pads including various oxidation considerations.
Probe technologies for probing over active circuitry (including flip-chip).
Probe force
Reduce per pin force required for good contact resistance to lower total load for high pin count and multi DUT probe applications. Evaluation and reduction of probe force requirements to eliminate die damage, including interlayer dielectric damage with lo
A chuck motion model is required to minimize probe damage
Probe cleaning
Development of high temperature (85°C–150°C) in situ cleaning mediums/methods, particularly for fine pitch, multi-DUT, and non-traditional probes.
Reduction of cleaning requirements while maintaining electrical performance to increase lifetime.
A self cleaning probe card is required for fine pitch bumped pad devices
Cost and delivery
Fine pitch or high pin count probe cards are too expensive and take too long to build.
Time and cost to repair fine pitch or high pin count probe cards is very high.
The time between chip design completion (“tape-out”) and the availability of wafers to be probed is less than the time required to design and build a probe card in almost every probe technology except traditional cantilever.
Space transformer lead times are too long, thus causing some vertical probe technologies to have lengthy lead-times.
Probe metrology Tools are required that support fine pitch probe characterization and pad damage measurements.
Metrology correlation is needed for post repair test versus on-floor usage.
High power devices Probe technologies will need to incorporate thermal management features capable of handling
device power dissipations approaching 1000 Watts and the higher currents (≥ 1.5 amp) flowing through individual probe points.
Contact resistance Probe technologies that achieve contact resistance <.5 Ohms initially and throughout use are needed.
A method to measure contact resistance is needed. The traditional continuity test is insufficient to monitor contact resistance.
High frequency probing Traditional probe technologies do not have the necessary electrical bandwidth for higher frequency
devices. At the top end are RF devices, requiring up to 40 GHz.
document.xls
2007_TST14
Year of Production 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
DRAM ½ Pitch (nm) (contacted) 65 57 50 45 40 36 32 28 25 22 20 18 16 14
MPU and ASIC ProductsWirebond - inline pad pitch 40 35 35 30 30 25 25 25 25 25 25 25 25 25Bump - array pad pitch 130 130 120 120 120 110 110 100 100 95 95 90 90 85I/O Pad Size (µm) X Y X Y X Y X Y X Y X Y X Y X Y X Y X Y X Y X Y X Y X YWirebond 30 55 30 55 30 55 25 45 25 45 20 35 20 35 20 35 15 25 15 25 15 25 15 25 15 25 15 25Bump 65 65 65 65 60 60 60 60 60 60 55 55 55 55 50 50 50 50 45 45 45 45 45 45 45 45 40 40Scrub (% of pad) AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTH Offline DEPTH AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTHWirebond 25 50 25 50 25 50 20 40 20 40 20 40 20 40 20 40 20 40 20 40 20 40 20 40 20 40 20 40Bump 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30
2050 2400 2400 2400 2400 2400 2400 2400 2400 2400 2400 2400 2400 2400
Number of Probe Points /Touchdown - Asics 5000 6000 7500 7500 7500 9000 9000 9000 9000 9000 9000 9000 9000 9000
Number of Probe Points / Touchdown - MPU 20000 20000 20000 20000 20000 30000 30000 30000 30000 30000 30000 30000 30000 30000
Maximum Current (mA) Probe Tip Probe Tip Probe Tip Probe Tip Probe Tip Probe Tip Probe Tip Probe Tip Probe Tip Probe Tip Probe Tip Probe Tip Probe Tip Probe TipASIC 400 <.001 500 <.001 500 <.001 500 <.001 500 <.001 1000 <.001 1000 <.001 1000 <.001 1000 <.001 1000 <.001 1000 <.001 1000 <.001 1000 <.001 1000 <.001MPU 1000 <.001 1000 <.001 1200 <.001 1200 <.001 1500 <.001 1500 <.001 1500 <.001 1500 <.001 1500 <.001 1500 <.001 1500 <.001 1500 <.001 1500 <.001 1500 <.001Maximum Resistance (Ohm) Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series
<0.5 <3 <0.5 <3 <0.5 <3 <0.5 <3 <0.5 <3 <0.5 <3 <0.5 <3 <0.5 <3 <0.5 <3 <0.5 <3 <0.5 <3 <0.5 <3 <0.5 <3 <0.5 <3
Memory ProductsWirebond - inline pad pitch 75 75 70 70 65 65 60 60 55 55 50 50 50 50I/O Pad Size (µm) X Y X Y X Y X Y X Y X Y X Y X Y X Y X Y X Y X Y X Y X YWirebond 65 80 65 80 60 80 60 80 55 80 55 80 55 80 55 80 50 80 50 80 50 80 65 80 65 80 65 80Scrub (% of pad) AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTHWirebond 25 50 25 50 25 50 25 50 25 50 25 50 25 50 25 50 25 50 25 50 25 50 25 50 25 50 25 50
100% of wafer 100% of wafer 100% of wafer 100% of wafer 100% of wafer 100% of wafer 100% of wafer 100% of wafer 100% of wafer 100% of wafer 100% of wafer 100% of wafer 100% of wafer 100% of wafer
Number of Probe Points / Touchdown - Memory 20000 20000 25000 25000 30000 30000 30000 30000 30000 30000 30000 30000 30000 30000
Maximum Current (mA) Probe Tip Probe Tip Probe Tip Probe Tip Probe Tip Probe Tip Probe Tip Probe Tip Probe Tip Probe Tip Probe Tip Probe Tip Probe Tip Probe Tip200 <.001 200 <.001 200 <.001 200 <.001 250 <.001 250 <.001 250 <.001 250 <.001 250 <.001 250 <.001 250 <.001 250 <.001 250 <.001 250 <.001
Maximum Resistance (Ohm) Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series<0.5 <3 <0.5 <3 <0.5 <3 <0.5 <3 <0.5 <3 <0.5 <3 <0.5 <3 <0.5 <3 <0.5 <3 <0.5 <3 <0.5 <3 <0.5 <3 <0.5 <3 <0.5 <3
RF and Mixed Signal ProductsWirebond - inline pad pitch 40 35 35 30 30 25 25 25 25 25 25 25 25 25Bump - array pad pitch 130 130 120 120 120 110 110 100 100 95 95 90 90 85I/O Pad Size (µm) X Y X Y X Y X Y X Y X Y X Y X Y X Y X Y X Y X Y X Y X YWirebond 30 55 30 55 30 55 25 45 25 45 20 35 20 35 20 35 15 25 15 25 15 25 15 25 15 25 15 25Bump 65 65 65 65 60 60 60 60 60 60 55 55 55 55 50 50 50 50 45 45 45 45 45 45 45 45 40 40Scrub (% of pad) AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTH Offline DEPTH AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTH AREA DEPTHWirebond 25 50 25 50 25 50 20 40 20 40 20 40 20 40 20 40 20 40 20 40 20 40 20 40 20 40 20 40Bump 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30
1600 1600 1600 1600 1600 1600 1600 1600 1600 1600 1600 1600 1600 1600Number of Probe Points /Touchdown 680 680 680 680 680 680 680 680 680 680 680 680 680 680Maximum Resistance (Ohm) Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series Contact Series
<0.5 <3 <0.5 <3 <0.4 <3 <0.4 <3 <0.4 <3 <0.4 <3 <0.4 <3 <0.4 <3 <0.4 <3 <0.4 <3 <0.4 <3 <0.4 <3 <0.4 <3 <0.4 <3
Size of Probed Area (mm2)
DC Leakage
DC Leakage
DC Leakage
DC Leakage
DC Leakage
DC Leakage
DC Leakage
DC Leakage
DC Leakage
DC Leakage
DC Leakage
DC Leakage
DC Leakage
DC Leakage
Size of Probed Area (mm2)
DC Leakage
DC Leakage
DC Leakage
DC Leakage
DC Leakage
DC Leakage
DC Leakage
DC Leakage
DC Leakage
DC Leakage
DC Leakage
DC Leakage
DC Leakage
DC Leakage
Size of Probed Area (mm2)
document.xls
2007_TST15
2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022
TSOP – Flash (NAND) – Contact blade [1]
Commodity NAND Memory
Lead Pitch (mm) 0.4 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3Data rate (MT/s) 50 50 50 66 66 100 100 100 100 133 133 133 133 266 266 266Contact blade
Inductance (nH) 10-15 5-10 5-10 5-10 5-10 5-10 5-10 5-10 5-10 5-10 5-10 5-10 5-10 5-10 5-10 5-10Contact Stroke (mm) 0.3-0.5 0.2-0.3 0.2-0.3 0.2-0.3 0.2-0.3 0.2-0.3 0.2-0.3 0.2-0.3 0.2-0.3 0.2-0.3 0.2-0.3 0.2-0.3 0.2-0.3 0.2-0.3 0.2-0.3 0.2-0.3Contact force (N) 0.2-0.4 0.2-0.3 0.2-0.3 0.2-0.3 0.2-0.3 0.2-0.3 0.2-0.3 0.2-0.3 0.2-0.3 0.2-0.3 0.2-0.3 0.2-0.3 0.2-0.3 0.2-0.3 0.2-0.3 0.2-0.3Contact resistance (m ohm) 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30Slit width (mm) 0.22 0.17 0.17 0.17 0.17 0.17 0.17 0.17 0.17 0.17 0.17 0.17 0.17 0.17 0.17 0.17
BGA – DRAM – Spring Probe [2]
Commodity DRAM (Mass production)
Lead Pitch (mm) 0.65 0.65 0.65 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5DRAM RM GT/S 1.1 1.3 1.3 1.6 1.6 2.1 2.7 2.7 3.2 3.2 4.3 5.3 5.4 6.4 6.4 8.5Spring Probe
Inductance (nH) 1.5 1.5 1.5 1 1 1 0.5 0.5 0.3 0.3 0.3 0.2 0.2 0.15 0.15 0.15Contact Stroke (mm) 0.3 0.3 0.3 0.3 0.3 0.3 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2Contact force (N) <0.4 <0.4 <0.4 <0.3 <0.3 <0.3 <0.2 <0.2 <0.2 <0.2 <0.2 <0.2 <0.2 <0.2 <0.2 <0.2Contact resistance (m ohm) 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100
BGA – SoC – Spring Probe (50 ohm) [3]
Logic (High volume microprocessor)
Lead Pitch (mm) 0.8 0.8 0.8 0.65 0.65 0.65 0.65 0.65 0.65 0.5 0.5 0.5 0.5 0.5 0.5 0.5I/O data (GT/s) 6 6 12 12 12 12 15 15 15 20 20 20 40 40 40 40Spring Probe (50 ohm)
Impedance (ohm) 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50Contact Stroke (mm) 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3Contact force (N) <0.4 <0.4 <0.4 <0.3 <0.3 <0.3 <0.3 <0.3 <0.3 <0.2 <0.2 <0.2 <0.2 <0.2 <0.2 <0.2Contact resistance (m ohm) 100 70 70 50 50 50 50 50 50 50 50 50 50 50 50 50
BGA – SoC – Conductive Rubber [4] [5]
Logic (High volume microprocessor)
Lead Pitch (mm) 0.8 0.8 0.8 0.65 0.65 0.65 0.65 0.65 0.65 0.5 0.5 0.5 0.5 0.5 0.5 0.5I/O data (GT/s) 6 6 12 12 12 12 15 15 15 20 20 20 40 40 40 40Spring Probe (50 ohm)
Inductance (nH) 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 <0.1 <0.1 <0.1 <0.1Contact Stroke (mm) 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15Contact force (N) 0.2 0.2 0.2 0.15 0.15 0.5 0.15 0.15 0.15 0.1 0.1 0.1 0.1 0.1 0.1 0.1Contact resistance (m ohm) 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50Thickness (mm) 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5
Manufacturable solutions exist, and are being optimized
Manufacturable solutions are known
Interim solutions are known uManufacturable solutions are NOT known
Notes:
[1] For pitches less than 0.3mm contactor molding becomes difficult due to the thin wall thickness between pins.
[2] For higher performance, a shorter probe spring is required which shortens the contact stroke. In 2019, the contact stroke will be 0.2mm so the contact resistance will be unstable.
[3] The spring probe must be coaxial for high-speed test. 20GT/s cannot be supported with finer pitches.
[4] Ball height is expected to change over the roadmap but amount of change is not known.
[5] A contact stroke of 0.15mm was assumed with a 0.5mm rubber thickness. For high ball count devices the contact pressure has been lowered.
Table RFAMS1 RF and Analog Mixed-Signal CMOS Technology Requirements
Year of Production 2007 2008 2009 2010
Performance RF/Analog [1]
Supply voltage (V) [2] 1.2 1.1 1.1 1
2 1.9 1.6 1.5
Gate Length (nm) [2] 53 45 37 32
32 30 30 30
160 140 100 90
6 6 5 5
13 11 9 8
170 200 240 280
200 240 290 340
0.25 0.22 0.2 <0.2
Precision Analog/RF Driver [1]
Supply voltage (V) 2.5 2.5 2.5 1.8
5 5 5 3
Gate Length (nm) [10] 250 250 250 180
220 220 220 160
IS 1000 1000 1000 360
9 9 9 6
40 40 40 50
70 70 70 90
CMOS NFET [1 HP CMOS lag 2 yrs]
1.1 1.1 1.1 1
EOT: Equivalent Oxide Thickness (Å) [13] 12 11 11 9
IS 53 37 32 29
IS 170 240 280 310
IS 200 290 340 380
IS 2.1 1.8 1.7 1.6
IS 4.0 3.5 3.4 3.3
ADD 5.3 4.7 4.5 4.4
Manufacturable solutions exist, and are being optimized
Manufacturable solutions are known
Interim solutions are known ¿Manufacturable solutions are NOT known
Notes for Table RFAMS1a and b:
Tox (nm) [2]
gm/gds at 5·Lmin-digital [3]
1/f-noise (µV²·µm²/Hz) [4]
s Vth matching (mV·µm) [5]
Ids (µA/µm) [6]
Peak Ft (GHz) [7]
Peak Fmax (GHz) [8]
NFmin (dB) [9]
Tox (nm) [10]
gm/gds at 10·Lmin-digital [11]
1/f Noise (µV²·µm²/Hz) [4]
s Vth matching (mV·µm) [5]
Peak Ft (GHz) [7]
Peak Fmax (GHz) [8]
Vdd: Power Supply Voltage (V) [13]
Lg: Physical Lgate for High Performance logic (nm) [13]
Peak Ft (GHz) [7]
Peak Fmax (GHz) [8]
NFmin (dB) at 24GHz[14]
NFmin (dB) at 60GHz[14]
NF min (dB) at 94GHz[14]
ORTCINDEX
2007 ITRS Chapters
2008INDEX
[9] This is the minimum transistor noise figure at 5GHz. 0.2dB represents the limitation of commercially available measurement equipment.
[14] This is the minimum transistor noise figure at 24 and 60GHz.
[1] Year of first digital product for a given technology generation as given in overall roadmap technology characteristics (ORTC) tables. Lithographic drivers for key technologies are indicated. Year of first RF and mixed-signal product at the same technology lag the low-standby power roadmap by one year. Beyond Planar CMOS, performance RF/Analog CMOS reflect DG CMOS, Precision Analog/RF driver device color change to yellow reflecting uncertainty on device integration. The supply voltage, Tox, Gate Length and Ids, Ft, Fmax color codes reflected the low-standby power roadmap. Any discrepancies, please refer to those of low-standby power roadmap.
[2] Nominal supply voltage, Vdd, SiO2 equivalent physical CMOS gate dielectric thickness, Tox, and minimum nominal gate length from low-standby power digital roadmap. For simplicity, only the Extended planar and DG technology options were used and the value was interpolated in the transition years.
[3] Measure for the low frequency amplification of a 5X minimum length, low-standby power CMOS transistor. Using different lengths is an extra degree of freedom in mixed signal designs. Long devices have better Gds amplification (at low frequencies). Operation point taken at 200 mV above the threshold voltage, V th, and at Vds = Vdd/2. The minimum value of 30 exceeds the projected technology capability with continued scaling for the standard logic device. When this occurs, the standard logic device should be replaced with an unique device designed for specifically for superior gain.
[4] Gate-referred 1/f noise spectral density, at a frequency of 1 Hz, normalized to an active gate area of 1 µm2. Operation point taken at 200 mV above the threshold voltage, Vth, and at Vds = Vdd/2.
[5] Matching specification for the NMOS transistor’s threshold voltage, assuming “near neighbor” devices at minimum practical separation. Careful layout and photolithographic uniformity, e.g. by using dummy structures, are required. Statistical dopant fluctuations start limiting further improvement with SiO 2. Matching behavior of high-k gate dielectrics very may be problematic. This parameter determines the lower boundary for the size of transistor in a mixed-signal circuit for a given accuracy and will limit dimensional, performance, and DC power consumption.
[6] Ids for Ft of 50 GHz for a minimum transistor length. Ft of 50 GHz is chosen for being 10X the application frequency for 5 GHz. An application frequency of 5 GHz is chosen as a mid-point for the frequency range of interest (1–10 GHz).
[7] Peak Ft measured from H21 extrapolated from 40 GHz with a 20 dB/dec slope.
[8] Peak Fmax measured from unilateral gain extrapolated from 40 GHz with a 20 dB/dec slope.
[10] This device is required to achieve direct modulation of the PA for applications from 2 to5 GHz and to support precision analog applications. Device with higher voltage tolerance are typically integrated with logic devices to support input-output interfaces. With continued scaling of logic devices alternate device structures may be required to support the required specifications.
[11] Measure for the low frequency amplification of a 10 minimum length, low-standby power CMOS transistor. Using different lengths is an extra degree of freedom in mixed signal designs. Long devices have better Gds amplification (at low frequencies). Operation point taken at 200 mV above the threshold voltage, V th, and at Vds = Vdd/2.
[12] Nominal supply voltage, Vdd, SiO2 equivalent electrical CMOS gate dielectric thickness, EOTelec and minimum nominal gate length from high-performance digital roadmap. For simplicity, only the Extended planar and DG technology options were used and the value was interpolated in the transition years.
[13] Nominal supply voltage, Vdd, SiO2 equivalent physical CMOS gate dielectric thickness, Tox, and minimum nominal gate length from high performance digital roadmap. For simplicity, only the Extended planar and DG technology options were used and the value was interpolated in the transition years.
2011 2012 2013 2014 2015 2016 2017 2018 2019
1 1 1 0.95 0.85 0.8 0.8 0.8 0.8
1.4 1.3 1.2 1.1 1.2 1.1 1.1 1 1
28 25 22 20 18 16 14 13 12
30 30 30 30 30 30 30 30 30
80 70 60 50 60 50 50 40 40
5 5 5 5 5 4 4 4 4
7 6 6 5 4 4 3 3 3
320 360 400 440 490 550 630 670 730
390 440 510 560 630 710 820 880 960
<0.2 <0.2 <0.2 <0.2 <0.2 <0.2 <0.2 <0.2 <0.2
1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.5
3 3 3 3 3 3 3 3 2.6
180 180 180 180 180 180 180 180 130
160 160 160 160 160 160 160 160 110
360 360 360 360 360 360 360 360 270
6 6 6 6 6 6 6 6 5
50 50 50 50 50 50 50 50 70
90 90 90 90 90 90 90 90 120
switch to DG device
1 1 0.95 0.9 0.9 0.9 0.8 0.8 0.7
7.5 6.5 5.5 5 6 6 6 5.5 5.5
27 24 22 20 18 17 15 14 12.8
330 370 400 440 490 520 590 630 680
410 460 510 560 630 670 760 820 900
1.6 1.5 1.4 1.4 1.3 1.3 1.2 1.2 1.1
3.2 3.0 3.0 2.9 2.7 2.7 2.6 2.5 2.4
4.3 4.1 4.0 3.9 3.8 3.7 3.5 3.5 3.4
Notes for Table RFAMS1a and b:
switch to DG device
[9] This is the minimum transistor noise figure at 5GHz. 0.2dB represents the limitation of commercially available measurement equipment.
[14] This is the minimum transistor noise figure at 24 and 60GHz.
[1] Year of first digital product for a given technology generation as given in overall roadmap technology characteristics (ORTC) tables. Lithographic drivers for key technologies are indicated. Year of first RF and mixed-signal product at the same technology lag the low-standby power roadmap by one year. Beyond Planar CMOS, performance RF/Analog CMOS reflect DG CMOS, Precision Analog/RF driver device color change to yellow reflecting uncertainty on device integration. The supply voltage, Tox, Gate Length and Ids, Ft, Fmax color codes reflected the low-standby power roadmap. Any discrepancies, please refer to those of low-standby power roadmap.
, and minimum nominal gate length from low-standby power digital roadmap. For simplicity, only the Extended planar and DG technology options were used and the value was interpolated in the transition years.
[3] Measure for the low frequency amplification of a 5X minimum length, low-standby power CMOS transistor. Using different lengths is an extra degree of freedom in mixed signal amplification (at low frequencies). Operation point taken at 200 mV above the threshold voltage, V th, and at Vds = Vdd/2. The minimum value of 30
exceeds the projected technology capability with continued scaling for the standard logic device. When this occurs, the standard logic device should be replaced with an unique device
. Operation point taken at 200 mV above the threshold voltage, Vth, and
[5] Matching specification for the NMOS transistor’s threshold voltage, assuming “near neighbor” devices at minimum practical separation. Careful layout and photolithographic uniformity, e.g. by using dummy structures, are required. Statistical dopant fluctuations start limiting further improvement with SiO 2. Matching behavior of high-k gate dielectrics very may be problematic. This parameter determines the lower boundary for the size of transistor in a mixed-signal circuit for a given accuracy and will limit dimensional, performance, and
of 50 GHz is chosen for being 10X the application frequency for 5 GHz. An application frequency of 5 GHz is chosen as a
[10] This device is required to achieve direct modulation of the PA for applications from 2 to5 GHz and to support precision analog applications. Device with higher voltage tolerance are typically integrated with logic devices to support input-output interfaces. With continued scaling of logic devices alternate device structures may be required to support the required
minimum length, low-standby power CMOS transistor. Using different lengths is an extra degree of freedom in mixed signal amplification (at low frequencies). Operation point taken at 200 mV above the threshold voltage, V th, and at Vds = Vdd/2.
equivalent electrical CMOS gate dielectric thickness, EOTelec and minimum nominal gate length from high-performance digital roadmap. For simplicity, only the Extended planar and DG technology options were used and the value was interpolated in the transition years.
, and minimum nominal gate length from high performance digital roadmap. For simplicity, only the Extended planar and DG technology options were used and the value was interpolated in the transition years.
2020 2021 2022
0.75 0.75 0.7
0.9 0.9 0.8
11 10 10
30 30 30
30 30 30
3 4 5
2 2 2
790 870 870
1050 1160 1160
<0.2 <0.2 <0.2
1.5 1.5 1.5
2.6 2.6 2.6
130 130 130
110 110 110
270 270 270 fixed error in scaling relationship
5 5 5
70 70 70
120 120 120
switch to DG device
0.7 0.7 0.65
5.5 5 5
11.7 10.7 9.7 following FEP, PIDS changes
740 810 890
990 1080 1200
1.1 1.0 1.0 matched published data and used more-physical scaling model
2.3 2.3 2.2 matched published data and used more-physical scaling model
3.2 3.1 3.0 added 94GH Nfmin starting in 2013
Table RFAMS2 RF and Analog Mixed-Signal Bipolar Technology Requirements
Year of Production 2007 2008 2009 2010
General Analog NPN Parameters
Emitter width (nm) (HS and HV NPN) 130 120 100 100
2 2 2 1.5
2 2 2 2
High Speed (HS) NPN (Common to mmWave Table)
IS 250 275 300 320
IS 280 305 330 350
Nfmin (dB) at 60GHz 3 2.5 2.2 1.9
BVceo (V) 1.8 1.7 1.65 1.6
13 15 17 18
High Voltage (HV) NPN
90 90 100 100
170 180 190 200
3.1 3.1 2.9 2.9
0.26 0.24 0.2 <0.2
28 22 16 15
Power Amplifier (PA) NPN (Common to PA Table)
IS 30 30 30 35
IS 80 80 80 80
IS Bvceo (V) 7 7 7 7
IS 17 17 17 17
Manufacturable solutions exist, and are being optimized
Manufacturable solutions are known
Interim solutions are known ¿Manufacturable solutions are NOT known
1/f-noise (µV²·µm²/Hz)
s current matching (%·µm)
Peak Ft (GHz)
Peak Fmax (GHz)
Jc at Peak Ft (mA/µm2)
Peak Ft (GHz) [Vcb=1V]
Peak Fmax (GHz)
BVceo
NFmin (dB) at 5GHz
Ic (µA/µm) at 50GHz Ft
Peak Ft (GHz) [Vcb=1V]
Peak Fmax (GHz)
BVcbo (V)
ORTCINDEX
2007 ITRS Chapters
2008INDEX
Table RFAMS2 RF and Analog Mixed-Signal Bipolar Technology Requirements
Year of Production
General Analog NPN Parameters
Emitter width (nm) (HS and HV NPN)
High Speed (HS) NPN (Common to mmWave Table)
Nfmin (dB) at 60GHz
BVceo (V)
High Voltage (HV) NPN
Power Amplifier (PA) NPN (Common to PA Table)
Bvceo (V)
Manufacturable solutions exist, and are being optimized
Manufacturable solutions are known
Interim solutions are known
Manufacturable solutions are NOT known
1/f-noise (µV²·µm²/Hz)
s current matching (%·µm)
Peak Ft (GHz)
Peak Fmax (GHz)
Jc at Peak Ft (mA/µm2)
Peak Ft (GHz) [Vcb=1V]
Peak Fmax (GHz)
BVceo
NFmin (dB) at 5GHz
Ic (µA/µm) at 50GHz Ft
Peak Ft (GHz) [Vcb=1V]
Peak Fmax (GHz)
BVcbo (V)
2011 2012 2013 2014 2015
100 90 90 90 80
1.5 1.5 1 1 1
2 2 2 2 2
340 360 380 395 415
370 390 410 425 445
1.7 1.5 1.4 1.3 1.2
1.55 1.5 1.45 1.4 1.35
19 21 22 23 24
110 110 120 120 130
210 220 230 240 250
2.8 2.8 2.6 2.6 2.5
<0.2 <0.2 <0.2 <0.2 <0.2
14 13 12 11 10
35 35 40 40 40
80 80 80 80 80
7 7 7 7 7
17 17 17 17 17
Table RFAMS2 RF and Analog Mixed-Signal Bipolar Technology Requirements
Year of Production
General Analog NPN Parameters
Emitter width (nm) (HS and HV NPN)
High Speed (HS) NPN (Common to mmWave Table)
Nfmin (dB) at 60GHz
BVceo (V)
High Voltage (HV) NPN
Power Amplifier (PA) NPN (Common to PA Table)
Bvceo (V)
Manufacturable solutions exist, and are being optimized
Manufacturable solutions are known
Interim solutions are known
Manufacturable solutions are NOT known
1/f-noise (µV²·µm²/Hz)
s current matching (%·µm)
Peak Ft (GHz)
Peak Fmax (GHz)
Jc at Peak Ft (mA/µm2)
Peak Ft (GHz) [Vcb=1V]
Peak Fmax (GHz)
BVceo
NFmin (dB) at 5GHz
Ic (µA/µm) at 50GHz Ft
Peak Ft (GHz) [Vcb=1V]
Peak Fmax (GHz)
BVcbo (V)
2016 2017 2018 2019 2020
80 80 70 70 70
1 1 1 1 1
2 2 2 2 2
430 445 455 470 480
460 475 485 500 510
1.1 1 1 0.9 0.9
1.35 1.3 1.3 1.3 1.3
25 26 27 28 29
130 140 140 150 150
260 270 280 290 300
2.5 2.4 2.4 2.4 2.4
<0.2 <0.2 <0.2 <0.2 <0.2
9 8 7 6 5
40 40 40 40 40
80 80 80 80 80
7 7 7 7 7
17 17 17 17 17
Table RFAMS2 RF and Analog Mixed-Signal Bipolar Technology Requirements
Year of Production
General Analog NPN Parameters
Emitter width (nm) (HS and HV NPN)
High Speed (HS) NPN (Common to mmWave Table)
Nfmin (dB) at 60GHz
BVceo (V)
High Voltage (HV) NPN
Power Amplifier (PA) NPN (Common to PA Table)
Bvceo (V)
Manufacturable solutions exist, and are being optimized
Manufacturable solutions are known
Interim solutions are known
Manufacturable solutions are NOT known
1/f-noise (µV²·µm²/Hz)
s current matching (%·µm)
Peak Ft (GHz)
Peak Fmax (GHz)
Jc at Peak Ft (mA/µm2)
Peak Ft (GHz) [Vcb=1V]
Peak Fmax (GHz)
BVceo
NFmin (dB) at 5GHz
Ic (µA/µm) at 50GHz Ft
Peak Ft (GHz) [Vcb=1V]
Peak Fmax (GHz)
BVcbo (V)
2021 2022
70 70
1 1
2 2
490 500 updated numbers, drop vcb defintion
520 530 updated numbers and colors0.9 0.8
1.25 1.25
29 30
160 160
310 320
2.3 2.3
<0.2 <0.2
5 5
40 40 updated numbers
80 80 updated numbers7 7 updated numbers
17 17 updated numbers
Table RFAMS2 RF and Analog Mixed-Signal Bipolar Technology Requirements
Year of Production
General Analog NPN Parameters
Emitter width (nm) (HS and HV NPN)
High Speed (HS) NPN (Common to mmWave Table)
Nfmin (dB) at 60GHz
BVceo (V)
High Voltage (HV) NPN
Power Amplifier (PA) NPN (Common to PA Table)
Bvceo (V)
Manufacturable solutions exist, and are being optimized
Manufacturable solutions are known
Interim solutions are known
Manufacturable solutions are NOT known
1/f-noise (µV²·µm²/Hz)
s current matching (%·µm)
Peak Ft (GHz)
Peak Fmax (GHz)
Jc at Peak Ft (mA/µm2)
Peak Ft (GHz) [Vcb=1V]
Peak Fmax (GHz)
BVceo
NFmin (dB) at 5GHz
Ic (µA/µm) at 50GHz Ft
Peak Ft (GHz) [Vcb=1V]
Peak Fmax (GHz)
BVcbo (V)
updated numbers, drop vcb defintion
Table RFAMS3 On-Chip Passives Technology Requirements
Year of Production 2007 2008 2009 2010
Analog
MOS Capacitor
7 7 7 11
<1e-9 <1e-9 <1e-9 <2e-6
Resistor
Thin Film BEOL
0.03 0.03 0.05 0.05
Temp. linearity (ppm/ºC) <100 <100 40-80 40-80
0.2 0.2 0.15 0.15
Sheet resistance, Rs (Ohm/sq) 50 50 50 50
P+ Polysilicon
0.1 0.1 0.1 0.1
Temp. linearity (ppm/ºC) <100 <100 40-80 40-80
1.7 1.7 1.7 1.7
Sheet resistance, Rs (Ohm/sq) 200–300 200–300 200–300 200–300
RF
Metal-Insulator-Metal Capacitor
IS 5 5 5 5
IS <100 <100 <100 < 100
IS <1e-8 <1e-8 <1e-8 <1e-8
IS 0.5 0.5 0.5 0.5
IS >50 >50 >50 >50
MOM Capacitor
3.7 5 5.3 6.2
Voltage linearity (ppm/V²) <100 <100 <100 <100
s Matching (% for 1pF) <0.15 <0.15 <0.15 <0.15
Inductor
29 30 32 34
MOS Varactor
Tuning Range [4] >5.5 >5.5 >5.5 >5.5
IS 35 40 40 45
PA
PA III-V Passives
IS Inductors Q (1GHz, 5nH) [5] 15 25 25 25
Capacitor Q [6] >100 >100 >100 >100
1.2 1.2 1.2 2
PA Silicon/SiGe Passives
IS Inductors Q (1GHz, 5nH) [5] 10 14 14 14
Capacitor Q [6] >100 >100 >100 >100
IS 2 2 2 2
Manufacturable solutions exist, and are being optimized
Manufacturable solutions are known
Interim solutions are known ¿Manufacturable solutions are NOT known
Notes for Table RFAMS3a and b:
Density (fF/µm²) [1]
Leakage (A/cm²) [8]
Parasitic capacitance (fF/µm²)
1s Matching (% µm)
Parasitic capacitance (fF/µm²)
1s Matching (% µm)
Density (fF/µm2) [2]
Voltage linearity (ppm/V²) Leakage (A/cm²) [9]
s Matching (%·µm)
Q (5 GHz for 1pF)
Density (fF/µm²)
Q (5 GHz, 1nH) [3]
Q (5 GHz, 0 V)
RF capacitor density (fF/µm2) [7]
RF capacitor density (fF/µm2) [7]
ORTCINDEX
2007 ITRS Chapters
2008INDEX
[1] This capacitance density corresponds to the highest end of the gate oxide thickness for precision analog device in the CMOS table.
[3] Q at 5 GHz for a single-ended 1nH inductor with a dedicated thick metal (analog metal).
[4] Defined as Cmax/Cmin in C-V curve of the varactor. Varactor align with performance RF device in the CMOS table.
[5] Inductor Q-quality factor of a 5nH inductor at 1 GHz achievable with the technology with a metallization suitable for handling the power requirements of the PA.
[6] Capacitor Q-quality factor of a 10 pF capacitor at 1 GHz achievable with the technology. Capacitor breakdown voltage must be rated for appropriate power amplification function.
[9] Leakage current is defined at room temperature and for the highest end of the supply voltage range for precision analog device in the CMOS table.
[2] No stacking (two capacitors on top of each other) is included. Coloring reflected MIM capacitor meeting all requirements including density, voltage linearity, leakage and matching on copper metallization.
[7] RF capacitor density-capacitor used for all other functions (matching, harmonic filtering, coupling, etc.). Capacitor must have adequate breakdown for the given application. No stacking.
[8] Leakage current is defined at room temperature and for the highest end of the supply voltage range and thickness end of the gate oxide thickness for precision analog device in the CMOS table.
2011 2012 2013 2014 2015 2016 2017 2018 2019
11 11 11 11 11 11 11 11 13
<2e-6 <2e-6 <2e-6 <2e-6 <2e-6 <2e-6 <2e-6 <2e-6 <2e-5
0.05 0.05 0.08 0.08 0.08 0.08 0.08 0.08 0.08
40-80 40-80 30 30 30 30 30 30 20
0.15 0.15 0.1 0.1 0.1 0.1 0.1 0.1 0.08
50 50 50 50 50 50 50 50 50
0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1
40-80 40-80 30 30 30 30 30 30 30
1.7 1.7 1 1 1 1 1 1 0.08
200–300 200–300 200–300 200–300 200–300 200–300 200–300 200–300 200–300
5 5 7 7 7 10 10 10 12
< 100 < 100 < 100 < 100 < 100 < 100 < 100 < 100 < 100
<1e-8 <1e-8 <1e-8 <1e-8 <1e-8 <1e-8 <1e-8 <1e-8 <1e-8
0.5 0.5 0.3 0.3 0.3 0.2 0.2 0.2 0.2
>50 >50 >50 >50 >50 >50 >50 >50 >50
7 6.5 7.5 8.6 9.9 11.4 13.1 15.1 17.4
<100 <100 <100 <100 <100 <100 <100 <100 <100
<0.15 <0.15 <0.1 <0.1 <0.1 <0.1 <0.1 <0.1 <0.08
36 38 40 42 44 46 48 50 52
>5.5 >5.5 >5.5 >5.5 >5.5 >5.5 >5.5 >5.5 >5.5
45 50 50 50 55 55 60 60 65
25 30 30 30 30 30 30 30 30
>100 >100 >100 >100 >100 >100 >100 >100 >100
2 2 2 2 2 2 2 2 2
14 18 18 18 18 18 18 18 18
>100 >100 >100 >100 >100 >100 >100 >100 >100
2 2 2 2 2 2 2 2 2
Notes for Table RFAMS3a and b:
[1] This capacitance density corresponds to the highest end of the gate oxide thickness for precision analog device in the CMOS table.
[3] Q at 5 GHz for a single-ended 1nH inductor with a dedicated thick metal (analog metal).
[4] Defined as Cmax/Cmin in C-V curve of the varactor. Varactor align with performance RF device in the CMOS table.
[5] Inductor Q-quality factor of a 5nH inductor at 1 GHz achievable with the technology with a metallization suitable for handling the power requirements of the PA.
[6] Capacitor Q-quality factor of a 10 pF capacitor at 1 GHz achievable with the technology. Capacitor breakdown voltage must be rated for appropriate power amplification function.
[9] Leakage current is defined at room temperature and for the highest end of the supply voltage range for precision analog device in the CMOS table.
[2] No stacking (two capacitors on top of each other) is included. Coloring reflected MIM capacitor meeting all requirements including density, voltage linearity, leakage and matching on
[7] RF capacitor density-capacitor used for all other functions (matching, harmonic filtering, coupling, etc.). Capacitor must have adequate breakdown for the given application. No
[8] Leakage current is defined at room temperature and for the highest end of the supply voltage range and thickness end of the gate oxide thickness for precision analog device in the
2020 2021 2022
13 13 13
<2e-5 <2e-5 <2e-5
0.08 0.08 0.08
20 20 20
0.08 0.08 0.08
50 50 50
0.1 0.1 0.1
30 20 20
0.08 0.08 0.08
200–300 200–300 200–300
12 12 12 updated numbers and colors
< 100 < 100 < 100 updated numbers and colors<1e-8 <1e-8 <1e-8 updated numbers and colors
0.2 0.2 0.2 updated numbers and colors>50 >50 >50 updated numbers and colors
20 23 26.4
<100 <100 <100
<0.08 <0.08 <0.08
54 56 58
>5.5 >5.5 >5.5
65 70 70 updated numbers and colors
30 30 30 updated colors>100 >100 >100
2 2 2
18 18 18 updated colors>100 >100 >100
2 2 2 updated numbers and colors, applications do not require the higher cap
updated numbers and colors, applications do not require the higher cap
Table RFAMS4 Embedded Passives Technology Requirements
Year of Production 2007 2008 2009
Resistor [1]
IS Max Sheet resistance, Rs (Ohm/sq) 1K 1K 1K
IS Tolerance (%) [2] <10% <10% <5%
IS Temp. linearity (ppm/ºC) <500 <300 <300
IS Min Sheet resistance, Rs (Ohm/sq) 100 100 100
IS Tolerance (%) [2] <10% <5% <3%
IS Temp. linearity (ppm/ºC) <300 <200 <200
Capacitor [3]
IS Density (nF/cm²) >1 >1 >5
Tolerance (%) [2] <10% <7% <10%
TCC (ppm) <500 <300 <500
Breakdown Voltage (V) >500V >1KV >300V
max Q [4] >25 >30 >25
Self Resonance Freq (GHz) [5] >0.5 >0.5 >0.1
Inductor [3]
Density (nH/mm2) 0.4 0.4 0.8
Tolerance (%) [2] <5% <5% <5%
max Q [6] >40 >40 >40
Self Resonance Freq (GHz) [7] >10 >10 >10
Manufacturable solutions exist, and are being optimized
Manufacturable solutions are known
Interim solutions are known ¿Manufacturable solutions are NOT known
Notes for Table RFAMS4a and b:
[1] For both thick and thin film process
[2] Untrimmed total tolerance including material and process
[3] For all material and process (lamination, buried, etc.)
[6] Maximum Q for 10nH inductor
[4] Maximum Q for 1cm2 capacitor
[5] SRF for 1cm2 capacitor
[7] SRF for 25mm2 inductor
ORTCINDEX
2007 ITRS Chapters
2008INDEX
2010 2011 2012 2013 2014 2015 2016 2017 2018
10K 100K 100K 500K 500K 500K 500K 500K 500K
<10% <10% <5% <10% <10% <5% <5% <5% <5%
<500 <300 <300 <500 <300 <300 <300 <300 <300
100 10 10 10 10 5 5 5 5
<1% <10% <5% <3% <1% <10% <5% <3% <1%
<200 <300 <200 <200 <200 <300 <200 <200 <200
>5 >5 >5 >10 >10 >10 >100 >100 >100
<7% <7% <5% <10% <7% <5% <10% <7% <5%
<400 <400 <300 <500 <300 <300 <300 <300 <200
>500V >700V >1KV >500V >700V >1KV >500V >500V >700V
>30 >30 >30 >25 >25 >30 >15 >20 >25
>0.1 >0.2 >0.2 >0.05 >0.1 >0.1 >0.001 >0.005 >0.01
0.8 0.8 0.8 2 2 2 2 4 4
<5% <5% <5% <5% <5% <3% <3% <3% <3%
>40 >40 >40 >40 >40 >40 >45 >45 >45
>10 >10 >10 >10 >10 >10 >10 >10 >10
Notes for Table RFAMS4a and b:
[1] For both thick and thin film process
[2] Untrimmed total tolerance including material and process
[3] For all material and process (lamination, buried, etc.)
[6] Maximum Q for 10nH inductor
2019 2020 2021 2022
500K 500K 500K 500K
<5% <5% <3% <1%
<300 <300 <200 <200
1 1 1 1
<10% <5% <3% <1%
<300 <200 <200 <200
>1000 >1000 >1000 >1000
<10% <10% <7% <5%
<300 <200 <200 <200
>500V >500V >700V >1KV
>10 >15 >20 >25
>0.001 >0.001 >0.001 >0.001
4 4 8 8
<3% <3% <3% <3%
>45 >45 >45 >45
>10 >10 >10 >10
Table RFAMS5 Power Amplifier Technology Requirements
Year of Production 2007 2008 2009 2010
Nominal battery voltage 3.2 3.2 3.2 3.2
IS End-of-life battery voltage 2.85 2.85 2.4 2.4
PA product solutions Radio/Baseband SIP [2]
PA frequency (GHz)
III-V HBT transistor
45 45 55 55
25 25 18 18
Linear efficiency (%) [1] 52 52 55 55
2.5 2.5 2.2 2.2
0.32 0.3 0.28 0.28
III-V HBT integration
Bias Control MESFET
Power management [4] N/A
Switch [5] (by-pass) HEMT
Filter [6] N/A
III-V PHEMT transistor
45 45 75 75
20 20 16 16
Linear efficiency (%) [1] 55 55 58 58
4 4 3.5 3.5
0.28 0.25 0.24 0.24
III-V PHEMT integration
Power management [4] N/A
Switch [7] logic integration E/D pHEMT
Filter [6] N/A
Silicon MOSFET transistor
60 60 35 35
45 45 60 60
12 12 10 10
Linear efficiency (%) [1] 45 45 45 45
6 6 4.5 4.5
0.08 0.08 0.06 0.06
Silicon MOSFET integration
Power management [4] Yes
MEMS switch [5] NO Stack Above IC
MEMS filter [6] Stack WLP
SiGe HBT transistor [9]
60 60 80 80
IS 17 17 17 17
Linear efficiency (%) [1] 50 50 52 52
2.5 2.5 2.2 2.2
0.12 0.12 0.11 0.11
SiGe integration
Power management Yes
Fmax (at Vcc) (GHz)
BVCBO (V)
Area (mm2) [2]
Cost/mm2 (US$) [3]
Fmax (at Vdd) (GHz)
BVDGO (V)
PA Area (mm2) [2]
Cost/mm2 (US$) [3]
Tox (PA) (Å) [8]
Fmax (at Vdd)
BVDSS (V)
PA Area (mm2) [2]
Cost/mm2 (US$) [3]
Fmax (GHz)
BVCBO (V)
PA Area (mm2) [2]
Cost/mm2 (US$) [3]
ORTCINDEX
2007 ITRS Chapters
2008INDEX
MEMS switch [5] NO Stack Above IC
MEMS filter [6] Stack WLP
Manufacturable solutions exist, and are being optimized
Manufacturable solutions are known
Interim solutions are known ¿Manufacturable solutions are NOT known
Notes for Table RFAMS5a and b:
[1] Linear efficiency—power added efficiency of the final PA stage under personal communication service (PCS) CDMA (IS-95) modulation.
[4] Power management—capability of the technology to provide RF power detection/DC power management for the PA.
[5] Switch—capability of the technology to integrate cost-effectively a stage by-pass switch into the PA active die.
[7] Switch Logic Integration—capability of the technology to integrate cost-effectively a control circuitry with the Tx/Rx antenna switch.
[9] Ideally, the Si requirements and GaAs requirements would be the same, but we use different values to account for the state-of-the-art performance differences between the technologies.
[2] Area—total semiconductor area necessary for the implementation of the quad-band GSM/general packet radio service (GPRS)/ Enhanced Data rates for GSM Evolution (EDGE) PA function, including matching/filtering.
[3] Cost/mm2—approximate commercial foundry cost of the area mentioned in [4].
[6] Filter—capability of the technology to integrate high-quality band selection filters needed for the assumed PA solution; currently performed with surface acoustic wave (SAW) filter technology.
[8] Tox (PA)—thickness of the MOSFET transistor in the RF power amplifier function.
2011 2012 2013 2014 2015 2016 2017 2018 2019
2.4 2.4 2.4 2.4 2.4 2.4 2.4 2.4 2.4
1.6 1.6 1.6 1.6 1.6 1.6 1.6 1.6 1.6
Radio/Baseband SIP [2] Radio/Baseband SIP [2]
55 65 65 65 65 65 65 65 65
18 18 18 18 18 18 18 18 18
55 55 55 55 55 55 55 55 55
2.2 2.2 2.2 2.2 2.2 2.2 2 2 2
0.28 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25
MESFET MESFET
N/A N/A
HEMT HEMT
N/A N/A
75 75 75 75 75 75 75 75 75
16 16 16 16 16 16 16 16 16
58 58 58 58 58 58 58 58 58
3.5 3.5 3.5 3.5 3.5 3.5 3.5 3.5 3.5
0.24 0.22 0.22 0.22 0.15 0.15 0.15 0.15 0.15
N/A N/A
E/D pHEMT E/D pHEMT
N/A N/A
35 35 35 35 35 35 35 35 35
60 60 60 60 60 60 60 60 60
10 10 10 10 10 10 10 10 10
45 45 45 45 45 45 45 45 45
4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5
0.06 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05
Yes Yes
Above IC Integrated Integrated
WLP Above IC Above IC
80 80 80 80 80 80 80 80 80
17 17 17 17 17 17 17 17 17
52 52 52 52 52 52 52 52 52
2.2 2.2 2.2 2.2 2.2 2.2 2 2 2
0.11 0.11 0.11 0.11 0.11 0.11 0.11 0.11 0.11
Yes Yes
0.8-6
Above IC Integrated Integrated
WLP Above IC Above IC
[1] Linear efficiency—power added efficiency of the final PA stage under personal communication service (PCS) CDMA (IS-95) modulation.
[4] Power management—capability of the technology to provide RF power detection/DC power management for the PA.
[5] Switch—capability of the technology to integrate cost-effectively a stage by-pass switch into the PA active die.
[7] Switch Logic Integration—capability of the technology to integrate cost-effectively a control circuitry with the Tx/Rx antenna switch.
[9] Ideally, the Si requirements and GaAs requirements would be the same, but we use different values to account for the state-of-the-art performance differences between the technologies.
[2] Area—total semiconductor area necessary for the implementation of the quad-band GSM/general packet radio service (GPRS)/ Enhanced Data rates for GSM Evolution (EDGE) PA
[6] Filter—capability of the technology to integrate high-quality band selection filters needed for the assumed PA solution; currently performed with surface acoustic wave (SAW) filter
2020 2021 2022
2.4 2.4 2.4
1.6 1.6 1.6 Delay 2.4V introRadio/Baseband SIP [2]
65 65 65
18 18 18
55 55 55
2 2 2
0.25 0.25 0.25
MESFET
N/A
HEMT
N/A
75 75 75
16 16 16
58 58 58
3.5 3.5 3.5
0.15 0.15 0.15
N/A
E/D pHEMT
N/A
35 35 35
60 60 60
10 10 10
45 45 45
4.5 4.5 4.5
0.05 0.05 0.05
Yes
Integrated
Above IC
80 80 80
17 17 17 updated numbers52 52 52
2 2 2
0.11 0.11 0.11
Yes
Integrated
Above IC
Table RFAMS6 Base Station Devices Technology Requirements
Year of Production 2007 2008 2009 2010
Application frequency (GHz) [1] 0.8–3.5 0.8–3.5 0.8–3.5 0.8–5
0.3 0.2 0.2 0.15
Packaging (C-Ceramic, P-Plastic) C, P C, P C, P Plastic
Si LDMOS
32, 48 32, 48 32, 48
240 300 400 500
1.8 1.8 1.8
55 57 60 55
120 150 200 250
39 40 42 39
GaAs FET
Operating voltage (V) 28 28
Saturated power (Watt) 240 240
Saturated power density (W/mm) 1.5 1.8
Saturated PAE (%) 65 67 70 65
Linear power (Watt) 120 120
Linear PAE (%) 46 47 50 46
GaN FET
Operating voltage (V) 48 48
Saturated power (Watt) 200 300
Saturated power density (W/mm) 4 5
Saturated PAE (%) 60 62 65 60
Manufacturable solutions exist, and are being optimized
Manufacturable solutions are known
Interim solutions are known ¿Manufacturable solutions are NOT known
Note for Table RFAMS6a and b:
[1] Application frequencies affected device saturated PAE scaling.
Cost ($$/Watt)
Operating voltage (V)
Saturated power (Watt)
Saturated power density (W/mm)
Saturated PAE (%)
Linear power (Watt)
Linear PAE (%)
ORTCINDEX
2007 ITRS Chapters
2008INDEX
2011 2012 2013 2014 2015 2016 2017 2018 2019
0.8–5 0.8–5 0.8–5 0.8–8 0.8–8 0.8–8
0.15 0.15 0.1 0.1 0.1 0.1
Plastic Plastic Plastic Plastic Plastic Plastic
32, 48 32, 48
500 500
1.8 1.8
57 60 60
250 250
40 42 42
28 28 28
240 240 240
1.8 1.8 1.8
67 70 70 70 72
120 120 120
47 50 50 50 51
48 48 48
400 500 500 500
5 5 5
62 65 65 65
2020 2021 2022
0.8–8
0.1
Plastic
32, 48
500
1.8
60
250
42
28
240
1.8
72
120
51
48
500
5
65
Year of Production 2007 2008 2009 2010
Device Technology—FET
GaAs PHEMT (low noise)
IS Gate length (nm) 150 150 150 100
IS 110 110 110 150
IS Breakdown (volts) 12 12 12 12
IS 650 650 650 700
IS 0.5 0.5 0.5 0.55
IS 1 1 1 0.8
IS 8.5 8.5 8.5 10.8
IS 2.8 2.8 2.8 2.5
IS 3 3 3 4
GaAs PHEMT (power)
IS Gate length (nm) 150 150 150 150
IS 150 150 150 150
IS Breakdown (volts) 12 12 12 12
IS 700 700 700 700
IS 0.5 0.5 0.5 0.5
IS 650 650 650 650
IS 45 45 45 45
IS 11 11 11 11
GaAs PHEMT (power)
IS Gate length (nm) 100 100 100 100
IS 200 200 200 200
IS Breakdown (volts) 8 8 8 8
IS 800 800 800 800
IS 0.65 0.65 0.65 0.65
IS 550 550 550 550
IS 30 30 30 30
IS 7 7 7 7
IS 350 350 350 350
IS 20 20 20 20
IS 5 5 5 5
InP HEMT (low noise)
Gate length (nm) 100 100 100 70
IS 200 200 200 250
IS Breakdown (volts) 4 4 4 3
IS 500 500 500 600
IS 1.1 1.1 1.1 1.5
IS 0.5 0.5 0.5 0.4
IS 15 15 15 16
Table RFAMS7 Millimeter Wave 10 GHz–100 GHz Technology Requirements
Ft (GHz)
Imax (mA/mm)
Gm (S/mm)
NFmin (dB) at 26 GHz
Associated Gain at 26 GHz
NFmin (dB) at 94 GHz
Associated Gain at 94 GHz
Fmax (GHz)
Imax (ma/mm)
Gm (S/mm)
Pout at 24 GHz and peak efficiency (mW/mm)
Peak efficiency at 24 GHz (%)
Gain at 24 GHz, at P1dB (dB)
Fmax (GHz)
Imax (ma/mm)
Gm (S/mm)
Pout at 60 GHz and peak efficiency (mW/mm)
Peak efficiency at 60 GHz (%)
Gain at 60 GHz, at P1dB (dB)
Pout at 94 GHz and peak efficiency (mW/mm)
Peak efficiency at 94 GHz (%)
Gain at 94 GHz, at P1dB (dB)
Ft (GHz)
Imax (ma/mm)
Gm (S/mm)
Fmin (dB) at 24 GHz
Associated Gain (dB) at 24 GHz
ORTCINDEX
2007 ITRS Chapters
2008INDEX
IS 1 1 1 0.8
IS 11 11 11 12
IS 1.5 1.5 1.5 1.3
IS 8 8 8 9
InP HEMT (power)
IS Gate length (nm) 100 100 100 70
IS 250 250 250 400
IS Breakdown (volts) 4 4 4 3
IS 500 500 500 600
IS 1.1 1.1 1.1 1.5
IS 450 450 450
IS 50 50 50
IS 14 14 14
IS 300 300 300 400
IS 40 40 45 50
IS 10 10 10 14
IS 150 150 160 200
IS 30 30 30 40
IS 7 7 7 10
GaAs MHEMT (low noise)—Ka through W-Band
IS Gate length (nm) 100 100 100 70
IS 200 200 200 250
IS Channel In content (%) 60 60 60 70
IS Offstate Breakdown (volts) 6 6 6 4
IS 900 900 900 900
IS 1.2 1.2 1.2 1.4
IS 0.5 0.5 0.5 0.4
IS Associated Gain (dB) @ 24GHz 15 15 15 16
IS 1 1 1 0.7
IS Associated Gain (dB) @ 60GHz 10 10 10 12
IS 1.5 1.5 1.5 1.2
IS Associated Gain (dB) @ 94GHz 8 8 8 10
GaAs MHEMT (Power) -Ka band
IS Gate length (nm) 150 150 150 100
IS Channel In content (%) 35 35 35 35
IS 200 200 200 250
IS Offstate Breakdown (volts) 8 8 8 10
IS 760 760 760 850
IS 0.85 0.85 0.85 1
IS 800 800 800 850
IS 45 45 45 50
IS 12 12 12 14
GaAs MHEMT (Power)
IS Gate length (nm) 100 100 100 70
IS Channel In content (%) 53 53 53 43
IS 300 300 300 300
Fmin (dB) at 60 GHz
Associated Gain (dB) at 60 GHz
Fmin (dB) at 94 GHz
Associated Gain (dB) at 94 GHz
Fmax (GHz)
Imax (ma/mm)
Gm (S/mm)
Pout at 24 GHz and peak efficiency (mW/mm)
Peak efficiency at 24 GHz (%)
Gain at 24 GHz, at P1dB (dB)
Pout at 60 GHz and peak efficiency (mW/mm)
Peak efficiency at 60 GHz (%)
Gain at 60 GHz, at P1dB (dB)
Pout at 94 GHz and peak efficiency (mW/mm)
Peak efficiency at 94 GHz (%)
Gain at 94 GHz, at P1dB (dB)
Ft (GHz)
Imax (ma/mm)
Gm (S/mm)
Fmin (dB) at 24 GHz
Fmin (dB) at 60 GHz
Fmin (dB) at 94 GHz
Fmax (GHz)
Imax (ma/mm)
Gm (S/mm)
Pout at 24 GHz and peak efficiency (mW/mm)
Peak efficiency at 24 GHz (%)
Gain at 24 GHz, at P1dB (dB)
Fmax (GHz)
IS Offstate Breakdown (volts) 7 7 7 9
IS 900 900 900 900
IS 1.2 1.2 1.2 1.4
IS 500 500 500 550
IS 40 40 40 45
IS 8 8 8 9
IS 225 225 225 300
IS 30 30 30 35
IS 6 6 6 7
GaN HEMT (low noise)
Gate Length (nm) 150 100
120 160
Breakdown (volts) 40 35
1000 1200
0.4 0.5
1.2 1
Associated Gain at 24 GHz 10 12
GaN HEMT (power)
Gate Length (nm) 150
150
Breakdown (volts) 60
1200
0.5
5000
IS 50
10
GaN HEMT (power)
Gate length (nm) 100
200
Breakdown (volts) 40
1200
0.55
4000
30
8
2500
20
6
Device Technology—RF CMOS
CMOS NFET [1 HP CMOS lag 2 yrs]
1.1 1.1 1.1 1
EOT: Equivalent Oxide Thickness (Å) [13] 12 11 11 9
IS 53 37 32 29
IS 170 240 280 310
IS 200 290 340 380
IS 2.1 1.8 1.7 1.6
IS 4.0 3.5 3.4 3.3
Imax (ma/mm)
Gm (S/mm)
Pout at 60 GHz and peak efficiency (mW/mm)
Peak efficiency at 60 GHz (%)
Gain at 60 GHz, at P1dB (dB)
Pout at 94 GHz and peak efficiency (mW/mm)
Peak efficiency at 94 GHz (%)
Gain at 94 GHz, at P1dB (dB)
Ft (GHz)
Imax (ma/mm)
Gm (S/mm)
Fmin (dB) at 24 GHz
Fmax (GHz)
Imax (ma/mm)
Gm (S/mm)
Pout at 24 GHz and peak efficiency (mW/mm)
Peak efficiency at 24 GHz (%)
Gain at 24 GHz, at P1dB (dB)
Fmax (GHz)
Imax (ma/mm)
Gm (S/mm)
Pout at 60 GHz and peak efficiency (mW/mm)
Peak efficiency at 60 GHz (%)
Gain at 60 GHz, at P1dB (dB)
Pout at 94 GHz and peak efficiency (mW/mm)
Peak efficiency at 94 GHz (%)
Gain at 94 GHz, at P1dB (dB)
Vdd: Power Supply Voltage (V) [13]
Lg: Physical Lgate for High Performance logic (nm) [13]
Peak Ft (GHz) [7]
Peak Fmax (GHz) [8]
NFmin (dB) at 24GHz[14]
NFmin (dB) at 60GHz[14]
ADD 5.3 4.7 4.5 4.4
Device Technology—HBT
InP HBT
IS Emitter width (nm) 1000 1000 500 500
IS 150 150 320 320
IS 200 200 320 320
IS 8 8 5 5
IS 1 1 5 5
SiGe HBT
Emitter width (nm) 130 120 100 100
250 275 300 320
280 305 330 350
Nfmin (dB) at 60GHz 3 2.5 2.2 1.9
ADD Nfmin (dB) at 94GHz 4.4 3.8 3.2 2.9
1.8 1.7 1.65 1.6
13 15 17 18
Manufacturable solutions exist, and are being optimized
Manufacturable solutions are known
Interim solutions are known ¿Manufacturable solutions are NOT known
NF min (dB) at 94GHz[14]
Peak Ft (GHz)
Peak Fmax (GHz)
BVceo
Jc at Peak Ft (mA/µm2)
Peak Ft (GHz)
Peak Fmax (GHz)
BVceo
Jc at Peak Ft (mA/µm2)
2011 2012 2013 2014 2015 2016
100 100 100 delayed start of 100 nm until 2010
150 150 150 delayed start of 100 nm until 2010
12 12 12 delayed start of 100 nm until 2010
700 700 700 delayed start of 100 nm until 2010
0.55 0.55 0.55 delayed start of 100 nm until 2010
0.8 0.8 0.8 delayed start of 100 nm until 2010
10.8 10.8 10.8 delayed start of 100 nm until 2010
2.5 2.5 2.5 delayed start of 100 nm until 2010
4 4 4 delayed start of 100 nm until 2010
150 150 150 150 150 150 deleted 100 nm data; extended 150 through 2016
150 150 150 150 150 150 deleted 100 nm data; extended 150 through 2016
12 12 12 12 12 12 deleted 100 nm data; extended 150 through 2016
700 700 700 700 700 700 deleted 100 nm data; extended 150 through 2016
0.5 0.5 0.5 0.5 0.5 0.5 deleted 100 nm data; extended 150 through 2016
650 650 650 650 650 650 deleted 100 nm data; extended 150 through 2016
45 45 45 45 45 45 deleted 100 nm data; extended 150 through 2016
11 11 11 11 11 11 deleted 100 nm data; extended 150 through 2016
70 70 70 70 nm data shifted 2 years out
250 250 250 70 nm data shifted 2 years out
8 8 8 70 nm data shifted 2 years out
850 850 850 70 nm data shifted 2 years out
0.75 0.8 0.8 70 nm data shifted 2 years out
550 550 550 70 nm data shifted 2 years out
35 40 40 70 nm data shifted 2 years out
8 9 9 70 nm data shifted 2 years out
350 350 350 70 nm data shifted 2 years out
25 30 30 70 nm data shifted 2 years out
6 7 7 70 nm data shifted 2 years out
70 50 50 35 35 25 shift 2 yrs. right; color changed for 35 nm data
250 350 350 420 420 500 shift 2 yrs. rigth
3 2.5 2.5 2 2 1.5 shift 2 yrs. right; color changed for 35 nm data
600 550 550 500 500 500 shift 2 yrs. rigth
1.5 1.8 1.8 2 2 2.2 shift 2 yrs. right; color changed for 35 nm data
0.4 0.3 0.3 0.3 0.3 0.25 shift 2 yrs. right; color changed for 35 nm data
16 17 17 18 18 20 shift 2 yrs. right; color changed for 35 nm data
0.8 0.6 0.6 0.6 0.6 0.5 shift 2 yrs. right; color changed for 35 nm data
12 13 13 14 14 16 shift 2 yrs. right; color changed for 35 nm data
1.3 1.1 1.1 1 1 0.9 shift 2 yrs. right; color changed for 35 nm data
9 11 11 12 12 14 shift 2 yrs. right; color changed for 35 nm data
70 70 50 50 50 50 shifted right 1 yr. and color changed for first 2 years @ 50 nm
400 400 450 450 450 450 shifted right 1 yr. and color changed for first 2 years @ 50 nm
3 3 2.5 2.5 2.5 2.5 shifted right 1 yr. and color changed for first 2 years @ 50 nm
600 600 600 600 600 600 shifted right 1 yr. and color changed for first 2 years @ 50 nm
1.5 1.5 1.7 1.7 1.7 1.7 shifted right 1 yr. and color changed for first 2 years @ 50 nm
shifted right one year
shifted right one year
shifted right one year
400 400 shifted right one year
50 50 shifted right one year
14 14 shifted right one year
200 200 200 200 200 200 shifted right one year
40 40 40 40 40 40 shifted right one year
10 10 11 11 11 11 shifted right one year
70 70 50 50 50 35 no change in data but shifted one year to the right
250 250 350 350 350 420 no change in data but shifted one year to the right
70 70 70 70 70 70 no change in data but shifted one year to the right
4 4 3 3 3 2.5 no change in data but shifted one year to the right
900 900 950 950 950 950 no change in data but shifted one year to the right
1.4 1.4 1.5 1.5 1.5 1.8 no change in data but shifted one year to the right
0.4 0.4 0.3 0.3 0.3 0.2 no change in data but shifted one year to the right
16 16 17 17 17 18 no change in data but shifted one year to the right
0.7 0.7 0.6 0.6 0.6 0.4 no change in data but shifted one year to the right
12 12 14 14 14 15 no change in data but shifted one year to the right
1.2 1.2 1 1 1 0.8 no change in data but shifted one year to the right
10 10 12 12 12 13 no change in data but shifted one year to the right
100 100 70 70 70 70 no change in data but shifted one year to the right
35 35 35 35 35 35 no change in data but shifted one year to the right
250 250 300 300 300 300 no change in data but shifted one year to the right
10 10 9 9 9 9 no change in data but shifted one year to the right
850 850 900 900 900 900 no change in data but shifted one year to the right
1 1 1.2 1.2 1.2 1.2 no change in data but shifted one year to the right
850 850 900 900 900 900 no change in data but shifted one year to the right
50 50 55 55 55 55 no change in data but shifted one year to the right
14 14 15 15 15 15 no change in data but shifted one year to the right
70 70 50 50 50 50 no change in data but shifted one year to the right
43 43 35 35 35 35 no change in data but shifted one year to the right
300 300 325 325 325 325 no change in data but shifted one year to the right
9 9 9 9 9 9 no change in data but shifted one year to the right
900 900 950 950 950 950 no change in data but shifted one year to the right
1.4 1.4 1.5 1.5 1.5 1.5 no change in data but shifted one year to the right
550 550 600 600 600 600 no change in data but shifted one year to the right
45 45 55 55 55 55 no change in data but shifted one year to the right
9 9 10 10 10 10 no change in data but shifted one year to the right
300 300 350 350 350 350 no change in data but shifted one year to the right
35 35 45 45 50 50 no change in data but shifted one year to the right
7 7 8 8 9 9 no change in data but shifted one year to the right
100 70 70 50 50 50
160 200 200 240 240 240
35 30 30 25 25 25
1300 1400 1400 1400 1400 1400
0.55 0.6 0.6 0.65 0.65 0.65
1 0.8 0.8 0.6 0.6 0.6
12 13 13 14 14 14
150 150 150 150 150 150
150 150 150 150 150 150
80 80 80 80 80 80
1200 1400 1400 1400 1400 1400
0.5 0.6 0.6 0.6 0.6 0.6
6000 7000 8000 8000 8000 8000
50 50 50 50 50 50 data changed
10 12 12 12 12 12
100 70 70 70 50 50
200 240 240 240 280 280
60 40 60 60 40 40
1200 1500 1500 1500 1500 1500
0.55 0.65 0.65 0.65 0.65 0.65
4500 5000 5000 5000 4500 4500
30 35 35 35 40 40
8.5 9 9.5 9.5 10 10
3000 3500 4000 4000 3500 3500
25 30 30 30 35 35
6.5 7 7.5 7.5 8 8
1 1 0.95 0.9 0.9 0.9
7.5 6.5 5.5 5 6 6
27 24 22 20 18 17 following FEP, PIDS changes
330 370 400 440 490 520
410 460 510 560 630 670
1.6 1.5 1.4 1.4 1.3 1.3 matched published data and used more-physical scaling model
3.2 3.0 3.0 2.9 2.7 2.7 matched published data and used more-physical scaling model
4.3 4.1 4.0 3.9 3.8 3.7 added 94GH Nfmin starting in 2013
250 250 250 130 130 130 no change in data but shifted one year to the right
400 400 400 560 560 560 no change in data but shifted one year to the right
560 560 560 800 800 800 no change in data but shifted one year to the right
4 4 4 3 3 3 no change in data but shifted one year to the right
10 10 10 20 20 20 no change in data but shifted one year to the right
100 90 90 90 80 80
340 360 380 395 415 430
370 390 410 425 445 460
1.7 1.5 1.4 1.3 1.2 1.1
2.6 2.3 2.1 2.0 1.8 1.7
1.55 1.5 1.45 1.4 1.35 1.35
19 21 22 23 24 25
deleted 100 nm data; extended 150 through 2016
deleted 100 nm data; extended 150 through 2016
deleted 100 nm data; extended 150 through 2016
deleted 100 nm data; extended 150 through 2016
deleted 100 nm data; extended 150 through 2016
deleted 100 nm data; extended 150 through 2016
deleted 100 nm data; extended 150 through 2016
deleted 100 nm data; extended 150 through 2016
shift 2 yrs. right; color changed for 35 nm data
shift 2 yrs. right; color changed for 35 nm data
shift 2 yrs. right; color changed for 35 nm data
shift 2 yrs. right; color changed for 35 nm data
shift 2 yrs. right; color changed for 35 nm data
shift 2 yrs. right; color changed for 35 nm data
shift 2 yrs. right; color changed for 35 nm data
shift 2 yrs. right; color changed for 35 nm data
shift 2 yrs. right; color changed for 35 nm data
shifted right 1 yr. and color changed for first 2 years @ 50 nm
shifted right 1 yr. and color changed for first 2 years @ 50 nm
shifted right 1 yr. and color changed for first 2 years @ 50 nm
shifted right 1 yr. and color changed for first 2 years @ 50 nm
shifted right 1 yr. and color changed for first 2 years @ 50 nm
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
matched published data and used more-physical scaling model
matched published data and used more-physical scaling model
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
no change in data but shifted one year to the right
Table RFAMS8 RF and Analog Mixed-Signal RFMEMS
Year of Production 2007 2008 2009
Design Tools
BAW (0) Separate tools (1) IRFM, (2) CM (3) DF
Resonator (0) Separate tools (1) IRFM, (2) CM (3) DF
Switch—capacitive contact (0) Separate tools, (2) CM (1) IRFM
Switch—metal contact (0) Separate tools, (2) CM (1) IRFM
All MEMS devices (4) MEMS TCAD (4) MEMS TCAD
Packaging
BAW Die stacking. Wafer level package. Micro cavity package (CWS).
Resonator Stacked die
Switch—capacitive contact Above IC integration with CWS
Switch—metal contact Above IC integration with CWS
Performance Driver
BAW
Resonator Real Time Clock (32 kHz)
Switch—capacitive contact t Cellular Frontend (Tuning): 20:1 tuning ratio, 40V actuation
Switch—metal contact
Cost Driver
BAW Die size / package
Resonator MEMS processing cost Packaging
Switch—capacitive contact Processing cost. Die size / microcavity package. Test.
Switch—metal contact Process cost. Reliability / size / microcavity package. Test.
NOTES:Design Tools (0) Separate Tools - Mechanical and RF simulation tools not integrated. Manual integration with package and IC.Design Tools (1) IRFM - integrated RF and mechanical 3D simulation toolsDesign Tools (2) CM - physically based compact models i.e. simplified versions of IRFMDesign Tools (3) DF - Design flow e.g. circuit level simulation, that includes IRFM Design Tools (4) MEMS TCAD-like 3D process sequence simulation tool to simulate deposition, roughness, thermal, underetch, etc…Packaging CWS = cap wafer seal (die to wafer, or wafer to wafer)Packaging TFS = thin film seal (e.g. a thin film "pinch-off" technique)
F= 900MHz to 2.5GHz. size and cost ; TCF=-20
ppm/K; K2*Q=100
F= 900MHz to 5GHz. Testability improved. TCF= -5ppm; K2*Q=150
Clock oscillator (10-100MHz) multi-frequency per die.
Cellular Frontend (Tuning, T/R): Insertion loss <0.3dB, lifetime >1e10 cycles
ORTCINDEX
2007 ITRS Chapters
2008INDEX
2010 2011 2012 2013
(3) DF DF + MEMS TCAD
(3) DF
(3) DF
(3) DF
Wafer level package. Micro cavity package (CWS). Above IC integration and TFS
Integration into IC and TFS
Above IC integration with CWS Embedded integration into IC and TFS
Above IC integration with CWS Embedded integration into IC and TFS
Nano resonator for filter function (800MHz–2.5GHz)
Cellular Frontend (Tuning): 20:1 tuning ratio, 40V actuation Cellular Frontend (Tuning): 30:1 tuning ratio, low-voltage actuation
Die size / package Integration with semiconductor die
Integration with semiconductor die
Processing cost. Die size / microcavity package. Test. Integration with semiconductor die
Process cost. Reliability / size / microcavity package. Test. Integration with semiconductor die
Design Tools (0) Separate Tools - Mechanical and RF simulation tools not integrated. Manual integration with package and IC.
Design Tools (4) MEMS TCAD-like 3D process sequence simulation tool to simulate deposition, roughness, thermal, underetch, etc…
Coupled Resonator Filter (CRF) ≥ increase functionality (e.g., impedance match).
F= 900MHz to 10GHz. Built In Self Test (BIST) structure. Tunable filter? TCF= -1ppm; K2*Q=200
Clock oscillator (10-100MHz) multi-frequency per die.
Cellular Frontend (Tuning, T/R): Insertion loss <0.3dB, lifetime >1e10 cycles
Cellular Frontend (Tuning, T/R): Insertion loss <0.2dB, lifetime >1e11 cycles
Table PIDS1 Process Integration Difficult Challenges—Near-term YearsDifficult Challenges ≥ 22 nm Summary of Issues
Scaling planar bulk CMOS will face significant challenges due to the high channel doping required, band-to-band tunneling across the junction and gate-induced drain leakage (GIDL), random doping variations, and difficulty in adequately controlling short channel effects. Also, keeping parasitics, such as series source/drain resistance with very shallow extensions and fringing capacitance, within tolerable limits will be significant issues.
Implementation into manufacturing of new structures such as ultra-thin body fully depleted silicon-on-insulator (SOI) and multiple-gate (e.g., FinFET) MOSFETs is expected at some point. This implementation will be challenging, with numerous new and difficult issues. A particularly challenging issue is the control of the thickness and its variability for these ultra-thin MOSFETs, as well as control of parasitic series source/drain resistance for very thin regions.
2. With scaling, difficulties in inducing adequate strain for enhanced mobility.
With scaling, it is critically important to maintain (or even increase) the current significantly enhanced CMOS channel mobility attained by applying strain to the channel. However, the strain due to current process-induced strain techniques tends to decrease with scaling.
Multiple major changes are projected over the next seven years, such as.:
Material: high-κ gate dielectric, metal gate electrodes, lead-free solder
Process: elevated S/D (selective epi) and advanced annealing and doping techniques
Structure: ultra-thin body (UTB) fully depleted (FD) SOI, multiple-gate MOSFETs, multi-chip package modules
It will be an important challenge to ensure the reliability of all these new materials, processes, and structures in a timely manner.
DRAM main issues with scaling—adequate storage capacitance for devices with reduced feature size, including difficulties in implementing high-κ storage dielectrics; access device design; holding the overall leakage to acceptably low levels; and deploying low sheet resistance materials for bit and word lines to ensure desired speed for scaled DRAMs.
SRAM—Difficulties with maintaining adequate noise margin and controlling key instabilities and soft error rate with scaling. Also, difficult lithography and etch issues with scaling.
Flash—Non-scalability of tunnel dielectric and interpoly dielectric. Dielectric material properties and dimensional control are key issues.
FeRAM—Continued scaling of stack capacitor is quite challenging. Eventually, continued scaling in 1T1C configuration. Sensitivity to IC processing temperatures and conditions.
MRAM—Magnetic material properties and dimensional control. Sensitivity to IC processing temperatures and conditions
Difficult Challenges<22 nm Summary of Issues
Advanced non-classical CMOS (e.g., multiple-gate MOSFETs) with ultra-thin, lightly doped body will be needed to scale MOSFETs to 10 nm gate length and below effectively. Control of parasitic resistance and capacitance will be critical.
To attain adequate drive current for the highly scaled MOSFETs, quasi-ballistic operation with enhanced thermal velocity and injection at the source end appears to be needed. Eventually, nanowires, carbon nanotubes, or other high transport channel materials (e.g., germanium or III-V thin channels on silicon) may be needed.
7. Dealing with fluctuations and statistical process variations in sub-11 nm gate length MOSFETs
Fundamental issues of statistical fluctuations for sub-10 nm gate length MOSFETs are not completely understood, including the impact of quantum effects, line edge roughness, and width variation.
Dense, fast, low operating voltage non-volatile memory will become highly desirable
Increasing difficulty is expected in scaling DRAMs, especially scaling down the dielectric equivalent oxide thickness and attaining the very low leakage currents and power dissipation that will be required.
All of the existing forms of nonvolatile memory face limitations based on material properties. Success will hinge on finding and developing alternative materials and/or development of alternative emerging technologies.
See Emerging Research Devices section for more detail.
Eventually, it is projected that the performance of copper/low-κ interconnect will become inadequate to meet the speed and power dissipation goals of highly scaled ICs.
Solutions (optical, microwave/RF, etc.) are currently unclear.
For detail, refer to ITRS Interconnect chapter.
Will drive major changes in process, materials, device physics, design, etc.
Performance, power dissipation, etc., of non-CMOS devices need to extend well beyond CMOS limits.
Non-CMOS devices need to integrate physically or functionally into a CMOS platform. Such integration may be difficult.
See Emerging Research Devices sections for more discussion and detail.
6. Implementation of advanced, non-classical CMOS with enhanced drive current and acceptable control of short channel effects for highly scaled MOSFETs
8. Identifying, selecting, and implementing new memory structures
9. Identifying, selecting, and implementing novel interconnect schemes
10. Eventually, identification, selection, and implementation of advanced, non-CMOS devices and architectures for advanced information processing
1. Scaling of MOSFETs to the 22 nm technology generation
3. Timely assurance for the reliability of multiple and rapid material, process, and structural changes
4. Scaling of DRAM and SRAM to the 22 nm technology generation
5. Scaling high-density non-volatile memory to the 22 nm technology generation
ORTCINDEX
2007 ITRS Chapters
2008INDEX
Table PIDS1 Process Integration Difficult Challenges—Near-term YearsDifficult Challenges ≥ 22 nm Summary of Issues
Scaling planar bulk CMOS will face significant challenges due to the high channel doping required, band-to-band tunneling across the junction and gate-induced drain leakage (GIDL), random doping variations, and difficulty in adequately controlling short channel effects. Also, keeping parasitics, such as series source/drain resistance with very shallow extensions and fringing capacitance, within tolerable limits will be significant issues.
Implementation into manufacturing of new structures such as ultra-thin body fully depleted silicon-on-insulator (SOI) and multiple-gate (e.g., FinFET) MOSFETs is expected at some point. This implementation will be challenging, with numerous new and difficult issues. A particularly challenging issue is the control of the thickness and its variability for these ultra-thin MOSFETs, as well as control of parasitic series source/drain resistance for very thin regions.
2. With scaling, difficulties in inducing adequate strain for enhanced mobility.
With scaling, it is critically important to maintain (or even increase) the current significantly enhanced CMOS channel mobility attained by applying strain to the channel. However, the strain due to current process-induced strain techniques tends to decrease with scaling.
Multiple major changes are projected over the next seven years, such as.:
Material: high-κ gate dielectric, metal gate electrodes, lead-free solder
Process: elevated S/D (selective epi) and advanced annealing and doping techniques
Structure: ultra-thin body (UTB) fully depleted (FD) SOI, multiple-gate MOSFETs, multi-chip package modules
It will be an important challenge to ensure the reliability of all these new materials, processes, and structures in a timely manner.
DRAM main issues with scaling—adequate storage capacitance for devices with reduced feature size, including difficulties in implementing high-κ storage dielectrics; access device design; holding the overall leakage to acceptably low levels; and deploying low sheet resistance materials for bit and word lines to ensure desired speed for scaled DRAMs.
SRAM—Difficulties with maintaining adequate noise margin and controlling key instabilities and soft error rate with scaling. Also, difficult lithography and etch issues with scaling.
Flash—Non-scalability of tunnel dielectric and interpoly dielectric. Dielectric material properties and dimensional control are key issues.
FeRAM—Continued scaling of stack capacitor is quite challenging. Eventually, continued scaling in 1T1C configuration. Sensitivity to IC processing temperatures and conditions.
MRAM—Magnetic material properties and dimensional control. Sensitivity to IC processing temperatures and conditions
Difficult Challenges<22 nm Summary of Issues
Advanced non-classical CMOS (e.g., multiple-gate MOSFETs) with ultra-thin, lightly doped body will be needed to scale MOSFETs to 10 nm gate length and below effectively. Control of parasitic resistance and capacitance will be critical.
To attain adequate drive current for the highly scaled MOSFETs, quasi-ballistic operation with enhanced thermal velocity and injection at the source end appears to be needed. Eventually, nanowires, carbon nanotubes, or other high transport channel materials (e.g., germanium or III-V thin channels on silicon) may be needed.
7. Dealing with fluctuations and statistical process variations in sub-11 nm gate length MOSFETs
Fundamental issues of statistical fluctuations for sub-10 nm gate length MOSFETs are not completely understood, including the impact of quantum effects, line edge roughness, and width variation.
Dense, fast, low operating voltage non-volatile memory will become highly desirable
Increasing difficulty is expected in scaling DRAMs, especially scaling down the dielectric equivalent oxide thickness and attaining the very low leakage currents and power dissipation that will be required.
All of the existing forms of nonvolatile memory face limitations based on material properties. Success will hinge on finding and developing alternative materials and/or development of alternative emerging technologies.
See Emerging Research Devices section for more detail.
Eventually, it is projected that the performance of copper/low-κ interconnect will become inadequate to meet the speed and power dissipation goals of highly scaled ICs.
Solutions (optical, microwave/RF, etc.) are currently unclear.
For detail, refer to ITRS Interconnect chapter.
Will drive major changes in process, materials, device physics, design, etc.
Performance, power dissipation, etc., of non-CMOS devices need to extend well beyond CMOS limits.
Non-CMOS devices need to integrate physically or functionally into a CMOS platform. Such integration may be difficult.
See Emerging Research Devices sections for more discussion and detail.
6. Implementation of advanced, non-classical CMOS with enhanced drive current and acceptable control of short channel effects for highly scaled MOSFETs
8. Identifying, selecting, and implementing new memory structures
9. Identifying, selecting, and implementing novel interconnect schemes
10. Eventually, identification, selection, and implementation of advanced, non-CMOS devices and architectures for advanced information processing
1. Scaling of MOSFETs to the 22 nm technology generation
3. Timely assurance for the reliability of multiple and rapid material, process, and structural changes
4. Scaling of DRAM and SRAM to the 22 nm technology generation
5. Scaling high-density non-volatile memory to the 22 nm technology generation
Table PIDS1 Process Integration Difficult Challenges—Near-term YearsDifficult Challenges ≥ 22 nm Summary of Issues
Scaling planar bulk CMOS will face significant challenges due to the high channel doping required, band-to-band tunneling across the junction and gate-induced drain leakage (GIDL), random doping variations, and difficulty in adequately controlling short channel effects. Also, keeping parasitics, such as series source/drain resistance with very shallow extensions and fringing capacitance, within tolerable limits will be significant issues.
Implementation into manufacturing of new structures such as ultra-thin body fully depleted silicon-on-insulator (SOI) and multiple-gate (e.g., FinFET) MOSFETs is expected at some point. This implementation will be challenging, with numerous new and difficult issues. A particularly challenging issue is the control of the thickness and its variability for these ultra-thin MOSFETs, as well as control of parasitic series source/drain resistance for very thin regions.
2. With scaling, difficulties in inducing adequate strain for enhanced mobility.
With scaling, it is critically important to maintain (or even increase) the current significantly enhanced CMOS channel mobility attained by applying strain to the channel. However, the strain due to current process-induced strain techniques tends to decrease with scaling.
Multiple major changes are projected over the next seven years, such as.:
Material: high-κ gate dielectric, metal gate electrodes, lead-free solder
Process: elevated S/D (selective epi) and advanced annealing and doping techniques
Structure: ultra-thin body (UTB) fully depleted (FD) SOI, multiple-gate MOSFETs, multi-chip package modules
It will be an important challenge to ensure the reliability of all these new materials, processes, and structures in a timely manner.
DRAM main issues with scaling—adequate storage capacitance for devices with reduced feature size, including difficulties in implementing high-κ storage dielectrics; access device design; holding the overall leakage to acceptably low levels; and deploying low sheet resistance materials for bit and word lines to ensure desired speed for scaled DRAMs.
SRAM—Difficulties with maintaining adequate noise margin and controlling key instabilities and soft error rate with scaling. Also, difficult lithography and etch issues with scaling.
Flash—Non-scalability of tunnel dielectric and interpoly dielectric. Dielectric material properties and dimensional control are key issues.
FeRAM—Continued scaling of stack capacitor is quite challenging. Eventually, continued scaling in 1T1C configuration. Sensitivity to IC processing temperatures and conditions.
MRAM—Magnetic material properties and dimensional control. Sensitivity to IC processing temperatures and conditions
Difficult Challenges<22 nm Summary of Issues
Advanced non-classical CMOS (e.g., multiple-gate MOSFETs) with ultra-thin, lightly doped body will be needed to scale MOSFETs to 10 nm gate length and below effectively. Control of parasitic resistance and capacitance will be critical.
To attain adequate drive current for the highly scaled MOSFETs, quasi-ballistic operation with enhanced thermal velocity and injection at the source end appears to be needed. Eventually, nanowires, carbon nanotubes, or other high transport channel materials (e.g., germanium or III-V thin channels on silicon) may be needed.
7. Dealing with fluctuations and statistical process variations in sub-11 nm gate length MOSFETs
Fundamental issues of statistical fluctuations for sub-10 nm gate length MOSFETs are not completely understood, including the impact of quantum effects, line edge roughness, and width variation.
Dense, fast, low operating voltage non-volatile memory will become highly desirable
Increasing difficulty is expected in scaling DRAMs, especially scaling down the dielectric equivalent oxide thickness and attaining the very low leakage currents and power dissipation that will be required.
All of the existing forms of nonvolatile memory face limitations based on material properties. Success will hinge on finding and developing alternative materials and/or development of alternative emerging technologies.
See Emerging Research Devices section for more detail.
Eventually, it is projected that the performance of copper/low-κ interconnect will become inadequate to meet the speed and power dissipation goals of highly scaled ICs.
Solutions (optical, microwave/RF, etc.) are currently unclear.
For detail, refer to ITRS Interconnect chapter.
Will drive major changes in process, materials, device physics, design, etc.
Performance, power dissipation, etc., of non-CMOS devices need to extend well beyond CMOS limits.
Non-CMOS devices need to integrate physically or functionally into a CMOS platform. Such integration may be difficult.
See Emerging Research Devices sections for more discussion and detail.
6. Implementation of advanced, non-classical CMOS with enhanced drive current and acceptable control of short channel effects for highly scaled MOSFETs
8. Identifying, selecting, and implementing new memory structures
9. Identifying, selecting, and implementing novel interconnect schemes
10. Eventually, identification, selection, and implementation of advanced, non-CMOS devices and architectures for advanced information processing
1. Scaling of MOSFETs to the 22 nm technology generation
3. Timely assurance for the reliability of multiple and rapid material, process, and structural changes
4. Scaling of DRAM and SRAM to the 22 nm technology generation
5. Scaling high-density non-volatile memory to the 22 nm technology generation
Table PIDS1 Process Integration Difficult Challenges—Near-term YearsDifficult Challenges ≥ 22 nm Summary of Issues
Scaling planar bulk CMOS will face significant challenges due to the high channel doping required, band-to-band tunneling across the junction and gate-induced drain leakage (GIDL), random doping variations, and difficulty in adequately controlling short channel effects. Also, keeping parasitics, such as series source/drain resistance with very shallow extensions and fringing capacitance, within tolerable limits will be significant issues.
Implementation into manufacturing of new structures such as ultra-thin body fully depleted silicon-on-insulator (SOI) and multiple-gate (e.g., FinFET) MOSFETs is expected at some point. This implementation will be challenging, with numerous new and difficult issues. A particularly challenging issue is the control of the thickness and its variability for these ultra-thin MOSFETs, as well as control of parasitic series source/drain resistance for very thin regions.
2. With scaling, difficulties in inducing adequate strain for enhanced mobility.
With scaling, it is critically important to maintain (or even increase) the current significantly enhanced CMOS channel mobility attained by applying strain to the channel. However, the strain due to current process-induced strain techniques tends to decrease with scaling.
Multiple major changes are projected over the next seven years, such as.:
Material: high-κ gate dielectric, metal gate electrodes, lead-free solder
Process: elevated S/D (selective epi) and advanced annealing and doping techniques
Structure: ultra-thin body (UTB) fully depleted (FD) SOI, multiple-gate MOSFETs, multi-chip package modules
It will be an important challenge to ensure the reliability of all these new materials, processes, and structures in a timely manner.
DRAM main issues with scaling—adequate storage capacitance for devices with reduced feature size, including difficulties in implementing high-κ storage dielectrics; access device design; holding the overall leakage to acceptably low levels; and deploying low sheet resistance materials for bit and word lines to ensure desired speed for scaled DRAMs.
SRAM—Difficulties with maintaining adequate noise margin and controlling key instabilities and soft error rate with scaling. Also, difficult lithography and etch issues with scaling.
Flash—Non-scalability of tunnel dielectric and interpoly dielectric. Dielectric material properties and dimensional control are key issues.
FeRAM—Continued scaling of stack capacitor is quite challenging. Eventually, continued scaling in 1T1C configuration. Sensitivity to IC processing temperatures and conditions.
MRAM—Magnetic material properties and dimensional control. Sensitivity to IC processing temperatures and conditions
Difficult Challenges<22 nm Summary of Issues
Advanced non-classical CMOS (e.g., multiple-gate MOSFETs) with ultra-thin, lightly doped body will be needed to scale MOSFETs to 10 nm gate length and below effectively. Control of parasitic resistance and capacitance will be critical.
To attain adequate drive current for the highly scaled MOSFETs, quasi-ballistic operation with enhanced thermal velocity and injection at the source end appears to be needed. Eventually, nanowires, carbon nanotubes, or other high transport channel materials (e.g., germanium or III-V thin channels on silicon) may be needed.
7. Dealing with fluctuations and statistical process variations in sub-11 nm gate length MOSFETs
Fundamental issues of statistical fluctuations for sub-10 nm gate length MOSFETs are not completely understood, including the impact of quantum effects, line edge roughness, and width variation.
Dense, fast, low operating voltage non-volatile memory will become highly desirable
Increasing difficulty is expected in scaling DRAMs, especially scaling down the dielectric equivalent oxide thickness and attaining the very low leakage currents and power dissipation that will be required.
All of the existing forms of nonvolatile memory face limitations based on material properties. Success will hinge on finding and developing alternative materials and/or development of alternative emerging technologies.
See Emerging Research Devices section for more detail.
Eventually, it is projected that the performance of copper/low-κ interconnect will become inadequate to meet the speed and power dissipation goals of highly scaled ICs.
Solutions (optical, microwave/RF, etc.) are currently unclear.
For detail, refer to ITRS Interconnect chapter.
Will drive major changes in process, materials, device physics, design, etc.
Performance, power dissipation, etc., of non-CMOS devices need to extend well beyond CMOS limits.
Non-CMOS devices need to integrate physically or functionally into a CMOS platform. Such integration may be difficult.
See Emerging Research Devices sections for more discussion and detail.
6. Implementation of advanced, non-classical CMOS with enhanced drive current and acceptable control of short channel effects for highly scaled MOSFETs
8. Identifying, selecting, and implementing new memory structures
9. Identifying, selecting, and implementing novel interconnect schemes
10. Eventually, identification, selection, and implementation of advanced, non-CMOS devices and architectures for advanced information processing
1. Scaling of MOSFETs to the 22 nm technology generation
3. Timely assurance for the reliability of multiple and rapid material, process, and structural changes
4. Scaling of DRAM and SRAM to the 22 nm technology generation
5. Scaling high-density non-volatile memory to the 22 nm technology generation
Table PIDS2 High-performance Logic Technology Requirements
Year of Production 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022
59 52 45 40 36 32 28 25 22 20 18 16 14 13 11
WAS 22 20 18 16 14 13 11 10 9 8 7 6 5.5 5 4.5
IS 29 27 24 22 20 18 17 15 14 12.8 11.7 10.7 9.7 8.9 8.1
EOT: Equivalent Oxide Thickness [2]
IS Extended planar bulk (Å) 12 10 9.5 8.8 7.5 6.5 6 5.3 5
IS UTB FD (Å) 7 6.5 6.8 5.5 5 5 5
IS DG (Å) 7.7 7 6 6 6 5.9 5.5 5.5
Gate Poly Depletion and Inversion-Layer Equivalent Thickness [3]
IS Extended Planar Bulk (Å) 7.4 3.3 3.2 3.1 2.9 2.8 2.75 2.65 2.6
IS UTB FD (Å) 4 4 4 4 4 4 4
IS DG (Å) 4 4 4 4 4 4 4 4
IS Extended Planar Bulk (Å) 19.4 13.3 12.7 11.9 10.4 9.3 8.75 7.95 7.6
IS UTB FD (Å) 11 10.5 9.8 9.5 9 9 9
IS DG (Å) 11.5 11 10 10 10 9.9 9.5 9.5
IS 450 650 830 900 1000 1100 1200 1300 1400
IS 1100 1200 1300 1400 1500 1700 1900
IS 1300 1400 1500 1700 1900 2100 2200 2500
IS Extended Planar Bulk (V) 1.1 1.1 1.07 1 1 1 0.975 0.925 0.9
IS UTB FD and DG (V) 1 1 0.95 0.9 0.9 0.9 0.9 0.8 0.8 0.7
IS Extended Planar Bulk (mV) 225 196 175 168 94 103 102 107 112
IS UTB FD (mV) 103 96 88 87 94 97 99
IS DG (mV) 110 105 104 106 109 111 110 109
IS Extended Planar Bulk (µA/µm) 0.13 0.17 0.46 0.71 0.7 0.64 0.69 0.71 0.68
IS UTB FD (µA/µm) 0.33 0.43 0.57 0.62 0.56 0.55 0.57
IS DG (µA/µm) 0.27 0.34 0.37 0.38 0.38 0.4 0.44 0.48
IS Extended Planar Bulk (µA/µm) 1006 1317 1370 1333 1639 1807 1816 1793 1762
IS UTB FD (µA/µm) 1948 1970 1970 1944 2123 2197 2181
IS DG (µA/µm) 1930 1943 2220 2309 2344 2395 2627 2533
IS 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8
IS Extended Planar Bulk 1.07 1.11 1.1 1.09 1.08 1.08 1.09 1.09 1.08
Grey cells delineate one of two time periods: either before initial production ramp has started for ultra-thin body fully depleted (UTB FD) SOI or double-gate (DG) MOSFETs, or beyond when planar bulk or UTB FD MOSFETs have reached the limits of practical scaling (see the text and the table notes for further discussion).
MPU/ASIC Metal 1 (M1) ½ Pitch (nm) (contacted)
Lg: Physical Lgate for High Performance logic (nm) [1]
Lg: Physical Lgate for High Performance logic (nm) [1]
EOTelec: Electrical Equivalent Oxide Thickness in inversion [4]
Jg,limit: Maximum gate leakage current density [5]
Extended Planar Bulk (A/cm2)
UTB FD (A/cm2)
DG (A/cm2)
Vdd: Power Supply Voltage (V) [6]
Vt,sat: Saturation Threshold Voltage [7]
Isd,leak: Source/Drain Subthreshold Off-State Leakage Current [8]
Id,sat: NMOS Drive Current [9]
Mobility enhancement factor due to strain [10]
Id,sat enhancement factor due to strain [11]
ORTCINDEX
2007 ITRS Chapters
IS UTB FD 1.07 1.07 1.06 1.06 1.06 1.05 1.05
IS DG 1.04 1.04 1.04 1.03 1.03 1.03 1.03 1.03
Effective Ballistic Enhancement Factor, Kbal [12]
IS Extended Planar Bulk 1 1 1 1 1 1 1 1 1
IS UTB FD 1.05 1.08 1.13 1.16 1.2 1.23 1.25
IS DG 1.21 1.25 1.32 1.35 1.42 1.57 1.67 1.87
IS Extended Planar Bulk (Ω-µm) 200 200 200 200 200 180 180 180 180
IS UTB FD (Ω-µm) 180 180 180 180 170 160 160
IS DG (Ω-µm) 180 180 170 160 160 160 160 150
IS Extended Planar Bulk (fF/µm) 0.515 0.700 0.652 0.637 0.663 0.670 0.671 0.652 0.633
IS UTB FD (F/µm) 0.565 0.559 0.530 0.508 0.490 0.449 0.410
IS DG (F/µm) 0.450 0.439 0.441 0.404 0.370 0.340 0.327 0.291
IS Extended Planar Bulk (fF/µm) 0.721 0.875 0.818 0.794 0.843 0.840 0.838 0.814 0.793
IS UTB FD (F/µm) 0.808 0.765 0.700 0.678 0.650 0.609 0.570
IS DG (F/µm) 0.640 0.629 0.621 0.583 0.550 0.520 0.507 0.481
τ =CV/I: NMOSFET intrinsic delay (ps) [16]
IS Extended Planar Bulk (ps) 0.79 0.73 0.64 0.60 0.51 0.46 0.45 0.42 0.4
IS UTB FD (ps) 0.41 0.39 0.34 0.31 0.28 0.25 0.22
IS DG (ps) 0.32 0.29 0.26 0.23 0.2 0.17 0.15 0.13
1/τ: NMOSFET intrinsic switching speed (GHz) [17]
IS Extended Planar Bulk (GHz) 1268 1368 1565 1679 1961 2174 2250 2413 2500
IS UTB FD (GHz) 2439 2610 3002 3226 3649 4076 4472
IS DG (GHz) 3195 3448 3938 4441 5000 5890 6667 7692
Manufacturable solutions exist, and are being optimized
Manufacturable solutions are known
Interim solutions are known ¿Manufacturable solutions are NOT known
Rsd: Effective Parasitic series source/drain resistance [13]
Cg,ideal: Ideal NMOS Device Gate Capacitance [14]
Cg,total: Total gate capacitance for calculation of CV/I [15]
Table PIDS3a and b Low Standby Power Technology Requirements
Year in Production 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022
59 52 45 40 36 32 28 25 22 20 18 16 14 13 11
29 27 24 22 20 18 17 15 14 12.8 11.7 10.7 9.7 8.9 8.1
IS Extended Planar Bulk and DG (nm) 38 32 29 27 22 18 17 15 14.0 12.8 11.7 10.7 9.7 8.9 8.1
IS UTB FD (nm) 20 18 17 16 15
EOT: Equivalent Oxide Thickness [2]
Extended planar bulk (Å) 16 15 14 13 12 11
IS UTB FD (Å) 12 11 10 9 8
IS DG (Å) 11 11 10 10 9 9 8 8
Gate Poly Depletion and Inversion-Layer Equivalent Thickness [3]
IS Extended planar bulk (Å) 6.0 3.4 3.3 3.2 3.1 3.1
IS UTB FD (Å) 4 4 4 4 4
IS DG (Å) 4 4 4 4 4 4 4 4
IS Extended planar bulk (Å) 22 18.4 17.3 16.2 15.1 14.1
IS UTB FD (Å) 16 15 14 13 12
IS DG (Å) 15 15 14 14 13 13 12 12
81 94 110 120 140 150
IS UTB FD (A/cm2) 150 170 180 190 200
IS DG (A/cm2) 190 210 230 250 270 300 330 380
Extended Planar Bulk (V) 1.1 1 1 1 1 0.95
IS UTB FD (V) 0.9 0.9 0.85 0.8 0.8
IS DG (V) 0.8 0.8 0.8 0.8 0.75 0.75 0.7 0.7
Extended Planar Bulk (mV) 567 535 535 544 552 547
IS UTB FD (mV) 399 401 404 404 405
IS DG (mV) 366 366 371 365 374 378 369 376
Extended Planar Bulk (pA/µm) 30.3 30.5 30.7 30.2 30.2 30.3
IS UTB FD (µA/µm) 30.9 31.7 30.2 31.0 32.7
IS DG (µA/µm) 26.5 29.7 25.5 33.8 26.2 23.9 33.8 28.9
IS Extended Planar Bulk (µA/µm) 499 501 528 542 560 519
IS UTB FD (µA/µm) 669 744 786 771 838
IS DG (µA/µm) 702 738 839 889 895 935 934 946
Mobility enhancement factor due to strain [10]
Extended Planar Bulk 1.8 1.8 1.8 1.8 1.8 1.8
IS UTB FD and DG 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8
Extended Planar Bulk 1.17 1.16 1.17 1.17 1.16 1.17
Grey cells delineate one of two time periods: either before initial production ramp has started for ultra-thin body fully depleted (UTB FD) SOI or double-gate (DG) MOSFETs, or beyond when planar bulk or UTB FD MOSFETs have reached the limits of practical scaling (see the text and the table notes for further discussion).
MPU/ASIC Metal 1 (M1) ½ Pitch (nm) (contacted)
Lg: Physical Lgate for High Performance logic (nm) [1]
Lg: Physical gate length for LSTP [1]
EOTelec: Electrical Equivalent Oxide Thickness in inversion [4]
Jg,limit: Maximum gate leakage current density [5]
Extended Planar Bulk (mA/cm2)
Vdd: Power Supply Voltage (V) [6]
Vt,sat: Saturation Threshold Voltage [7]
Isd,leak: Source/Drain Subthreshold Off-State Leakage Current [8]
Id,sat: NMOS Drive Current [9]
Id,sat enhancement factor due to strain [11]
ORTCINDEX
2007 ITRS Chapters
IS UTB FD 1.07 1.07 1.07 1.08 1.07
IS DG 1.05 1.04 1.04 1.04 1.04 1.04 1.04 1.04
Effective Ballistic Enhancement Factor [12]
Extended Planar Bulk 1 1 1 1 1 1
IS UTB FD 1 1 1.1 1.15 1.18
IS DG 1.1 1.15 1.22 1.27 1.4 1.45 1.5 1.55
Extended Planar Bulk (Ω-µm) 180 180 180 180 180 180
IS UTB FD (Ω-µm) 200 180 160 150 150
IS DG (Ω-µm) 200 200 180 180 170 160 140 140
IS Extended Planar Bulk (fF/µm) 0.581 0.601 0.558 0.532 0.502 0.490
IS UTB FD (F/µm) 0.431 0.414 0.419 0.425 0.431
IS DG (F/µm) 0.368 0.322 0.320 0.296 0.292 0.265 0.259 0.230
IS Extended Planar Bulk (fF/µm) 0.791 0.821 0.768 0.732 0.692 0.670
IS UTB FD (F/µm) 0.671 0.654 0.639 0.625 0.631
IS DG (F/µm) 0.608 0.562 0.560 0.526 0.502 0.465 0.449 0.420
τ =CV/I: NMOSFET intrinsic delay (ps) [16]
Extended Planar Bulk (ps) 1.74 1.64 1.46 1.35 1.24 1.23
IS UTB FD (ps) 0.9 0.79 0.69 0.65 0.6
IS DG (ps) 0.69 0.61 0.53 0.47 0.42 0.37 0.34 0.31
1/τ: NMOSFET intrinsic switching speed (GHz) [17]
Extended Planar Bulk (GHz) 575 610 685 741 806 813
IS UTB FD (GHz) 1111 1266 1449 1538 1667
IS DG (GHz) 1449 1639 1887 2128 2381 2703 2941 3226
Manufacturable solutions exist, and are being optimized
Manufacturable solutions are known
Interim solutions are known ¿Manufacturable solutions are NOT known
Rsd: Effective Parasitic series source/drain resistance [13]
Cg,ideal: Ideal NMOS Device Gate Capacitance [14]
Cg,total: Total gate capacitance for calculation of CV/I [15]
Table PIDS3c and d Low Operating Power Technology Requirements
Year in Production 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022
59 52 45 40 36 32 28 25 22 20 18 16 14 13 11
IS 29 27 24 22 20 18 17 15 14 12.8 11.7 10.7 9.7 8.9 8.1
WAS 28 25 22 20 18 16 14 13 11 10 9 8 7 6.5 6
IS 32 29 27 24 22 18 17 15 14 12.8 11.7 10.7 9.7 8.9 8.1
EOT: Equivalent Oxide Thickness [2]
IS Extended planar bulk (Å) 12 11 10 10 9 8
IS UTB FD (Å) 9 8.5 8 8 8 7.3
IS DG (Å) 8.5 8 8 8 7.5 7 7 7
Gate Poly Depletion and Inversion-Layer Equivalent Thickness [3]
IS Extended planar bulk (Å) 6.4 3.4 3.4 3.3 3.3 3.2
IS UTB FD (Å) 4 4 4 4 4 4
IS DG (Å) 4 4 4 4 4 4 4 4
IS Extended planar bulk (Å) 18.4 14.4 13.4 13.3 12.3 11.2
IS UTB FD (Å) 13 12.5 12 12 11.9 11.4
IS DG (Å) 12.5 12 12 12 11.7 11 11 11
IS 78 86 95 100 110 140
IS 140 150 170 180 200 220
IS 170 180 200 220 230 260 280 310
IS Extended Planar Bulk (V) 0.8 0.8 0.8 0.77 0.7 0.7
IS UTB FD (V) 0.7 0.65 0.6 0.6 0.59 0.54
IS DG (V) 0.6 0.6 0.6 0.6 0.57 0.5 0.5 0.5
IS Extended Planar Bulk (mV) 310 313 322 336 259 249
IS UTB FD (mV) 209 202 199 202 201 193
IS DG (mV) 202 201 202 202 198 190 194 190
IS Extended Planar Bulk (nA/µm) 9.08 7.78 7.89 12.1 18.3 35.7
IS UTB FD (µA/µm) 11.9 16.1 19.4 18.6 21.5 31.0
IS DG (µA/µm) 9.66 10.7 11.2 12.4 15.9 21.4 20.0 24.9
IS Extended Planar Bulk (µA/µm) 541 634 651 600 682 760
IS UTB FD (µA/µm) 788 768 755 763 801 749
IS DG (µA/µm) 790 826 895 908 884 821 855 895
Grey cells delineate one of two time periods: either before initial production ramp has started for ultra-thin body fully depleted (UTB FD) SOI or double-gate (DG) MOSFETs, or beyond when planar bulk or UTB FD MOSFETs have reached the limits of practical scaling (see the text and the table notes for further discussion).
MPU/ASIC Metal 1 (M1) ½ Pitch (nm) (contacted)
Lg: Physical Lgate for High Performance logic (nm) [1]
Lg: Physical gate length for LOP (nm) [1]
Lg: Physical gate length for LOP (nm) [1]
EOTelec: Electrical Equivalent Oxide Thickness in inversion [4]
Jg,limit: Maximum gate leakage current density [5]
Extended Planar Bulk (A/ cm2)
UTB FD (A/ cm2)
DG (A/ cm2)
Vdd: Power Supply Voltage (V) [6]
Vt,sat: Saturation Threshold Voltage [7]
Isd,leak: Source/Drain Subthreshold Off-State Leakage Current [8]
Id,sat: NMOS Drive Current [9]
ORTCINDEX
2007 ITRS Chapters
2008INDEX
1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8
IS Extended Planar Bulk 1.15 1.15 1.15 1.15 1.11 1.09
IS UTB FD 1.07 1.07 1.07 1.06 1.06 1.06
IS DG 1.05 1.05 1.04 1.04 1.04 1.04 1.04 1.04
Effective Ballistic Enhancement Factor [12]
IS Extended Planar Bulk 1 1 1 1 1 1
IS UTB FD 1 1.05 1.1 1.1 1.16 1.21
IS DG 1.17 1.18 1.25 1.26 1.29 1.37 1.43 1.45
IS Extended Planar Bulk (Ω-µm) 190 190 190 190 190 190
IS UTB FD (Ω-µm) 190 185 175 170 165 162
IS DG (Ω-µm) 180 180 170 170 166 154 149 141
IS Extended Planar Bulk (fF/µm) 0.601 0.653 0.663 0.636 0.613 0.554
IS UTB FD (F/µm) 0.478 0.469 0.431 0.402 0.368 0.335
IS DG (F/µm) 0.414 0.402 0.368 0.336 0.315 0.304 0.279 0.254
IS Extended Planar Bulk (fF/µm) 0.789 0.869 0.867 0.789 0.813 0.754
IS UTB FD (F/µm) 0.688 0.674 0.631 0.602 0.550 0.529
IS DG (F/µm) 0.654 0.643 0.608 0.576 0.546 0.511 0.478 0.445
τ = CV/I: NMOSFET intrinsic delay (ps) [16]
IS Extended Planar Bulk (ps) 1.17 1.10 1.07 1.01 0.83 0.69
IS UTB FD (ps) 0.61 0.57 0.5 0.47 0.41 0.38
IS DG (ps) 0.5 0.47 0.41 0.38 0.35 0.31 0.28 0.24
1/τ: NMOSFET intrinsic switching speed (GHz) [17]
IS Extended Planar Bulk (GHz) 857 912 939 988 1205 1449
IS UTB FD (GHz) 1639 1763 2008 2128 2473 2660
IS DG (GHz) 2008 2128 2473 2660 2882 3259 3631 4107
Manufacturable solutions exist, and are being optimized
Manufacturable solutions are known
Interim solutions are known ¿Manufacturable solutions are NOT known
Mobility enhancement factor due to strain [10]
Id,sat enhancement factor due to strain [11]
Rsd: Effective Parasitic series source/drain resistance [13]
Cg,ideal: Ideal NMOS Device Gate Capacitance [14]
Cg,total: Total gate capacitance for calculation of CV/I [15]
Table PIDS4 DRAM Technology Requirements
Year in Production 2007 2008 2009 2010
DRAM ½ Pitch (nm) [1] 68 58 50 45
0.0277 0.0202 0.015 0.0122
1.2 0.9 0.8 0.6
DRAM storage node cell capacitor voltage (V) [4] 0.65 0.65 0.6 0.6
Equivalent Electric field of capacitor dielectric, (MV/cm) [5] 5.7 7.2 7.5 10
DRAM cell FET structure [6] RCAT RCAT RCAT FinFET
5 5 4.5 4
Maximum Wordline (WL) level (V) [8] 3 2.8 2.7 2.7
Negative Wordline (WL) use [9] yes yes yes yes
Equivalent Electric field of cell FET device dielectric (MV/cm) [10] 6 5.6 6 6.75
Cell Size Factor: a [11] 6 6 6 6
Array Area Efficiency [12] 0.56 0.56 0.56 0.56
Minimum DRAM retention time (ms) [13] 64 64 64 64
DRAM soft error rate (fits) [14] 1000 1000 1000 1000
1.3 1.2 1.1 1.1
Support nMOS EOT [nm] [16] 3.2 3 2.6 2.6
Support PMOS Gate Electrode [17] P+Poly/W P+Poly/W P+Poly/W P+Poly/W
Support Gate Oxide [18] SiON SiON SiON SiON
100 90 75 75
500 465 470 450
0.4 0.4 0.38 0.37
220 210 220 210
-0.45 -0.4 -0.38 -0.38
Notes for Table PIDS4a and b:
[4] The DRAM storage node capacitor voltage must be low enough that the resulting electric field in the dielectric (see Note [5]) is within acceptable limits.
DRAM cell size (µm2) [2]
DRAM storage node cell capacitor dielectric: equivalent oxide thickness EOT (nm) [3]
DRAM cell FET dielectric: equivalent oxide thickness, EOT (nm) [7]
Vint (support FET voltage) [V] [15]
Support min. Lgate for NMOS FET, physical [nm] [19]
Support Isat-n [µA/µm] (25C, Vg=Vd=Vint) [20]
Support min. Vtn (25C, Gm,max, Vd=55mV) [21]
Support Isat-p [µA/µm] (25C, Vg=Vd=-Vint) [22]
Support min. Vtp (25C, Gm,max, Vd=55mV) [23]
[1] From ORTC (Overall Roadmap Technology Characteristics) Table 1a and b. These DRAM half pitch numbers are the same as those in the 2006 ITRS due to no further speed up in the pace of DRAM half pitch scaling during 2006.
[2] The DRAM cell size is driven by the values for DRAM capacity (bits per chip) and chip size, as discussed in more detail in the Front End Process chapter. The capacity and chip size numbers are based on the ORTC Tables 1a and 1b. Since the DRAM capacity and chip size numbers are quite aggressive, the cell size must also be scaled aggressively. The difficulty will lie in reducing the value of the cell size factor “a”, where “a” equals (cell size /F2) and F is the DRAM half pitch. The required values of “a” are 6 for 68 nm and beyond.
[3] Storage node cell dielectric EOT is defined as (dielectric physical thickness / [κ/3.9]), where κ is the relative dielectric constant of the storage node cell dielectric and 3.9 is the relative dielectric constant of thermal SiO2. The value of EOT is driven by the values for DRAM capacity (bits per chip) and chip size, as discussed in more detail in the Front End Process chapter. The capacity and the chip size numbers used by FEP are from the ORTC Tables 1a and 1b. Since the values of DRAM capacity and chip size from FEP are quite aggressive, the EOT must also be scaled very aggressively. Up to 2009, the dielectric material is based on Al2O3 or HfO2, and hence the color is white. Beyond 2009, breakthroughs such as MIM structure with higher κ insulator material of epsilon more than 40 and physical thickness of less than 9nm are needed, so the color is yellow. Finally, for 2012 and beyond, there are no known solutions with demonstrated credibility, and hence the color is red. The actual EOT required for each year also depends on other factors such as cell height and/or 3D structure, film leakage current and contact formation.
ORTCINDEX
2007 ITRS Chapters
2008INDEX
[14] This is a typical FIT rate and depends on cycle time and the quality of cell capacitor and sensing circuits.
[17] Support PMOS FET Gate electrode material migrates from P+Poly/W to TiN.
[18] DRAM support MOS FET dielectric material migrates from SiON to HfSiON in order to leakage current to meet.
[5] The equivalent electric field in the capacitor dielectric is (DRAM storage node capacitor voltage / DRAM storage node dielectric equivalent oxide thickness, EOT). The equivalent field is the electric field if the dielectric is silicon dioxide; if the dielectric is high-k, the actual electric field is [equivalent field]/[k/3.9]. Note the sharp increase in the equivalent field with scaling. The color turns yellow in 2009, when the field is 7.5 MV/cm, and red in 2012, when the field becomes 13.75 MV/cm..
[6] DRAM cell MOSFET structure migrates from RCAT (recessed channel array transistor) to FinFET. RCAT is a technology to improve retention time characteristics by introducing recessed channel structure. FinFET is used to increase the drive current in the limited cell FET area and also to improve retention time characteristics.
[7] DRAM cell FET dielectric EOT is defined as (dielectric physical thickness / [κ/3.9]), where κ is the relative dielectric constant of the DRAM cell FET dielectric and 3.9 is the relative dielectric constant of thermal SiO2. The EOT values here are large, mainly because of the high word line voltage levels (see Note 8) and the need to keep the electric field in the dielectric within tolerable limits (see Note 9)
[8] Maximum word line level is the (highly boosted) gate voltage for cell FET devices. The high gate voltage is required to get enough device drive current with high threshold voltage due to back gate voltage at the operating condition.
[9] Negative word line is used to suppress sub-threshold leakage current of cell transistor even in the case of lower level of V t value of cell FET. The low Vt is preferable to get higher drive current of cell FET.
[10] The equivalent electric field in the cell FET device dielectric is (maximum word line level / DRAM cell FET dielectric equivalent oxide thickness, EOT). The equivalent field is the electric field if the dielectric is silicon dioxide; if the dielectric is high-k, the actual field is [equivalent field]/[k/3.9].
[11] Cell size factor = a = (DRAM cell size/F2), where F is the DRAM ½ pitch. The required values of a are 6 for 2007 and beyond. In contrast, the 2005 version of the DRAM table has a = 8 in 2005, 2006 and 2007.
[12] Array area efficiency is the ratio of cell array area to total chip area. Hence, array area efficiency = 1 / (1 + [peripheral circuit area]/NaF2), where N is the DRAM capacity (number of bits per chip), F is the DRAM ½ pitch, and a is the cell size factor (see Note 9). For a = 8, array area efficiency is estimated to be 0.63, so when a is decreased to 6 from 2007, the array area efficiency of 0.56 is made in conjunction with a = 6, assuming the same relative peripheral circuit area.
[13] Retention time is defined at 85ºC, and is the minimum time during which the data from memory can still be sensed correctly without refreshing a row bit line. The 64 ms specified here is the value needed for PC applications. The retention time depends on the combined interaction of device leakage current, signal strength, and signal sensing circuit sensitivity, and also depends on operational frequency and temperature.
[15] Vint is the nominal power supply voltage for DRAM support FET in peripheral circuit area. It has been chosen to maintain sufficient voltage over-drive in order to meet the required saturation current drive values while still maintaining reasonable vertical gate dielectric electric field strengths.
[16] DRAM support MOS FET dielectric EOT is defined as (dielectric physical thickness / [κ/3.9]), where κ is the relative dielectric constant of the DRAM cell FET dielectric and 3.9 is the relative dielectric constant of thermal SiO2.
[19] Physical support min. Lgate for NMOS FET is the final, as-etched length of the bottom of the gate electrode.
[20] Support Isat-n (the saturation drive current for support NMOS FET) is defined as the NMOSFET drain current per micron device width with the gate bias and the drain bias set equal to V int
(see Note [15]) and the source and substrate biases set to zero at 25°C, namely Vg=Vd=Vint).
[21] Support min Vtn is the saturation threshold voltage measured at 25°C, Gm max,Vd=55mV.
[22] Support Isat-p (the saturation drive current for support PMOS FET) is defined as the PMOSFET drain current per micron device width with the gate bias and the drain bias set equal to -
Vint (see Note [15]) and the source and substrate biases set to zero at 25°C ,namely, Vg=Vd=-Vint).
[23] Support min Vtp is the saturation threshold voltage measured at 25°C, Gm max,Vd=55mV.
2011 2012 2013 2014 2015 2016 2017 2018 2019
40 36 32 30 25 22 20 18 16
0.0096 0.0078 0.0061 0.0054 0.0038 0.0029 0.0024 0.0019 0.00154
0.5 0.4 0.3 0.3 0.3 0.3 0.3 0.3 0.25
0.55 0.55 0.5 0.5 0.45 0.45 0.4 0.4 0.35
11 13.8 16.7 16.7 15 15 13.3 13.3 14
FinFET FinFET FinFET FinFET FinFET FinFET FinFET FinFET FinFET
4 4 4 4 4 3.5 3.5 3.5 3.5
2.7 2.7 2.6 2.6 2.4 2.3 2.3 2.3 2
yes yes yes yes yes yes yes yes yes
6.75 6.75 6.5 6.5 6 6.57 6.57 6.57 5.71
6 6 6 6 6 6 6 6 6
0.56 0.56 0.56 0.56 0.56 0.56 0.56 0.56 0.56
64 64 64 64 64 64 64 64 64
1000 1000 1000 1000 1000 1000 1000 1000 1000
1.1 1.1 1.1 1 0.9 0.9 0.9 0.9 0.9
2.5 2.2 2 1.8 1.6 1.5 1.4 1.4 1.3
P+Poly/W P+Poly/W P+Poly/W TiN TiN TiN TiN TiN TiN
SiON SiON HfSiON HfSiON HfSiON HfSiON HfSiON HfSiON HfSiON
65 60 50 48 40 35 31 28 25
410 430 450 440 480 550 550 550
0.37 0.33 0.33 0.31 0.31 0.31 0.31 0.31 0.31
165 170 175 170 190 215 215 215 215
-0.38 -0.34 -0.34 -0.32 -0.32 -0.32 -0.32 -0.32 -0.32
Notes for Table PIDS4a and b:
[4] The DRAM storage node capacitor voltage must be low enough that the resulting electric field in the dielectric (see Note [5]) is within acceptable limits.
445
[1] From ORTC (Overall Roadmap Technology Characteristics) Table 1a and b. These DRAM half pitch numbers are the same as those in the 2006 ITRS due to no further speed up in the pace
[2] The DRAM cell size is driven by the values for DRAM capacity (bits per chip) and chip size, as discussed in more detail in the Front End Process chapter. The capacity and chip size numbers are based on the ORTC Tables 1a and 1b. Since the DRAM capacity and chip size numbers are quite aggressive, the cell size must also be scaled aggressively. The difficulty will lie in
) and F is the DRAM half pitch. The required values of “a” are 6 for 68 nm and beyond.
[3] Storage node cell dielectric EOT is defined as (dielectric physical thickness / [κ/3.9]), where κ is the relative dielectric constant of the storage node cell dielectric and 3.9 is the relative . The value of EOT is driven by the values for DRAM capacity (bits per chip) and chip size, as discussed in more detail in the Front End Process chapter. The
capacity and the chip size numbers used by FEP are from the ORTC Tables 1a and 1b. Since the values of DRAM capacity and chip size from FEP are quite aggressive, the EOT must also be , and hence the color is white. Beyond 2009, breakthroughs such as MIM structure with higher κ insulator
material of epsilon more than 40 and physical thickness of less than 9nm are needed, so the color is yellow. Finally, for 2012 and beyond, there are no known solutions with demonstrated credibility, and hence the color is red. The actual EOT required for each year also depends on other factors such as cell height and/or 3D structure, film leakage current and contact formation.
[14] This is a typical FIT rate and depends on cycle time and the quality of cell capacitor and sensing circuits.
[17] Support PMOS FET Gate electrode material migrates from P+Poly/W to TiN.
[18] DRAM support MOS FET dielectric material migrates from SiON to HfSiON in order to leakage current to meet.
[5] The equivalent electric field in the capacitor dielectric is (DRAM storage node capacitor voltage / DRAM storage node dielectric equivalent oxide thickness, EOT). The equivalent field is the actual electric field is [equivalent field]/[k/3.9]. Note the sharp increase in the equivalent field with scaling.
The color turns yellow in 2009, when the field is 7.5 MV/cm, and red in 2012, when the field becomes 13.75 MV/cm..
[6] DRAM cell MOSFET structure migrates from RCAT (recessed channel array transistor) to FinFET. RCAT is a technology to improve retention time characteristics by introducing recessed channel structure. FinFET is used to increase the drive current in the limited cell FET area and also to improve retention time characteristics.
[7] DRAM cell FET dielectric EOT is defined as (dielectric physical thickness / [κ/3.9]), where κ is the relative dielectric constant of the DRAM cell FET dielectric and 3.9 is the relative . The EOT values here are large, mainly because of the high word line voltage levels (see Note 8) and the need to keep the electric field in the dielectric within
[8] Maximum word line level is the (highly boosted) gate voltage for cell FET devices. The high gate voltage is required to get enough device drive current with high threshold voltage due to
[9] Negative word line is used to suppress sub-threshold leakage current of cell transistor even in the case of lower level of V t value of cell FET. The low Vt is preferable to get higher drive
[10] The equivalent electric field in the cell FET device dielectric is (maximum word line level / DRAM cell FET dielectric equivalent oxide thickness, EOT). The equivalent field is the electric
), where F is the DRAM ½ pitch. The required values of a are 6 for 2007 and beyond. In contrast, the 2005 version of the DRAM table has a = 8 in
[12] Array area efficiency is the ratio of cell array area to total chip area. Hence, array area efficiency = 1 / (1 + [peripheral circuit area]/NaF2), where N is the DRAM capacity (number of bits per chip), F is the DRAM ½ pitch, and a is the cell size factor (see Note 9). For a = 8, array area efficiency is estimated to be 0.63, so when a is decreased to 6 from 2007, the array area
[13] Retention time is defined at 85ºC, and is the minimum time during which the data from memory can still be sensed correctly without refreshing a row bit line. The 64 ms specified here is the value needed for PC applications. The retention time depends on the combined interaction of device leakage current, signal strength, and signal sensing circuit sensitivity, and also depends
is the nominal power supply voltage for DRAM support FET in peripheral circuit area. It has been chosen to maintain sufficient voltage over-drive in order to meet the required
[16] DRAM support MOS FET dielectric EOT is defined as (dielectric physical thickness / [κ/3.9]), where κ is the relative dielectric constant of the DRAM cell FET dielectric and 3.9 is the
(the saturation drive current for support NMOS FET) is defined as the NMOSFET drain current per micron device width with the gate bias and the drain bias set equal to V int
[22] Support Isat-p (the saturation drive current for support PMOS FET) is defined as the PMOSFET drain current per micron device width with the gate bias and the drain bias set equal to -
2020 2021 2022
14 13 12
0.00118 0.00101 0.00086
0.2 0.15 0.12
0.35 0.35 0.35
17.5 23.3 29.2
FinFET FinFET FinFET
3.5 3.5 3.5
2 2 2
yes yes yes
5.71 5.71 5.71
6 6 6
0.56 0.56 0.56
64 64 64
1000 1000 1000
0.9 0.7 0.7
1.3 1.3 1.2
TiN TiN TiN
HfSiON HfSiON HfSiON
23 21 19
550 550 550
0.31 0.31 0.31
215 215 215
-0.32 -0.32 -0.32
Table PIDS5 Non-Volatile Memory Technology Requirements
Near-term
Year of Production 2007 2008 2009 2010
DRAM ½ Pitch (nm) (contacted) 65 57 50 45
67 60 54 48
NAND Flash poly 1/2 Pitch (nm) 51 45 40 36
NAND Flash
51 45 40 36
32 32 64 64
Cell type (FG, CT, 3D, etc.) [3] FG FG FG FG/CT
3D NAND number of memory layers 1 1 1 1
A. Floating Gate NAND Flash
4.0/2.0 4.0/2.0 4.0/1.3 4.0/1.0
Tunnel oxide thickness (nm) [5] 6-7 6-7 6-7 6-7
Interpoly dielectric material [6] ONO ONO ONO ONO
Interpoly dielectric thickness (nm) 10-13 10-13 10-13 10-13
Gate coupling ratio (GCR) [7] 0.6–0.7 0.6–0.7 0.6–0.7 0.6–0.7
Control gate material [8] n-Poly n-Poly n-Poly n-Poly
Highest W/E voltage (V) [9] 17-19 17-19 15-17 15-17
Endurance (erase/write cycles) [10] 1E+05 1E+05 1E+05 1E+05
Nonvolatile data retention (years) [11] 10-20 10-20 10-20 10-20
2 2 3 4
4.0/1.0
Tunnel dielectric material [14] SiO2 or ONO
Tunnel dielectric thickness EOT (nm) 3-4
Blocking dielectric material [15] SiO2 or Al2O3
Blocking dielectric thickness EOT (nm) 6–8
Charge trapping layer material [16] SiN
Charge trapping layer thickness (nm) [17] 5–7
Gate material [18]
Highest W/E voltage (V) 15-17
MPU/ASIC Metal 1 (M1) ½ Pitch (nm)(contacted)
NAND Flash technology node – F (nm) [1]
Number of word lines in one NAND string [2]
Cell size – area factor a in multiples of F2 SLC/MLC [4]
Maximum number of bits per cell (MLC) [12]
B. Charge trapping NAND Flash (MANOS or Barrier Engineering) [13]
Cell size – area factor a in multiples of F2 SLC/MLC
p-Poly/Metal
ORTCINDEX
2007 ITRS Chapters
2008INDEX
Endurance (erase/write cycles) [19] 1E+05
Nonvolatile data retention (years) [20] 10-20
Maximum number of bits per cell (MLC) 4
NOR Flash
65 57 50 45
A. Floating gate NOR Flash
9-11 9-11 9-11 9-11
Gate length Lg, physical (nm) [26] 130 120 100 90
Tunnel oxide thickness (nm) [27] 8–9 8–9 8–9 8
Interpoly dielectric material [28] ONO ONO ONO ONO
Interpoly dielectric thickness EOT (nm) 13-15 13-15 13-15 13-15
Gate coupling ratio [29] 0.6–0.7 0.6–0.7 0.6–0.7 0.6–0.7
Highest W/E voltage (V) [30] 7-9 7-9 7-9 7-9
Iread (µA) [31] 25-34 23-31 21-27 20-26
Endurance (erase/write cycles) [32] 1.00E+05 1.00E+05 1.00E+05 1.00E+06
Nonvolatile data retention (years) [33] 10–20 10–20 10–20 10–20
2 2 2 2
CC CC CC CC
SONOS/NROM technology node, F (nm) 65 57 50 45
6-7 6-7 6-7 6-7
3.3/1.6 3.3/1.6 3.3/1.6 3.3/1.6
Gate length Lg, physical (nm) [38] 0.14 0.13 0.12 0.11
Tunnel oxide thickness (nm) [39] 5 5 5 4.5
5-7 5-7 5-7 4-6
7–9 7–9 7–9 6–8
Highest W/E voltage (V) 7-9 7-9 7-9 6-8
Iread (µA) [31] 25-34 23-31 21-27 20-26
Endurance (erase/write cycles) [32] 1.00E+05 1.00E+05 1.00E+05 1.00E+06
Nonvolatile data retention (years) [33] 10–20 10–20 10–20 10–20
4 4 4 4
Non-charge-storage NVM
A. FeRAM (Ferroelectric RAM)
NOR Flash technology node – F (nm) [21]
Cell size – area factor a in multiples of F2 [22], [23], [24], [25]
Maximum number of bits per cell (MLC) [34]
Array architecture (with cell contact (CC) or virtual ground (VG))[ 35]
B. Charge trapping NOR Flash (SONOS/NROM) [36]
SONOS/NROM cell size - area factor a in multiples of F2
Cell size (per bit) – area factor a in multiples of F2 (SLC/MLC) [37]
Charge trapping layer thickness (nm) [40]
Blocking (top) dielectric thickness EOT (nm) [41]
Maximum number of bits per cell (physical 2-bit/cell + MLC) [37]
WAS FeRAM technology node – F (nm) [42] 180 180 180 130
IS FeRAM technology node – F (nm) [42] 180 180 180 150
22 22 22 20
FeRAM cell size ( µm2) 0.713 0.713 0.713 0.450
FeRAM cell structure [44] 2T2C 1T1C 1T1C 1T1C
FeRAM capacitor structure [45] stack stack stack stack
WAS FeRAM capacitor footprint (µm2) [46] 0.330 0.330 0.330 0.200
IS FeRAM capacitor footprint (µm2) [46] 0.330 0.330 0.330 0.20
FeRAM capacitor active area (µm2) [47] 0.330 0.330 0.330 0.20
1.00 1.00 1.00 1.00
Ferro capacitor voltage (V) [49] 1.50 1.50 1.50 1.20
13.5 13.5 13.5 20
1.0E+14 ### ### ###
10 Years 10 Years 10 Years 10 Years
B. MRAM (Magnetic RAM)
MRAM technology node F (nm) [52] 90 65 65 45
20 22 19 20
MRAM typical cell size (µm2) 0.16 0.09 0.08 0.041
MRAM switching field (Oe) [53] 35 35 35 35
MRAM write energy (pJ/bit) [54] 70 35 35 25
MRAM active area per cell (µm2) [55] 0.05 0.025 0.025 0.013
2 1.1 1 0.8
MRAM magnetoresistance ratio(%) [57] 70 70 70 70
MRAM nonvolatile data retention (years) >10 >10 >10 >10
>3e16 >3e16 >3e16 >3e16
>10 >10 >10 >10
C. PCRAM (Phase-Change RAM)
PCRAM technology node F (nm) [58] 72 58 46 40
4.8 4.0 4.0 4.0
15.0 14.0 12.0 11.0
24883 13456 8464 6400
77760 47096 25392 17600
FeRAM cell size – area factor a inmultiples of F2 [43]
FeRAM cap active area/footprint ratio [48]
FeRAM minimum switching chargedensity (µC/cm2) [50]
FeRAM endurance (read/write cycles) [51]
FeRAM nonvolatile data retention(years)
MRAM cell size area factor a in multiples of F2
MRAM resistance-area product (Kohm-(µm2) [56]
MRAM write endurance (read/write cycles)
MRAM endurance – tunnel junction reliability (years at bias) [58]
PCRAM cell size area factor a in multiples of F2 (BJT access device) [59]
PCRAM cell size area factor a in multiples of F2 (nMOSFET access device) [60]
PCRAM typical cell size (nm2) (BJT access device) [61]
PCRAM typical cell size (nm2) (nMOSFET access device) [62]
1 1 2 2
24883 13456 4232 3200
77760 47096 12696 8800
PCRAM storage element CD (nm) [66] 45 36 30 25
373,000 195,000 112,000 64,000
PCRAM reset current (µA) [68] 235 170 130 100
PCRAM set resistance (Kohm) [69] 3.54 4.57 5.68 7.08
1.50E+07 1.50E+07 1.50E+07 1.50E+07
PCRAM BJT emitter area (nm2) [71] 4072 2642 1662 1257
1.5 1.5 1.8 1.8
239 171 108 82
>10 >10 >10 >10
1.0E+08 1.0E+08 1.0E+10 1.0E+10
Manufacturable solutions exist, and are being optimized
Manufacturable solutions are known
Interim solutions are known ¿
Manufacturable solutions are NOT known
Notes for Table PIDS5a and b:
[1] NAND Flash has surpassed CMOS and DRAM technology since 2005. This entry provides the F value for designs in the indicated time period.
PCRAM number of bits per cell (MLC) [63]
PCRAM typical cell area per bit size (µm2) (BJT access device) [64]
PCRAM typical cell area per bit size (µm2) (nMOSFET access device) [65]
PCRAM phase change volume (nm3) [67]
PCRAM BJT current density (A/cm2) [70]
PCRAM nMOSFET apparent current density for reset (µA/nm) [72]
PCRAM nMOSFET apparent device width (nm) [73]
PCRAM nonvolatile data retention (years) [74]
PCRAM write endurance (read/write cycles) [75]
[2] NAND Flash architecture consists of bit line strings of a number of storage devices. Long bit line strings reduce the overhead for bit line transistors and increase the packing density, however, at the expense of higher overall resistance and consequently lower read current. The number of word lines in a bit line string has increased from 16 to 32 nm.
[3] Because of the difficulty in maintaining high gate coupling ratio and preventing cross talk between neighboring cells, NAND technology is forecasted to migrate gradually from floating gate devices (FG) to charge trapping devices (CT). (Ref: K. Kim, "Technology for sub 50nm DRAM and NAND Flash manufacturing,” in Tech. Digest International Electron Devices Meeting, pp. 539-543, 2005.) The statistical fluctuation limit of storing too few electrons imposes new challenges to data retention (Ref: G. Molas, et al., "Impact of few electron phenomena on floating gate memory reliability,” Tech. Digest 2004 International Electron Devices Meeting, pp. 877-880, 2004.) and 3D integration of multiple layers of devices may be required to continue the scaling. (Refs: S.H. Lee et al., "Three dimensionally stacked NAND Flash memory technology using stacking single crystal Si layers on ILD and TANOS structure for beyond 30nm node,” Tech. Digest 2006 International Electron Devices Meeting, pp. 37-40, 2006. E-K. Lai, et al., "A multi-layer stackable thin-film transistor (TFT) NAND-type Flash memory,” Tech. Digest 2006 International Electron Devices Meeting, pp. 41-44, 2006.)
[4] The area factor “a” = cell area per bit/F2, so this entry presents the expected range for Flash cell area in multiples of the implementation technology F 2. It is possible to store more than 1 bit of information in a Flash cell but increasing the logic levels from (1, 0) to (11, 10, 00, 01), etc. by using multi-level cell (MLC). Therefore, the area factor includes both single-level cell (SLC) and multi-level cell (MLC) devices.
[9] Low write and erase voltage is desirable but EOT for tunnel oxide and IPD must be decreased to allow lower W/E voltage without compromising W/E speed.
[11] Data retention is controlled by both tunnel oxide integrity and statistical distribution of the number of stored electrons. Both thinner tunnel oxide and fewer number of stored electrons in the long term contribute to the shorter retention forecast.
[18] High work function metal gate is the best to prevent gate injection. However, p-type polysilicon may become an interim solution because of its easy processing, low cost, and reasonably good performance.
[5] The scaling of tunnel oxide for NAND Flash faces the same challenge as that for NOR Flash. However, the use of error code correction (ECC) in NAND tolerates tunnel oxide defects to a higher level than that for NOR, thus allowing tunnel oxide of 6–7 nm. Currently there are no known solutions to scale tunnel oxide significantly below 6 nm.
[6] ONO has been used as the interpoly dielectric (IPD) up to now and will continue in the near future. However, below 40 nm the spacing between floating gates becomes too narrow to fill effectively with ONO and word line polysilicon, and the loss of sidewall control gate to floating gate coupling will severely degrade the gate coupling ratio (GCR) and the device becomes inoperable. Since it is impossible to create additional space, higher dielectric constant IPD or charge trapping (CT) device must be used. Here the path of migrating to high-κ IPD is shown. It is shown in red color since engineering solutions have not been demonstrated yet.
[7] Gate coupling ratio (GCR) is defined as (control gate to floating gate capacitance)/(total floating gate capacitance). GCR represents the fraction of voltage drop across the tunnel oxide and must be higher than 0.6 for the device to function during write and erase operations. High GCR is normally achieved by wrapping the control gate around the sidewalls of the floating gate. This requires tall floating gate and the cross talk along the bit line direction with neighboring cells is a challenge for MLC operation. At below 40 nm , the spacing between floating gates may become too narrow for the IPD and control gate to wrap around and maintaining sufficiently high GCR is a difficult challenge. Planar device with high-κ IPD is a potential solution.
[8] n-type polysilicon (and polycide) gate has been used for the control gate so far and will continue in the near future. The introduction of high-κ IPD, with a lower barrier height to Si, will cause severe gate injection during erase operation and high work function material such as p-type polysilicon or metal gate may have to be adopted.
[10] Write and erase cycling endurance reflects the tunnel oxide damage caused by repeated passing of charges under high electric field. Scaling does not worsen the oxide damage, however, larger array strains the ECC capability and the tolerance for defects is thus reduced. High-κ IPD may also trap charge and cause degradation. Current projection is gradually reduced cycling endurance for future technology. Note that this is not suitable for certain applications, e.g., solid-state drive storage that requires high cycling endurance.
[12] Multi-level cell (MLC) with 4 logic levels (2-bit/cell) is commonly used for NAND Flash today and devices with 8 logic levels (3-bit/cell) and 16 logic levels (4-bit/cell) are being developed. 8-bit/cell MLC device requires 256 logic levels and so far seems beyond reach of current technology, thus no forecast is made for 8-bit/cell MLC even in the long-term years.
[13] MANOS (Metal-Al2O3-Nitride-Oxide-Si) devices use a relatively thin tunnel oxide, a SiN trapping layer for charge storage, Al 2O3 to increase voltage drop across the tunnel oxide and high work function metal gate to stop gate injection. (Ref: C.H. Lee, et al., "A novel SONOS structure of SiO2/SiN/Al2O3 with TaN metal gate for multi-giga bit Flash memories,” Tech. Digest 2003 International Electron Devices Meeting, pp. 613-616, 2003.) Barrier engineering uses composite tunneling barriers to allow easy erase operation by substrate hole tunneling, yet prevents low field hole direct tunneling during retention. (Ref: H.T. Lue, et al., "BE-SONOS: a bandgap engineered SONOS with excellent performance and reliability,” Tech. Digest 2005 International Electron Devices Meeting, pp. 555-558, 2005.) Without floating gate GCR and cross talk issues, these two CT approaches promise to help NAND scaling below 30 nm.
[14] Tunnel dielectric for MANOS type device is a relatively thin silicon oxide (3-4nm). Tunnel dielectric for barrier engineered (BE) device may be a composite ONO (e.g., 2 nm/2 nm/2 nm) or other composites such as OAO.
[15] For MANOS device Al2O3 is the preferred blocking oxide since it has a high barrier height. Other composite blocking layers such as AN, ANO, or AHO are also potential candidates. For BE device the blocking oxide may be SiO 2, although Al2O3 and other composite layers may also be used.
[16] SiN is the most common and best known charge trapping layer with relatively deep electron traps that provide good data retention. Other exotic high-κ material with even deeper traps may be used in the long term years. (Ref: A. Chin, et al., "Low voltage high speed SiO2/AlGaN/AlLaO3/TaN memory with good retention,” Tech. Digest 2005 International Electron Devices Meeting, pp. 165-168, 2005.) Note that for CT device the charge loss mechanism may be mainly substrate hole tunneling in low field and thus deeper traps do not necessarily improve data retention except for high temperature applications.
[17] Trapping efficiency in SiN seems thickness dependent. (Ref: H.T. Lue, et al., Proc. 2007 International Reliability Physics Symposium, 2007). Therefore, thinner SiN or other high-κ trapping layers forecasted for long-term years may suffer from reduced programming efficiency.
[19] The mechanism for endurance degradation for CT devices is not well understood yet. Unlike floating gate device, CT devices are not sensitive to tunnel oxide damage since the charge is stored in discrete traps and one weak spot does not cause all stored charge to leak out as in floating gate devices. Published endurance data seem to indicate similar endurance as floating gate device.
[20] The mechanism for data retention loss for BE device seems understood. (Ref: H.T. Lue, et al., "Reliability model of bandgap engineered SONOS (BE-SONOS),” Tech. Digest 2006 International Electron Devices Meeting, pp. 495-498, 2006.) Data retention mechanism for MANOS device is still not well understood. Published data suggest that data retention is comparable to floating gate devices under certain conditions.
[21] NOR Flash traditionally falls behind CMOS and DRAM but has caught up in recent years and is now on par with DRAM.
[29] The gate coupling ratio (GCR) is the (control gate to floating gate capacitance)/(total floating gate capacitance). GCR must be greater than about 0.6 for proper device operation.
[32] E/W endurance requirements vary with the specifics of an application, but 1E5 cycles have been accepted as the historical minimum acceptable level for a useful NOR product.
[22] High-κ interpoly dielectric is projected at 32 nm and beyond to achieve gate coupling ratio of >0.6 but this offers only limited help to the area factor. (Ref: 2005 Symposium on VLSI Technology, 11B-3, E. S. Cho, et al.,” Hf-silicate Inter-Poly Dielectric Technology for sub 70 nm Body Tied FinFET Flash Memory,” pp. 208-209.)
[23] Although virtual ground (VG) array may significantly decrease the cell size in the near term years (Ref: 2005 Symposium on VLSI Technology, 11B-1, R. Koval, et al. "Flash ETOX Virtual Ground Architecture: A Future Scaling Direction,” pp. 204-205.), this effect has not been included in the current table because VG is radically different from the conventional array and large development effort is needed to implement it and so far there is no industrial consensus to take up this endeavor.
[24] Although non-planar devices (such as FinFET) are being developed for future Flash scaling, their impact has not been included in the current table. The dilemma of filling ONO and control and floating gates in the narrow gap between adjacent vertical devices has not been completely resolved yet.
[25] Both the cell size and the gate length for NOR Flash have been more aggressively scaled recently with an area factor of approximately 10F2. (Refs: 2005 ISSCC, "A 90 nm 512Mb 166 MHz Multilevel Cell Flash Memory with 1.5MBytes/s Programming,” pp. 54-55. 2003 Symposium on VLSI Technology, "Highly Manufacturable 90nm NOR Flash Technology with 0.081µm2 Cell Size,” pp. 91-92. 2004 Symposium on VLSI Technology, "A 70nm NOR Flash Technology with 0.049µm2 Cell Size,” pp. 238-239.)
[26] NOR Flash uses channel hot electron programming which requires steep junction profile to generate, with the consequence of difficulties in controlling the short channel effect, but yet the NOR architecture is vulnerable to device leakage. Since the tunnel oxide thickness cannot be scaled, controlling the short channel effect imposes a very difficult challenge for scaling. The gate length generally is substantially larger than F in recent years. Despite this difficulty, the area factor has been maintained at 10F 2 in recent years (see note 25).
[27] Tunnel oxides must be thick enough to assure retention but thin enough to allow ease of erase/write. This difficult trade-off problem hinders scaling. Tunnel oxides less than 7 nm pose fundamental problems for retention reliability.
[28] ONO has been used as the interpoly dielectric up to now and most likely in the near future. However, at 32 nm and beyond high-κ IPD may be necessary to maintain GCR at 0.6 or above. Currently, the GCR is achieved by wrapping the control gate over the sidewalls of the floating gate thus increasing the control gate to floating gate capacitor area. At 32 nm or below, the gap between adjacent gloating gates becomes too narrow for the ONO and control gate to fill in and this semi-vertical structure will cease to function.
[30] This is the highest voltage relative to ground seen in the cell array, usually supplied by on-chip charge pumping circuits. Low voltage is desired to reduce the charge pumping circuit overhead and simplify processing. The introduction of high-κ IPD will help to reduce the erase voltage.
[31] In principle the read current decreases with scaling at a rate W/(L*Cox) to prevent voltage overdrive (read disturb). Since access time depends critically on the read current and is an important performance parameter for NOR Flash the read current decreases slower than W/(L*Cox) to maintain performance. This may cause read disturb issues in long term years.
[33] Retention is a defect related parameter rather than an intrinsic device characteristic. Improvement in defect control and accumulation of device history is expected to eventually allow specification of 20 years retention. Also, it should become possible to accept a reduced retention specification as a tradeoff for increased E/W endurance.
[34] Cell read out distinguishes between four levels of charge storage to provide two storage bits (Multilevel cell MLC). Progression to 8 or 16 levels is potentially possible but maintaining reasonable V t, read speed and array efficiency beyond 2-bit/cell are challenging. Unlike NAND Flash where density is a key competitive advantage, performance and reliability tend to hold higher importance than density for NOR Flash, thus the pressure to higher level MLC is not as strong as for NAND Flash.
[35] Virtual ground array uses junction isolation and buried diffusion for bit line, thus requires no STI isolation or bit line contact in the cell. In principle the area factor for VG array can be as small as 4F 2, compared to ~ 10F2 for an array with cell STI isolation and contact. The scaling of buried diffusion is a difficult challenge and junction isolation produces more leakage paths and complicates the design. Large R&D effort is needed to implement VG array and overcome its shortcomings to take advantage of its smaller cell size.
[36] Charge trapping device for NOR application uses mainly a SONOS structure and thus often is confused with the conventional SONOS device. Conventional SONOS is more suitable for a NAND array. The device is programmed by Fowler-Nordheim tunneling of electrons from the substrate and the charge is stored in the SiN layer of SONOS. Since the electrons are stored in deep SiN traps it is difficult to de-trap by Fowler-Nordheim tunneling. Instead, a very thin (2–3 nm) tunnel oxide is used to allow substrate hole tunneling into the SiN to erase the device. However, such thin tunnel oxide also allows direct tunneling of holes from the substrate even under weak electric field produced by the stored electrons and good data retention is difficult to achieve. NROM is a device proposed to solve the SONOS problem in a NOR array15. Using channel hot electron to program the cell, electrons are stored in the SiN layer near the edge of S/D junctions. To erase, band-to-band tunneling generated hot holes are injected into the SiN. A relatively thick (4–5 nm) tunnel oxide is adopted and this solves the data retention issue. NROM also has the advantage of storing two bits of information in one device (source side and drain side) and it applies a reverse read method to distinguish the different states. NROM is built on a virtual ground array architecture and has a relatively small cell size. It is used not only for NOR Flash, but also for some data storage applications even though it is not a NAND structure.
[37] CT NOR SLC Flash stores one bit on the source side and one bit on the drain side, or 2-bit/cell. The MLC Flash stores 2 bits on the source side and two bits on the drain side, thus 4-bit/cell.
[39] Because electrons are trapped in deep levels in SiN, the tunnel oxide can be scaled more aggressively than for floating gate device.
[40] Reducing the thickness of charge trapping SiN will reduce the EOT but will degrade trapping efficiency.
[41] Interpoly dielectric must be thick enough to assure retention.
[42] This entry is the critical dimension “F” within the FeRAM cell for stand-alone memory devices (not embedded devices).
[45] The geometry of the capacitor is a key factor in determining cell size. Stacked planar films are expected to be replaced by more efficient 3D structures.
[47] This is the actual effective area of the capacitor. It is larger than the footprint for 3D capacitor because of the utilization of area in the third dimension.
[48] This ratio of the effective area to the footprint gives a measure of the impact of utilization of the third dimension.
[52] MRAM devices are expected to lag the CMOS current technology up until 45 nm half pitch in 2010. This entry provides the F value for designs in the indicated time period.
[53] The MRAM switching field is the magnetic intensity H required to change the direction of magnetization of the cell.
[55] MRAM active bit area is the area of the magnetic material stack within the cell. It represents the “A” in the R*A product.
[38] Although physically storing two bits in the same device, the gate length scaling is not limited by left-right bit interference. The scaling is limited by the same factors for floating gate device - junction breakdown voltage and short channel effect. 4-bit/cell MLC device window, on the other hand, is affected by left-right bit interference and the Lg scaling for MLC may be more gradual than SLC.
[43] This is the area factor “a” = cell size/F2. FeRAM cell size is presented in terms of F2 multiples of the FeRAM implementation technology.
[44] FeRAM cell structures have migrated to one transistor, one capacitor (1T1C) formats. (Refs. J.H. Park, et al., "Fully Logic Compatible (1.6V Vcc, 2 Additional FRAM Masks) Highly Reliable Sub 10F 2 Embedded FRAM with Advanced Direct Via Technology and Robust 100 nm thick MOCVD PZT Technology", 2004 IEDM, 23.7.1, pp. 591-594. Y. M. Kang et al., "Sub-1.2V Operational, 0.15µm/12F 2 Cell FRAM Technologies for Next Generation SoC Applications", 2005 Symposium on VLSI Technology, 6B-4, pp. 102-103.) Other alternative configurations are under investigation such as Chain-FeRAM. (Refs. H. Kanaya et al., "A 0.602µm2 Nestled Chain Cell Structure Formed by One Mask Etching Process for 64Mbit FeRAM,” 2004 Symposium for VLSI Technology, pp. 150-151. N. Nagel et al., "New Highly Scalable 3 Dimensional Chain FeRAM Cell with Vertical Capacitor,” 2004 Symposium on VLSI Technology, pp. 146-147.)
[46] This is the footprint of the capacitor in micrometers squared. It is this area that constitutes the capacitor area contribution to the cell size. For 2005–2006 ~19F 2, for 2007 - 2009 ~16F2, and for 2010–2020 ~10F2 (3D capacitor) are assumed.
[49] This is the operating voltage (Vop) applied to the capacitor. Low voltage operation is a difficult key design issue. Generally the ferroelectric film thickness needs to be decreased in order to reduce the V op, with great technological challenges. (Ref. D. C. Yoo et al., "Highly Reliable 50nm-thick PZT Capacitor and Low Voltage FRAM Device Using Ir/SrRuO2/MOCVD PZT Capacitor Technology", 2005 Symposium on VLSI Technology, 6B-3, pp. 100-101.)
[50] The minimum switching charge density in µC/cm2 is a useful design parameter. It is equal to the cell minimum switching charge divided by the capacitor actual effective area. The capacitor voltage is taken as V op.
[51] FeRAM is a destructive read-out technology, so every read is accompanied by a write to restore the data. Endurance cycles are taken as the sum of all read and all write cycles. For FeRAM to compete with DRAM and SRAM the cycle endurance should be about 1E15. Testing time is a serious concern. Note that operation at 100 MHz for 3 years would accumulate 1E16 cycles.
[54] MRAM switching energy per bit is calculated as (write current * power supply voltage * write time). It is preferred to use the median value of switching energy measured on a multi-megabit array. A good estimate of power drain is (switching energy * number of writes per second).
[56] MRAM resistance-area product (i.e., the R*A product) is an intrinsic property of the magnetic material stack that provides a convenient basis for comparing cells of different sizes. The R*A product can be computed by measuring the effective low state resistance (R low) of the magnetic tunnel junction and multiply it by the active bit area of the magnetic stack.
[58]This is the critical dimension, F.
[61] The expected “typical” PCRAM cell size with BJT access device is presented in micrometers squared.
[62] The expected “typical” PCRAM cell size with nMOSFET access device is presented in micrometers squared.
[63] PCRAM is capable of MLC multi-bit/cell operation since the resistance ratio between amorphous and crystalline state is typically 100–1,000. This entry is the expected number of MLC bits per cell.
[64] The expected cell size per MLC bit for the PCRAM with BJT cell. It is the physical cell size divided by the number of MLC bits per cell.
[65] The expected cell size per MLC bit for the PCRAM with nMOSFET cell. It is the physical cell size divided by the number of MLC bits per cell.
[66] PCRAM phase change element must be substantially smaller than the technology F to have efficiency reset operation with reasonable current. This entry is the expected dimension for the phase change element in nanometers.
[67] PCRAM phase change volume is a key factor for device design and peak power requirement. This entry is the expected phase change volume in nanometer cubed.
[68] This entry is the expected reset current for PCRAM in microamperes.
[69] The set resistance is a key design factor for PCRAM read speed.
[71] This entry is the expected BJT emitter area that can provide the needed reset current, assuming the BJT current density is met.
[73] This entry is the expected nMOSFET gate width that can provide the needed reset current, assuming the MOSFET output current density is met.
[57] MRAM magnetoresistive ratio is calculated as 100*(Rhigh – Rlow)/Rlow. This ratio summarizes the difference between a logic ONE and a logic ZERO, and as such it represents the intrinsic capability of the magnetic stack. The magnetic tunnel junction resistance values are to be measured at low currents.
[59] The area factor “a” = cell area/F2. This entry is the expected PCRAM cell area in multiples of the implementation technology F2. PCRAM requires significant reset current to change the phase-change element from crystalline to amorphous. A BJT transistor is capable of providing more current per unit area compared to a MOSFET, thus helps to reduce the cell size. Both BJT and nMOSFET access device cells are represented in this table. PCRAM is capable of MLC multi-bit per cell. This area factor is per cell, not per bit.
[60] The area factor “a” = cell area/F2. This entry is the expected PCRAM cell area in multiples of the implementation technology F2. PCRAM requires significant reset current to change the phase-change element from crystalline to amorphous. A BJT transistor is capable of providing more current per unit area compared to a MOSFET, thus helps to reduce the cell size. An nMOSFET transistor has larger cell size in the near term years, but offers simple process and low voltage operation. Both BJT and nMOSFET access device cells are represented in this table. PCRAM is capable of MLC multi-bit per cell. This area factor is per cell, not per bit.
[70] This entry is the expected current density output from the BJT access device required to reset the PCRAM cell (from crystalline to amorphous state). It is a compromise between larger area BJT (which causes larger cell size) and higher output current (which requires higher operation voltage).
[72] This entry is the expected current density output from the nMOSFET access device required to reset the PCRAM cell (from crystalline to amorphous state). It is a compromise between larger width nMOSFET (which causes larger cell size) and higher output current (which requires higher operation voltage or less reliable device).
[74] This entry is the expected PCRAM data retention that will allow it to be used as a nonvolatile memory. Data retention mechanism for PCRAM is not yet thoroughly studied. Recent published data indicate >10 years of retention at elevated temperatures. (Refs. S. J. Ahn et al., "Highly Manufacturable High Density Phase Change Memory of 64Mb and Beyond,” 2004 IEDM, 37.2, pp. 907-910. A. L. Lacaita et al., "Electrothermal and Phase Change Dynamics in Chalcogenide-Based Materials,” 2004 IEDM, 37.3, pp. 911-914.)
[75] This entry is the expected PCRAM W/E cycling endurance. Recent published data indicate cycling endurance from 1E+9 to 1E+13. (Refs. S.J. Ahn et al., "Highly Manufacturable High Density Phase Change Memory of 64Mb and Beyond,” 2004 IEDM, 37.2, pp. 907-910. S. Lai et al., "Current Status of Phase Change Memory and Its Future,” 2003 IEDM, pp. 255-258.)
Near-term Long-term
2011 2012 2013 2014 2015 2016 2017
40 35 32 28 25 22 20
42 38 34 30 27 24 21
32 28 25 22 20 19 18
32 28 25 22 20 19 18
64 64 64 64 64 64 64
CT CT CT-3D CT-3D CT-3D CT-3D CT-3D
1 1 2 2 2 2 2
4.0/1.0 4.0/1.0 4.0/1.0 4.0/1.0 4.0/1.0 4.0/1.0 4.0/1.0
6-7 6-7 6-7 6-7 6-7 4 4
ONO High-K High-K High-K High-K High-K High-K
10-13 9-10 9-10 9-10 9-10 9-10 9-10
0.6–0.7 0.6–0.7 0.6–0.7 0.6–0.7 0.6–0.7 0.6–0.7 0.6–0.7
n-Poly Metal Metal Metal
15-17 15-17 15-17 15-17 15-17 15-17 15-17
1E+05 1E+05 1E+05 1E+05 1E+05 1E+04 1E+04
10-20 10-20 10-20 10-20 10-20 5-10 5-10
4 4 4 4 4 4 4
4.0/1.0 4.0/1.0 4.0/1.0 4.0/1.0 4.0/1.0 4.0/1.0 4.0/1.0
SiO2 or ONO SiO2 or ONO SiO2 or ONO SiO2 or ONO SiO2 or ONO SiO2 or ONO SiO2 or ONO
3-4 3-4 3-4 3-4 3-4 3-4 3-4
SiO2 or Al2O3 SiO2 or Al2O3 SiO2 or Al2O3 SiO2 or Al2O3 SiO2 or Al2O3 SiO2 or Al2O3 SiO2 or Al2O3
6–8 6–8 6–8 6–8 6–8 6–8 6–8
SiN SiN SiN SiN SiN SiN / High-K SiN / High-K
5–7 5–7 5–7 5–7 4–6 4–6 4–6
Metal Metal Metal
15-17 15-17 15-17 15-17 15-17 15-17 15-17
Poly/metal
Poly/metal
Poly/metal
p-Poly/Metal
p-Poly/Metal
p-Poly/Metal
p-Poly/Metal
1E+05 1E+05 1E+05 1E+05 1E+05 1E+04 1E+04
10-20 10-20 10-20 10-20 10-20 5-10 5-10
4 4 4 4 4 4 4
40 35 32 28 25 22 20
9-11 9-11 9-11 9-11 9-11 10-13 10-13
80 70 64 56 50 44 40
8 8 8 7 - 8 7 - 8 7 - 8 7 - 8
ONO ONO High-K High-K High-K High-K High-K
13-15 13-15 8-10 8-10 8-10 8-10 8-10
0.6–0.7 0.6–0.7 0.6–0.7 0.6–0.7 0.6–0.7 0.6–0.7 0.6–0.7
7-9 7-9 6-8 6-8 6-8 6-8 6-8
19-25 17-22 15-20 14-19 13-18 12–17 11–16
1.00E+06 1.00E+06 1.00E+06 1.00E+06 1.00E+06 1.00E+07 1.00E+07
10–20 10–20 20 20 20 20 20
2 2 2 2 2 2 2
CC CC CC/VG CC/VG CC/VG CC/VG CC/VG
40 35 32 28 25 22 20
7-8 7-8 7-8 7-8 8-9 8-9 8-9
3.7/1.9 3.7/1.9 3.7/1.9 3.7/1.9 4.3/2.2 4.3/2.2 4.3/2.2
0.11 0.1 0.1 0.09 0.09 0.08 0.08
4.5 4.5 4 4 4 4 4
4-6 4-6 4-6 4-6 4-5 4-5 4-5
6–8 6–8 6–8 6–8 5–7 5–7 5–7
6-8 6-8 6-8 5–7 5–7 5–7 5–7
19-25 17-22 15-20 14-19 13-18 12–17 11–16
1.00E+06 1.00E+06 1.00E+06 1.00E+06 1.00E+06 1.00E+06 1.00E+06
10–20 10–20 10–20 10–20 10–20 10–20 10–20
4 4 4 4 6 6 6
130 130 90 90 90 90 90
150 150 130 130 130 90 90
20 20 16 16 16 14 14
0.450 0.450 0.270 0.270 0.270 0.113 0.113
1T1C 1T1C 1T1C 1T1C 1T1C 1T1C 1T1C
stack stack stack stack stack 3D 3D
0.200 0.200 0.106 0.106 0.106 0.041 0.041
0.20 0.20 0.106 0.106 0.106 0.041 0.041
0.20 0.20 0.106 0.106 0.106 0.100 0.100
1.00 1.00 1.00 1.00 1.00 2.46 2.46
1.20 1.20 1.20 1.20 1.20 1.00 1.00
20 20 34.0 34.0 34.0 30 30
### ### ### ### ### >1.0E16 >1.0E16
10 Years 10 Years 10 Years 10 Years 10 Years 10 Years 10 Years
45 45 32 32 32 22 22
18 18 19 17 17 18 16
0.036 0.036 0.019 0.017 0.017 0.009 0.0077
35 35 35 35 35 35 35
25 25 20 20 20 20 20
0.013 0.013 0.009 0.009 0.009 0.007 0.007
0.8 0.8 0.6 0.6 0.6 0.6 0.6
70 70 70 70 70 70 70
>10 >10 >10 >10 >10 >10 >10
>3e16 >3e16 >3e16 >3e16 >3e16 >3e16 >3e16
>10 >10 >10 >10 >10 >10 >10
35 32 28 25 22 20 18
4.0 4.0 4.0 4.0 4.0 4.0 4.0
10.0 8.9 8.8 8.4 7.4 7.3 7.3
4900 4096 3136 2500 1936 1600 1296
12250 9114 6899 5250 3582 2920 2365
2 4 4 4 4 4 4
2450 1024 784 625 484 400 324
6125 2278 1725 1313 895 730 591
22 20 18 16 14 13 12
43,000 33,000 25,000 18,000 12,000 9,000 6,700
80 70 62 52 43 37 32
8.29 9.21 10.20 11.66 13.56 15.17 17.18
1.50E+07 1.50E+07 1.50E+07 1.60E+07 1.70E+07 1.90E+07 2.00E+07
962 804 616 491 380 314 254
1.8 2.1 2.1 2.1 2.4 2.4 2.4
68 51 43 36 26 23 21
>10 >10 >10 >10 >10 >10 >10
1.0E+10 1.0E+12 1.0E+12 1.0E+12 1.0E+15 1.0E+15 1.0E+15
[1] NAND Flash has surpassed CMOS and DRAM technology since 2005. This entry provides the F value for designs in the indicated time period.
[2] NAND Flash architecture consists of bit line strings of a number of storage devices. Long bit line strings reduce the overhead for bit line transistors and increase the packing density, however, at the expense of higher overall resistance and consequently lower read
[3] Because of the difficulty in maintaining high gate coupling ratio and preventing cross talk between neighboring cells, NAND technology is forecasted to migrate gradually from floating gate devices (FG) to charge trapping devices (CT). (Ref: K. Kim, "Technology for sub 50nm DRAM and NAND Flash manufacturing,” in Tech. Digest International Electron Devices Meeting, pp. 539-543, 2005.) The statistical fluctuation limit of storing too few electrons imposes new challenges to data retention (Ref: G. Molas, et al., "Impact of few electron phenomena on floating gate memory reliability,” Tech. Digest 2004 International Electron Devices Meeting, pp. 877-880, 2004.) and 3D integration of multiple layers of devices may be required to continue the scaling. (Refs: S.H. Lee et al., "Three dimensionally stacked NAND Flash memory technology using stacking single crystal Si layers on ILD and TANOS structure for beyond 30nm node,” Tech. Digest 2006 International Electron Devices Meeting, pp. 37-40, 2006. E-K. Lai, et al., "A multi-layer stackable thin-film transistor (TFT) NAND-type Flash memory,” Tech. Digest 2006 International Electron Devices Meeting, pp. 41-44, 2006.)
, so this entry presents the expected range for Flash cell area in multiples of the implementation technology F 2. It is possible to store more than 1 bit of information in a Flash cell but increasing the logic levels from (1, 0) to (11, 10, 00, 01), etc. by using multi-level cell (MLC). Therefore, the area factor includes both single-level cell (SLC) and multi-level cell (MLC) devices.
[9] Low write and erase voltage is desirable but EOT for tunnel oxide and IPD must be decreased to allow lower W/E voltage without compromising W/E speed.
[11] Data retention is controlled by both tunnel oxide integrity and statistical distribution of the number of stored electrons. Both thinner tunnel oxide and fewer number of stored electrons in the long term contribute to the shorter retention forecast.
[18] High work function metal gate is the best to prevent gate injection. However, p-type polysilicon may become an interim solution because of its easy processing, low cost, and reasonably good performance.
[5] The scaling of tunnel oxide for NAND Flash faces the same challenge as that for NOR Flash. However, the use of error code correction (ECC) in NAND tolerates tunnel oxide defects to a higher level than that for NOR, thus allowing tunnel oxide of 6–7 nm. Currently
[6] ONO has been used as the interpoly dielectric (IPD) up to now and will continue in the near future. However, below 40 nm the spacing between floating gates becomes too narrow to fill effectively with ONO and word line polysilicon, and the loss of sidewall control gate to floating gate coupling will severely degrade the gate coupling ratio (GCR) and the device becomes inoperable. Since it is impossible to create additional space, higher dielectric constant IPD or charge trapping (CT) device must be used. Here the path of migrating to high-κ IPD is shown. It is shown in red color since engineering solutions have not been demonstrated yet.
[7] Gate coupling ratio (GCR) is defined as (control gate to floating gate capacitance)/(total floating gate capacitance). GCR represents the fraction of voltage drop across the tunnel oxide and must be higher than 0.6 for the device to function during write and erase operations. High GCR is normally achieved by wrapping the control gate around the sidewalls of the floating gate. This requires tall floating gate and the cross talk along the bit line direction with neighboring cells is a challenge for MLC operation. At below 40 nm , the spacing between floating gates may become too narrow for the IPD and control gate to wrap around and maintaining sufficiently high GCR is a difficult challenge. Planar device with high-κ IPD is a potential solution.
[8] n-type polysilicon (and polycide) gate has been used for the control gate so far and will continue in the near future. The introduction of high-κ IPD, with a lower barrier height to Si, will cause severe gate injection during erase operation and high work function
[10] Write and erase cycling endurance reflects the tunnel oxide damage caused by repeated passing of charges under high electric field. Scaling does not worsen the oxide damage, however, larger array strains the ECC capability and the tolerance for defects is thus reduced. High-κ IPD may also trap charge and cause degradation. Current projection is gradually reduced cycling endurance for future technology. Note that this is not suitable for certain applications, e.g., solid-state drive storage that requires high cycling endurance.
[12] Multi-level cell (MLC) with 4 logic levels (2-bit/cell) is commonly used for NAND Flash today and devices with 8 logic levels (3-bit/cell) and 16 logic levels (4-bit/cell) are being developed. 8-bit/cell MLC device requires 256 logic levels and so far seems beyond reach
[13] MANOS (Metal-Al2O3-Nitride-Oxide-Si) devices use a relatively thin tunnel oxide, a SiN trapping layer for charge storage, Al 2O3 to increase voltage drop across the tunnel oxide and high work function metal gate to stop gate injection. (Ref: C.H. Lee, et al., "A novel with TaN metal gate for multi-giga bit Flash memories,” Tech. Digest 2003 International Electron Devices Meeting, pp. 613-616, 2003.) Barrier engineering uses composite tunneling barriers to allow easy erase operation by substrate
hole tunneling, yet prevents low field hole direct tunneling during retention. (Ref: H.T. Lue, et al., "BE-SONOS: a bandgap engineered SONOS with excellent performance and reliability,” Tech. Digest 2005 International Electron Devices Meeting, pp. 555-558, 2005.) Without floating gate GCR and cross talk issues, these two CT approaches promise to help NAND scaling below 30 nm.
[14] Tunnel dielectric for MANOS type device is a relatively thin silicon oxide (3-4nm). Tunnel dielectric for barrier engineered (BE) device may be a composite ONO (e.g., 2 nm/2 nm/2 nm) or other composites such as OAO.
is the preferred blocking oxide since it has a high barrier height. Other composite blocking layers such as AN, ANO, or AHO are also potential candidates. For BE device the blocking oxide may be SiO 2, although Al2O3 and other composite
[16] SiN is the most common and best known charge trapping layer with relatively deep electron traps that provide good data retention. Other exotic high-κ material with even deeper traps may be used in the long term years. (Ref: A. Chin, et al., "Low voltage high speed SiO2/AlGaN/AlLaO3/TaN memory with good retention,” Tech. Digest 2005 International Electron Devices Meeting, pp. 165-168, 2005.) Note that for CT device the charge loss mechanism may be mainly substrate hole tunneling in low field and thus deeper traps do not
[17] Trapping efficiency in SiN seems thickness dependent. (Ref: H.T. Lue, et al., Proc. 2007 International Reliability Physics Symposium, 2007). Therefore, thinner SiN or other high-κ trapping layers forecasted for long-term years may suffer from reduced programming
[19] The mechanism for endurance degradation for CT devices is not well understood yet. Unlike floating gate device, CT devices are not sensitive to tunnel oxide damage since the charge is stored in discrete traps and one weak spot does not cause all stored charge to leak out as in floating gate devices. Published endurance data seem to indicate similar endurance as floating gate device.
[20] The mechanism for data retention loss for BE device seems understood. (Ref: H.T. Lue, et al., "Reliability model of bandgap engineered SONOS (BE-SONOS),” Tech. Digest 2006 International Electron Devices Meeting, pp. 495-498, 2006.) Data retention mechanism for MANOS device is still not well understood. Published data suggest that data retention is comparable to floating gate devices under certain conditions.
[21] NOR Flash traditionally falls behind CMOS and DRAM but has caught up in recent years and is now on par with DRAM.
[29] The gate coupling ratio (GCR) is the (control gate to floating gate capacitance)/(total floating gate capacitance). GCR must be greater than about 0.6 for proper device operation.
[32] E/W endurance requirements vary with the specifics of an application, but 1E5 cycles have been accepted as the historical minimum acceptable level for a useful NOR product.
[22] High-κ interpoly dielectric is projected at 32 nm and beyond to achieve gate coupling ratio of >0.6 but this offers only limited help to the area factor. (Ref: 2005 Symposium on VLSI Technology, 11B-3, E. S. Cho, et al.,” Hf-silicate Inter-Poly Dielectric Technology
[23] Although virtual ground (VG) array may significantly decrease the cell size in the near term years (Ref: 2005 Symposium on VLSI Technology, 11B-1, R. Koval, et al. "Flash ETOX Virtual Ground Architecture: A Future Scaling Direction,” pp. 204-205.), this effect has not been included in the current table because VG is radically different from the conventional array and large development effort is needed to implement it and so far there is no industrial consensus to take up this endeavor.
[24] Although non-planar devices (such as FinFET) are being developed for future Flash scaling, their impact has not been included in the current table. The dilemma of filling ONO and control and floating gates in the narrow gap between adjacent vertical devices has
[25] Both the cell size and the gate length for NOR Flash have been more aggressively scaled recently with an area factor of approximately 10F2. (Refs: 2005 ISSCC, "A 90 nm 512Mb 166 MHz Multilevel Cell Flash Memory with 1.5MBytes/s Programming,” pp. 54-55. 2003 Symposium on VLSI Technology, "Highly Manufacturable 90nm NOR Flash Technology with 0.081µm2 Cell Size,” pp. 91-92. 2004 Symposium on VLSI Technology, "A 70nm NOR Flash Technology with 0.049µm2 Cell Size,” pp. 238-239.)
[26] NOR Flash uses channel hot electron programming which requires steep junction profile to generate, with the consequence of difficulties in controlling the short channel effect, but yet the NOR architecture is vulnerable to device leakage. Since the tunnel oxide thickness cannot be scaled, controlling the short channel effect imposes a very difficult challenge for scaling. The gate length generally is substantially larger than F in recent years. Despite this difficulty, the area factor has been maintained at 10F 2 in recent years (see
[27] Tunnel oxides must be thick enough to assure retention but thin enough to allow ease of erase/write. This difficult trade-off problem hinders scaling. Tunnel oxides less than 7 nm pose fundamental problems for retention reliability.
[28] ONO has been used as the interpoly dielectric up to now and most likely in the near future. However, at 32 nm and beyond high-κ IPD may be necessary to maintain GCR at 0.6 or above. Currently, the GCR is achieved by wrapping the control gate over the sidewalls of the floating gate thus increasing the control gate to floating gate capacitor area. At 32 nm or below, the gap between adjacent gloating gates becomes too narrow for the ONO and control gate to fill in and this semi-vertical structure will cease to function.
[30] This is the highest voltage relative to ground seen in the cell array, usually supplied by on-chip charge pumping circuits. Low voltage is desired to reduce the charge pumping circuit overhead and simplify processing. The introduction of high-κ IPD will help to reduce
[31] In principle the read current decreases with scaling at a rate W/(L*Cox) to prevent voltage overdrive (read disturb). Since access time depends critically on the read current and is an important performance parameter for NOR Flash the read current decreases slower
[33] Retention is a defect related parameter rather than an intrinsic device characteristic. Improvement in defect control and accumulation of device history is expected to eventually allow specification of 20 years retention. Also, it should become possible to accept a
[34] Cell read out distinguishes between four levels of charge storage to provide two storage bits (Multilevel cell MLC). Progression to 8 or 16 levels is potentially possible but maintaining reasonable V t, read speed and array efficiency beyond 2-bit/cell are challenging. Unlike NAND Flash where density is a key competitive advantage, performance and reliability tend to hold higher importance than density for NOR Flash, thus the pressure to higher level MLC is not as strong as for NAND Flash.
[35] Virtual ground array uses junction isolation and buried diffusion for bit line, thus requires no STI isolation or bit line contact in the cell. In principle the area factor for VG array can be as small as 4F 2, compared to ~ 10F2 for an array with cell STI isolation and contact. The scaling of buried diffusion is a difficult challenge and junction isolation produces more leakage paths and complicates the design. Large R&D effort is needed to implement VG array and overcome its shortcomings to take advantage of its smaller cell size.
[36] Charge trapping device for NOR application uses mainly a SONOS structure and thus often is confused with the conventional SONOS device. Conventional SONOS is more suitable for a NAND array. The device is programmed by Fowler-Nordheim tunneling of electrons from the substrate and the charge is stored in the SiN layer of SONOS. Since the electrons are stored in deep SiN traps it is difficult to de-trap by Fowler-Nordheim tunneling. Instead, a very thin (2–3 nm) tunnel oxide is used to allow substrate hole tunneling into the SiN to erase the device. However, such thin tunnel oxide also allows direct tunneling of holes from the substrate even under weak electric field produced by the stored electrons and good data retention is difficult to achieve. NROM is a device proposed to solve the
. Using channel hot electron to program the cell, electrons are stored in the SiN layer near the edge of S/D junctions. To erase, band-to-band tunneling generated hot holes are injected into the SiN. A relatively thick (4–5 nm) tunnel oxide is adopted and this solves the data retention issue. NROM also has the advantage of storing two bits of information in one device (source side and drain side) and it applies a reverse read method to distinguish the different states. NROM is built on a virtual ground array architecture and has a relatively small cell size. It is used not only for NOR Flash, but also for some data storage applications even though it is not a NAND structure.
[37] CT NOR SLC Flash stores one bit on the source side and one bit on the drain side, or 2-bit/cell. The MLC Flash stores 2 bits on the source side and two bits on the drain side, thus 4-bit/cell.
[39] Because electrons are trapped in deep levels in SiN, the tunnel oxide can be scaled more aggressively than for floating gate device.
[40] Reducing the thickness of charge trapping SiN will reduce the EOT but will degrade trapping efficiency.
[41] Interpoly dielectric must be thick enough to assure retention.
[42] This entry is the critical dimension “F” within the FeRAM cell for stand-alone memory devices (not embedded devices).
[45] The geometry of the capacitor is a key factor in determining cell size. Stacked planar films are expected to be replaced by more efficient 3D structures.
[47] This is the actual effective area of the capacitor. It is larger than the footprint for 3D capacitor because of the utilization of area in the third dimension.
[48] This ratio of the effective area to the footprint gives a measure of the impact of utilization of the third dimension.
[52] MRAM devices are expected to lag the CMOS current technology up until 45 nm half pitch in 2010. This entry provides the F value for designs in the indicated time period.
[53] The MRAM switching field is the magnetic intensity H required to change the direction of magnetization of the cell.
[55] MRAM active bit area is the area of the magnetic material stack within the cell. It represents the “A” in the R*A product.
[38] Although physically storing two bits in the same device, the gate length scaling is not limited by left-right bit interference. The scaling is limited by the same factors for floating gate device - junction breakdown voltage and short channel effect. 4-bit/cell MLC device scaling for MLC may be more gradual than SLC.
multiples of the FeRAM implementation technology.
[44] FeRAM cell structures have migrated to one transistor, one capacitor (1T1C) formats. (Refs. J.H. Park, et al., "Fully Logic Compatible (1.6V Vcc, 2 Additional FRAM Masks) Highly Reliable Sub 10F 2 Embedded FRAM with Advanced Direct Via Technology and Robust 100 nm thick MOCVD PZT Technology", 2004 IEDM, 23.7.1, pp. 591-594. Y. M. Kang et al., "Sub-1.2V Operational, 0.15µm/12F 2 Cell FRAM Technologies for Next Generation SoC Applications", 2005 Symposium on VLSI Technology, 6B-4, pp. 102-103.) Other alternative configurations are under investigation such as Chain-FeRAM. (Refs. H. Kanaya et al., "A 0.602µm2 Nestled Chain Cell Structure Formed by One Mask Etching Process for 64Mbit FeRAM,” 2004 Symposium for VLSI Technology, pp. 150-151. N. Nagel et al., "New Highly Scalable 3 Dimensional Chain FeRAM Cell with Vertical Capacitor,” 2004 Symposium on VLSI Technology, pp. 146-147.)
[46] This is the footprint of the capacitor in micrometers squared. It is this area that constitutes the capacitor area contribution to the cell size. For 2005–2006 ~19F 2, for 2007 - 2009 ~16F2, and for 2010–2020 ~10F2 (3D capacitor) are assumed.
) applied to the capacitor. Low voltage operation is a difficult key design issue. Generally the ferroelectric film thickness needs to be decreased in order to reduce the V op, with great technological challenges. (Ref. D. C. Yoo et al., /MOCVD PZT Capacitor Technology", 2005 Symposium on VLSI Technology, 6B-3, pp. 100-101.)
is a useful design parameter. It is equal to the cell minimum switching charge divided by the capacitor actual effective area. The capacitor voltage is taken as V op.
[51] FeRAM is a destructive read-out technology, so every read is accompanied by a write to restore the data. Endurance cycles are taken as the sum of all read and all write cycles. For FeRAM to compete with DRAM and SRAM the cycle endurance should be about 1E15. Testing time is a serious concern. Note that operation at 100 MHz for 3 years would accumulate 1E16 cycles.
[54] MRAM switching energy per bit is calculated as (write current * power supply voltage * write time). It is preferred to use the median value of switching energy measured on a multi-megabit array. A good estimate of power drain is (switching energy * number of
[56] MRAM resistance-area product (i.e., the R*A product) is an intrinsic property of the magnetic material stack that provides a convenient basis for comparing cells of different sizes. The R*A product can be computed by measuring the effective low state resistance (R low)
[58]This is the critical dimension, F.
[61] The expected “typical” PCRAM cell size with BJT access device is presented in micrometers squared.
[62] The expected “typical” PCRAM cell size with nMOSFET access device is presented in micrometers squared.
[63] PCRAM is capable of MLC multi-bit/cell operation since the resistance ratio between amorphous and crystalline state is typically 100–1,000. This entry is the expected number of MLC bits per cell.
[64] The expected cell size per MLC bit for the PCRAM with BJT cell. It is the physical cell size divided by the number of MLC bits per cell.
[65] The expected cell size per MLC bit for the PCRAM with nMOSFET cell. It is the physical cell size divided by the number of MLC bits per cell.
[66] PCRAM phase change element must be substantially smaller than the technology F to have efficiency reset operation with reasonable current. This entry is the expected dimension for the phase change element in nanometers.
[67] PCRAM phase change volume is a key factor for device design and peak power requirement. This entry is the expected phase change volume in nanometer cubed.
[68] This entry is the expected reset current for PCRAM in microamperes.
[69] The set resistance is a key design factor for PCRAM read speed.
[71] This entry is the expected BJT emitter area that can provide the needed reset current, assuming the BJT current density is met.
[73] This entry is the expected nMOSFET gate width that can provide the needed reset current, assuming the MOSFET output current density is met.
. This ratio summarizes the difference between a logic ONE and a logic ZERO, and as such it represents the intrinsic capability of the magnetic stack. The magnetic tunnel junction resistance values
. This entry is the expected PCRAM cell area in multiples of the implementation technology F2. PCRAM requires significant reset current to change the phase-change element from crystalline to amorphous. A BJT transistor is capable of providing more current per unit area compared to a MOSFET, thus helps to reduce the cell size. Both BJT and nMOSFET access device cells are represented in this table. PCRAM is capable of MLC multi-bit per cell. This area factor is per cell, not per bit.
. This entry is the expected PCRAM cell area in multiples of the implementation technology F2. PCRAM requires significant reset current to change the phase-change element from crystalline to amorphous. A BJT transistor is capable of providing more current per unit area compared to a MOSFET, thus helps to reduce the cell size. An nMOSFET transistor has larger cell size in the near term years, but offers simple process and low voltage operation. Both BJT and nMOSFET access device cells are represented in this table. PCRAM is capable of MLC multi-bit per cell. This area factor is per cell, not per bit.
[70] This entry is the expected current density output from the BJT access device required to reset the PCRAM cell (from crystalline to amorphous state). It is a compromise between larger area BJT (which causes larger cell size) and higher output current (which requires
[72] This entry is the expected current density output from the nMOSFET access device required to reset the PCRAM cell (from crystalline to amorphous state). It is a compromise between larger width nMOSFET (which causes larger cell size) and higher output current
[74] This entry is the expected PCRAM data retention that will allow it to be used as a nonvolatile memory. Data retention mechanism for PCRAM is not yet thoroughly studied. Recent published data indicate >10 years of retention at elevated temperatures. (Refs. S. J. Ahn et al., "Highly Manufacturable High Density Phase Change Memory of 64Mb and Beyond,” 2004 IEDM, 37.2, pp. 907-910. A. L. Lacaita et al., "Electrothermal and Phase Change Dynamics in Chalcogenide-Based Materials,” 2004 IEDM, 37.3, pp. 911-914.)
[75] This entry is the expected PCRAM W/E cycling endurance. Recent published data indicate cycling endurance from 1E+9 to 1E+13. (Refs. S.J. Ahn et al., "Highly Manufacturable High Density Phase Change Memory of 64Mb and Beyond,” 2004 IEDM, 37.2, pp. 907-910. S. Lai et al., "Current Status of Phase Change Memory and Its Future,” 2003 IEDM, pp. 255-258.)
Long-term
2018 2019 2020 2021 2022
18 16 14 12 10
19 17 15 13 11
16 14 13 11 10
16 14 13 11 10
64 64 64 64 64
CT-3D CT-3D CT-3D CT-3D CT-3D
4 4 4 4 4
4.0/1.0 4.0/1.0 4.0/1.0 4.0/1.0 4.0/1.0
4 4 4 4 4
High-K High-K High-K High-K High-K
9-10 9-10 9-10 9-10 9-10
0.6–0.7 0.6-0.7 0.6-0.7 0.6-0.7 0.6-0.7
Metal Metal Metal Metal Metal
15-17 15-17 15-17 15-17 15-17
1E+04 1E+04 1E+04 1E+04 1E+04
5-10 5-10 5-10 5-10 5-10
4 4 4 4 4
4.0/1.0 4.0/1.0 4.0/1.0 4.0/1.0 4.0/1.0
SiO2 or ONO SiO2 or ONO SiO2 or ONO SiO2 or ONO SiO2 or ONO
3-4 3-4 3-4 3-4 3-4
SiO2 or Al2O3 SiO2 or Al2O3 SiO2 or Al2O3 SiO2 or Al2O3 SiO2 or Al2O3
6–8 6–8 6–8 6–8 6–8
SiN / High-K SiN / High-K SiN / High-K SiN / High-K SiN / High-K
4–6 4–6 4–6 3-4 3-4
Metal Metal Metal Metal Metal
15-17 15-17 15-17 15-17 15-17
1E+04 1E+04 1E+04 1E+04 1E+04
5-10 5-10 5-10 5-10 5-10
4 4 4 4 4
18 16 14 12 10
10-13 10-13 10-13 10-13 10-13
36 32 28 24 20
7 - 8 7 - 8 7 - 8 7 - 8 7 - 8
High-K High-K High-K High-K High-K
7-9 6-8 6-8 6-8 6-8
0.6–0.7 0.6-0.7 0.6-0.7 0.6-0.7 0.6-0.7
6-8 6-8 6-8 6-8 6-8
10–15 9-14 8-13 7-12 6-10
1.00E+07 1.00E+07 1.00E+07 1.00E+07 1.00E+07
20 20 20 20 20
2 2 2 2 2
CC/VG CC/VG CC/VG CC/VG CC/VG
18 16 14 12 10
8-9 9-10 9-10 9-10 9-10
4.3/2.2 4.8/2.4 4.8/2.4 4.8/2.4 4.8/2.4
0.07 0.07 0.07 0.06 0.06
4 3.5 3.5 3.5 3.5
4-5 4 4 4 4
5–7 5–7 5–7 5–7 5–7
5–7 5–7 5–7 4-6 4-6
10–15 9-14 8-13 7-12 6-10
1.00E+06 1.00E+06 1.00E+06 1.00E+06 1.00E+06
10–20 10–20 10–20 10–20 10–20
6 6 6 6 6
90 65 65 65 65
90 65 65 65 65
14 12 12 12 12
0.113 0.051 0.051 0.051 0.051
1T1C 1T1C 1T1C 1T1C 1T1C
3D 3D 3D 3D 3D
0.041 0.016 0.016 0.016 0.016
0.041 0.016 0.016 0.016 0.016
0.100 0.069 0.069 0.069 0.069
2.46 4.25 4.25 4.25 4.25
1.00 0.70 0.70 0.70 0.70
30 30 30 30 30
>1.0E16 >1.0E16 >1.0E16 >1.0E16 >1.0E16
10 Years 10 Years 10 Years 10 Years 10 Years
22 16 16 16 16
16 17 16 17 16
0.0077 0.0044 0.0041 0.0044 0.0041
35 35 35 35 35
20 20 20 20 20
0.007 0.005 0.005 0.005 0.005
0.6 0.6 0.6 0.6 0.6
70 70 70 70 70
>10 >10 >10 >10 >10
>3e16 >3e16 >3e16 >3e16 >3e16
>10 >10 >10 >10 >10
16 14 12 10 8
4.0 4.0 4.0 4.0 4.0
6.0 6.0 6.0 5.5 5.5
1024 784 576 480 340
1536 1176 864 650 450
4 4 4 4 4
256 196 144 120 85
384 294 216 162 112
10 9 8 8 7
4,700 3,200 2,000 1,300 900
27 22 18 15 13
19.74 23.11 27.72 31.00 35.00
2.10E+07 2.20E+07 2.40E+07 2.50E+07 2.70E+07
201 154 113 91 73
2.88 2.88 2.88 2.88 2.88
16 14 12 10 9
>10 >10 >10 >10 >10
1.0E+15 1.0E+15 1.0E+15 1.0E+15 1.0E+15
[1] NAND Flash has surpassed CMOS and DRAM technology since 2005. This entry provides the F value for designs in the indicated time period.
[2] NAND Flash architecture consists of bit line strings of a number of storage devices. Long bit line strings reduce the overhead for bit line transistors and increase the packing density, however, at the expense of higher overall resistance and consequently lower read
[3] Because of the difficulty in maintaining high gate coupling ratio and preventing cross talk between neighboring cells, NAND technology is forecasted to migrate gradually from floating gate devices (FG) to charge trapping devices (CT). (Ref: K. Kim, "Technology for sub 50nm DRAM and NAND Flash manufacturing,” in Tech. Digest International Electron Devices Meeting, pp. 539-543, 2005.) The statistical fluctuation limit of storing too few electrons imposes new challenges to data retention (Ref: G. Molas, et al., "Impact of few electron phenomena on floating gate memory reliability,” Tech. Digest 2004 International Electron Devices Meeting, pp. 877-880, 2004.) and 3D integration of multiple layers of devices may be required to continue the scaling. (Refs: S.H. Lee et al., "Three dimensionally stacked NAND Flash memory technology using stacking single crystal Si layers on ILD and TANOS structure for beyond 30nm node,” Tech. Digest 2006 International Electron Devices Meeting, pp. 37-40, 2006. E-K. Lai, et al., "A multi-layer stackable thin-film transistor
. It is possible to store more than 1 bit of information in a Flash cell but increasing the logic levels from (1, 0) to
[9] Low write and erase voltage is desirable but EOT for tunnel oxide and IPD must be decreased to allow lower W/E voltage without compromising W/E speed.
[11] Data retention is controlled by both tunnel oxide integrity and statistical distribution of the number of stored electrons. Both thinner tunnel oxide and fewer number of stored electrons in the long term contribute to the shorter retention forecast.
[18] High work function metal gate is the best to prevent gate injection. However, p-type polysilicon may become an interim solution because of its easy processing, low cost, and reasonably good performance.
[5] The scaling of tunnel oxide for NAND Flash faces the same challenge as that for NOR Flash. However, the use of error code correction (ECC) in NAND tolerates tunnel oxide defects to a higher level than that for NOR, thus allowing tunnel oxide of 6–7 nm. Currently
[6] ONO has been used as the interpoly dielectric (IPD) up to now and will continue in the near future. However, below 40 nm the spacing between floating gates becomes too narrow to fill effectively with ONO and word line polysilicon, and the loss of sidewall control gate to floating gate coupling will severely degrade the gate coupling ratio (GCR) and the device becomes inoperable. Since it is impossible to create additional space, higher dielectric constant IPD or charge trapping (CT) device must be used. Here the path of migrating
[7] Gate coupling ratio (GCR) is defined as (control gate to floating gate capacitance)/(total floating gate capacitance). GCR represents the fraction of voltage drop across the tunnel oxide and must be higher than 0.6 for the device to function during write and erase operations. High GCR is normally achieved by wrapping the control gate around the sidewalls of the floating gate. This requires tall floating gate and the cross talk along the bit line direction with neighboring cells is a challenge for MLC operation. At below 40 nm , the
[8] n-type polysilicon (and polycide) gate has been used for the control gate so far and will continue in the near future. The introduction of high-κ IPD, with a lower barrier height to Si, will cause severe gate injection during erase operation and high work function
[10] Write and erase cycling endurance reflects the tunnel oxide damage caused by repeated passing of charges under high electric field. Scaling does not worsen the oxide damage, however, larger array strains the ECC capability and the tolerance for defects is thus reduced. High-κ IPD may also trap charge and cause degradation. Current projection is gradually reduced cycling endurance for future technology. Note that this is not suitable for certain applications, e.g., solid-state drive storage that requires high cycling endurance.
[12] Multi-level cell (MLC) with 4 logic levels (2-bit/cell) is commonly used for NAND Flash today and devices with 8 logic levels (3-bit/cell) and 16 logic levels (4-bit/cell) are being developed. 8-bit/cell MLC device requires 256 logic levels and so far seems beyond reach
to increase voltage drop across the tunnel oxide and high work function metal gate to stop gate injection. (Ref: C.H. Lee, et al., "A novel with TaN metal gate for multi-giga bit Flash memories,” Tech. Digest 2003 International Electron Devices Meeting, pp. 613-616, 2003.) Barrier engineering uses composite tunneling barriers to allow easy erase operation by substrate
hole tunneling, yet prevents low field hole direct tunneling during retention. (Ref: H.T. Lue, et al., "BE-SONOS: a bandgap engineered SONOS with excellent performance and reliability,” Tech. Digest 2005 International Electron Devices Meeting, pp. 555-558, 2005.)
, although Al2O3 and other composite
[16] SiN is the most common and best known charge trapping layer with relatively deep electron traps that provide good data retention. Other exotic high-κ material with even deeper traps may be used in the long term years. (Ref: A. Chin, et al., "Low voltage high speed SiO2/AlGaN/AlLaO3/TaN memory with good retention,” Tech. Digest 2005 International Electron Devices Meeting, pp. 165-168, 2005.) Note that for CT device the charge loss mechanism may be mainly substrate hole tunneling in low field and thus deeper traps do not
[17] Trapping efficiency in SiN seems thickness dependent. (Ref: H.T. Lue, et al., Proc. 2007 International Reliability Physics Symposium, 2007). Therefore, thinner SiN or other high-κ trapping layers forecasted for long-term years may suffer from reduced programming
[19] The mechanism for endurance degradation for CT devices is not well understood yet. Unlike floating gate device, CT devices are not sensitive to tunnel oxide damage since the charge is stored in discrete traps and one weak spot does not cause all stored charge to
[20] The mechanism for data retention loss for BE device seems understood. (Ref: H.T. Lue, et al., "Reliability model of bandgap engineered SONOS (BE-SONOS),” Tech. Digest 2006 International Electron Devices Meeting, pp. 495-498, 2006.) Data retention mechanism
[21] NOR Flash traditionally falls behind CMOS and DRAM but has caught up in recent years and is now on par with DRAM.
[29] The gate coupling ratio (GCR) is the (control gate to floating gate capacitance)/(total floating gate capacitance). GCR must be greater than about 0.6 for proper device operation.
[32] E/W endurance requirements vary with the specifics of an application, but 1E5 cycles have been accepted as the historical minimum acceptable level for a useful NOR product.
[22] High-κ interpoly dielectric is projected at 32 nm and beyond to achieve gate coupling ratio of >0.6 but this offers only limited help to the area factor. (Ref: 2005 Symposium on VLSI Technology, 11B-3, E. S. Cho, et al.,” Hf-silicate Inter-Poly Dielectric Technology
[23] Although virtual ground (VG) array may significantly decrease the cell size in the near term years (Ref: 2005 Symposium on VLSI Technology, 11B-1, R. Koval, et al. "Flash ETOX Virtual Ground Architecture: A Future Scaling Direction,” pp. 204-205.), this effect
[24] Although non-planar devices (such as FinFET) are being developed for future Flash scaling, their impact has not been included in the current table. The dilemma of filling ONO and control and floating gates in the narrow gap between adjacent vertical devices has
[25] Both the cell size and the gate length for NOR Flash have been more aggressively scaled recently with an area factor of approximately 10F2. (Refs: 2005 ISSCC, "A 90 nm 512Mb 166 MHz Multilevel Cell Flash Memory with 1.5MBytes/s Programming,” pp. 54-55. Cell Size,” pp. 91-92. 2004 Symposium on VLSI Technology, "A 70nm NOR Flash Technology with 0.049µm2 Cell Size,” pp. 238-239.)
[26] NOR Flash uses channel hot electron programming which requires steep junction profile to generate, with the consequence of difficulties in controlling the short channel effect, but yet the NOR architecture is vulnerable to device leakage. Since the tunnel oxide thickness cannot be scaled, controlling the short channel effect imposes a very difficult challenge for scaling. The gate length generally is substantially larger than F in recent years. Despite this difficulty, the area factor has been maintained at 10F 2 in recent years (see
[28] ONO has been used as the interpoly dielectric up to now and most likely in the near future. However, at 32 nm and beyond high-κ IPD may be necessary to maintain GCR at 0.6 or above. Currently, the GCR is achieved by wrapping the control gate over the sidewalls of the floating gate thus increasing the control gate to floating gate capacitor area. At 32 nm or below, the gap between adjacent gloating gates becomes too narrow for the ONO and control gate to fill in and this semi-vertical structure will cease to function.
[30] This is the highest voltage relative to ground seen in the cell array, usually supplied by on-chip charge pumping circuits. Low voltage is desired to reduce the charge pumping circuit overhead and simplify processing. The introduction of high-κ IPD will help to reduce
[31] In principle the read current decreases with scaling at a rate W/(L*Cox) to prevent voltage overdrive (read disturb). Since access time depends critically on the read current and is an important performance parameter for NOR Flash the read current decreases slower
[33] Retention is a defect related parameter rather than an intrinsic device characteristic. Improvement in defect control and accumulation of device history is expected to eventually allow specification of 20 years retention. Also, it should become possible to accept a
, read speed and array efficiency beyond 2-bit/cell are challenging.
for an array with cell STI isolation and contact. The scaling of buried diffusion is a difficult challenge and junction isolation produces more leakage paths and complicates the design. Large R&D effort is needed to implement VG array and overcome its shortcomings to take advantage of its smaller cell size.
[36] Charge trapping device for NOR application uses mainly a SONOS structure and thus often is confused with the conventional SONOS device. Conventional SONOS is more suitable for a NAND array. The device is programmed by Fowler-Nordheim tunneling of electrons from the substrate and the charge is stored in the SiN layer of SONOS. Since the electrons are stored in deep SiN traps it is difficult to de-trap by Fowler-Nordheim tunneling. Instead, a very thin (2–3 nm) tunnel oxide is used to allow substrate hole tunneling into the SiN to erase the device. However, such thin tunnel oxide also allows direct tunneling of holes from the substrate even under weak electric field produced by the stored electrons and good data retention is difficult to achieve. NROM is a device proposed to solve the
. Using channel hot electron to program the cell, electrons are stored in the SiN layer near the edge of S/D junctions. To erase, band-to-band tunneling generated hot holes are injected into the SiN. A relatively thick (4–5 nm) tunnel oxide is adopted and this solves the data retention issue. NROM also has the advantage of storing two bits of information in one device (source side and drain side) and it applies a reverse read method to distinguish the different states. NROM is built
[37] CT NOR SLC Flash stores one bit on the source side and one bit on the drain side, or 2-bit/cell. The MLC Flash stores 2 bits on the source side and two bits on the drain side, thus 4-bit/cell.
[39] Because electrons are trapped in deep levels in SiN, the tunnel oxide can be scaled more aggressively than for floating gate device.
[40] Reducing the thickness of charge trapping SiN will reduce the EOT but will degrade trapping efficiency.
[41] Interpoly dielectric must be thick enough to assure retention.
[42] This entry is the critical dimension “F” within the FeRAM cell for stand-alone memory devices (not embedded devices).
[45] The geometry of the capacitor is a key factor in determining cell size. Stacked planar films are expected to be replaced by more efficient 3D structures.
[47] This is the actual effective area of the capacitor. It is larger than the footprint for 3D capacitor because of the utilization of area in the third dimension.
[48] This ratio of the effective area to the footprint gives a measure of the impact of utilization of the third dimension.
[52] MRAM devices are expected to lag the CMOS current technology up until 45 nm half pitch in 2010. This entry provides the F value for designs in the indicated time period.
[53] The MRAM switching field is the magnetic intensity H required to change the direction of magnetization of the cell.
[55] MRAM active bit area is the area of the magnetic material stack within the cell. It represents the “A” in the R*A product.
[38] Although physically storing two bits in the same device, the gate length scaling is not limited by left-right bit interference. The scaling is limited by the same factors for floating gate device - junction breakdown voltage and short channel effect. 4-bit/cell MLC device
Embedded FRAM with Advanced Direct Via Technology and Cell FRAM Technologies for Next Generation SoC Applications", 2005 Symposium on VLSI Technology, 6B-4, pp. 102-103.) Other
Nestled Chain Cell Structure Formed by One Mask Etching Process for 64Mbit FeRAM,” 2004 Symposium for VLSI Technology, pp. 150-151. N. Nagel et al.,
(3D capacitor) are assumed.
, with great technological challenges. (Ref. D. C. Yoo et al.,
[51] FeRAM is a destructive read-out technology, so every read is accompanied by a write to restore the data. Endurance cycles are taken as the sum of all read and all write cycles. For FeRAM to compete with DRAM and SRAM the cycle endurance should be about
[54] MRAM switching energy per bit is calculated as (write current * power supply voltage * write time). It is preferred to use the median value of switching energy measured on a multi-megabit array. A good estimate of power drain is (switching energy * number of
[56] MRAM resistance-area product (i.e., the R*A product) is an intrinsic property of the magnetic material stack that provides a convenient basis for comparing cells of different sizes. The R*A product can be computed by measuring the effective low state resistance (R low)
[58]This is the critical dimension, F.
[61] The expected “typical” PCRAM cell size with BJT access device is presented in micrometers squared.
[62] The expected “typical” PCRAM cell size with nMOSFET access device is presented in micrometers squared.
[63] PCRAM is capable of MLC multi-bit/cell operation since the resistance ratio between amorphous and crystalline state is typically 100–1,000. This entry is the expected number of MLC bits per cell.
[64] The expected cell size per MLC bit for the PCRAM with BJT cell. It is the physical cell size divided by the number of MLC bits per cell.
[65] The expected cell size per MLC bit for the PCRAM with nMOSFET cell. It is the physical cell size divided by the number of MLC bits per cell.
[66] PCRAM phase change element must be substantially smaller than the technology F to have efficiency reset operation with reasonable current. This entry is the expected dimension for the phase change element in nanometers.
[67] PCRAM phase change volume is a key factor for device design and peak power requirement. This entry is the expected phase change volume in nanometer cubed.
[68] This entry is the expected reset current for PCRAM in microamperes.
[69] The set resistance is a key design factor for PCRAM read speed.
[71] This entry is the expected BJT emitter area that can provide the needed reset current, assuming the BJT current density is met.
[73] This entry is the expected nMOSFET gate width that can provide the needed reset current, assuming the MOSFET output current density is met.
. This ratio summarizes the difference between a logic ONE and a logic ZERO, and as such it represents the intrinsic capability of the magnetic stack. The magnetic tunnel junction resistance values
. PCRAM requires significant reset current to change the phase-change element from crystalline to amorphous. A BJT transistor is capable of providing more current per unit area compared to a MOSFET, thus helps to reduce the cell size. Both BJT and nMOSFET access device cells are represented in this table. PCRAM is capable of MLC multi-bit per cell. This area factor is per cell, not per bit.
. PCRAM requires significant reset current to change the phase-change element from crystalline to amorphous. A BJT transistor is capable of providing more current per unit area compared to a MOSFET, thus helps to reduce the cell size. An nMOSFET transistor has larger cell size in the near term years, but offers simple process and low voltage operation. Both BJT and nMOSFET access device
[70] This entry is the expected current density output from the BJT access device required to reset the PCRAM cell (from crystalline to amorphous state). It is a compromise between larger area BJT (which causes larger cell size) and higher output current (which requires
[72] This entry is the expected current density output from the nMOSFET access device required to reset the PCRAM cell (from crystalline to amorphous state). It is a compromise between larger width nMOSFET (which causes larger cell size) and higher output current
[74] This entry is the expected PCRAM data retention that will allow it to be used as a nonvolatile memory. Data retention mechanism for PCRAM is not yet thoroughly studied. Recent published data indicate >10 years of retention at elevated temperatures. (Refs. S. J. Ahn et al., "Highly Manufacturable High Density Phase Change Memory of 64Mb and Beyond,” 2004 IEDM, 37.2, pp. 907-910. A. L. Lacaita et al., "Electrothermal and Phase Change Dynamics in Chalcogenide-Based Materials,” 2004 IEDM, 37.3, pp. 911-914.)
[75] This entry is the expected PCRAM W/E cycling endurance. Recent published data indicate cycling endurance from 1E+9 to 1E+13. (Refs. S.J. Ahn et al., "Highly Manufacturable High Density Phase Change Memory of 64Mb and Beyond,” 2004 IEDM, 37.2, pp. 907-
Table PIDS6 Reliability Difficult Challenges
Difficult Challenges ≥ 22 nm Summary of Issues
Transistor Reliability Time dependent dielectric breakdown
Negative bias temperature instability
Threshold voltage shifts due to traps, carrier injection, program or erase
Mobility degradation due to mechanical stress relaxation or interface state density change
New or changed failure mechanisms (TDDB, PBTI, NBTI< moisture absorption, etc.) resulting from high κ/metal gate
Interconnect Reliability Copper electromigration and stress voiding in scaled interconnects (lines and vias)
Electrical breakdown of interconnect dielectrics, especially low κ and ultra low κ
Moisture absorption/transport due to voids in porous low κ dielectrics
Cu (ionic) migration through cracked or thin barrier metals
Packaging Reliability New failure mechanisms associated with Pb-free solders and new mold compounds
Electromigration in package traces, vias, and bumps
Impact of multichip modules and stacked dies on failure rate
Solder ball electromigration, for example in CSP and flip chip
Radioactive contaminants in packaging materials
Reliability in Extreme and/or Critical Applications
Automotive (define mission profile for HOT underhood versus passenger and substantial cycling)
Military (rugged versus shock and dust, highly diverse environmental requirements)
Space, i.e., radiation hard
Aeronautical (singe event effects tolerant and large, fast temperature swings)
Medical (corrosive, hermeticity, and safety)
Impact of Variability on Reliability Statistic variation growing larger and defect size is comparable to feature size: Distribution of dopant atoms; subtle ultra-thin gate oxide defects; line edge roughness and other litho "fidelity" issues; surface scattering
How to cope with cost-effective screens and qualifications that capture some "good" units
Design for Reliability in face of large percentage process variability
How to use yield to drive reliability
Difficult Challenges<22 nm Summary of Issues
Reliability of novel devices, structures, materials and applications
ITRS proposes many new materials and structures, yet currently very little known about failure mechanisms
Need to have reliability characterization in place well in advance of application to develop appropriate reliability requirements and qualification procedures
Design for Reliability tools
ORTCINDEX
2007 ITRS Chapters
2008INDEX
Table PIDS6 Reliability Difficult Challenges
Difficult Challenges ≥ 22 nm Summary of Issues
Transistor Reliability Time dependent dielectric breakdown
Negative bias temperature instability
Threshold voltage shifts due to traps, carrier injection, program or erase
Mobility degradation due to mechanical stress relaxation or interface state density change
New or changed failure mechanisms (TDDB, PBTI, NBTI< moisture absorption, etc.) resulting from high κ/metal gate
Interconnect Reliability Copper electromigration and stress voiding in scaled interconnects (lines and vias)
Electrical breakdown of interconnect dielectrics, especially low κ and ultra low κ
Moisture absorption/transport due to voids in porous low κ dielectrics
Cu (ionic) migration through cracked or thin barrier metals
Packaging Reliability New failure mechanisms associated with Pb-free solders and new mold compounds
Electromigration in package traces, vias, and bumps
Impact of multichip modules and stacked dies on failure rate
Solder ball electromigration, for example in CSP and flip chip
Radioactive contaminants in packaging materials
Reliability in Extreme and/or Critical Applications
Automotive (define mission profile for HOT underhood versus passenger and substantial cycling)
Military (rugged versus shock and dust, highly diverse environmental requirements)
Space, i.e., radiation hard
Aeronautical (singe event effects tolerant and large, fast temperature swings)
Medical (corrosive, hermeticity, and safety)
Impact of Variability on Reliability Statistic variation growing larger and defect size is comparable to feature size: Distribution of dopant atoms; subtle ultra-thin gate oxide defects; line edge roughness and other litho "fidelity" issues; surface scattering
How to cope with cost-effective screens and qualifications that capture some "good" units
Design for Reliability in face of large percentage process variability
How to use yield to drive reliability
Difficult Challenges<22 nm Summary of Issues
Reliability of novel devices, structures, materials and applications
ITRS proposes many new materials and structures, yet currently very little known about failure mechanisms
Need to have reliability characterization in place well in advance of application to develop appropriate reliability requirements and qualification procedures
Design for Reliability tools
Table PIDS7 Reliability Technology Requirements
Year of Production 2007 2008 2009 2010
Early failures (ppm) (First 4000 operating hours) [1] 50–2000 50–2000 50–2000 50–2000
Long term reliability (FITS = failures in 1E9 hours) [2] 50–2000 50–2000 50–2000 50–2000
SRAM Soft error rate (FITs/MBit) 1000-2000 1000-2000 1000-2000 1000-2000
Relative failure rate per transistor (normalized to 2007 value) [3] 1 0.83 0.71 0.66
1 0.5 0.5 0.5
Manufacturable solutions exist, and are being optimized
Manufacturable solutions are known
Interim solutions are known ¿Manufacturable solutions are NOT known
Notes for Table PIDS7a and b:
[1] Failures during the first 4000 hours of operation (~1 year's use at 50% duty cycle). Early failures are associated with defects.
[2] Long term reliability rate applies for the specified lifetime of the IC.
[3] While the overall IC failure rate does not change with time, as the number of transistors per chip increases [from ORTC], the relative failure rate per transistor must decrease
Relative failure rate per m of interconnect (normalized to 2007 value) [4]
Reliability requirements vary with different applications. For many mainstream customers it will be sufficient to hold current reliability levels steady during this period of rapid technological change. However, other customers would like reliability levels to be improved. Degradation of current reliability levels is not acceptable. Reliability requirements are for the packaged device and include both chip and package related failure modes.
A reliability qualification can always be attempted with available knowledge. The better the knowledge the less risk in the qualification and vice versa. Yellow coloring indicates some risk. Striped indicates a greater risk (due to changed and possible new failure modes). Finally, red indicates an unspecified solution (e.g., what technology will be used for post-Cu) for which the reliability risk cannot be assessed until details about the solution are provided.
[4] As the length of interconnect per chip increases [from Interconnect Technology Requirements tables], the failure rate per m of interconnect must decrease. Even more important for reliability is the increase in the number of vias.
ORTCINDEX
2007 ITRS Chapters
2008INDEX
2011 2012 2013 2014 2015 2016 2017 2018 2019
50–2000 50–2000 50–2000 50–2000 50–2000 50–2000 50–2000 50–2000 50–2000
50–2000 50–2000 50–2000 50–2000 50–2000 50–2000 50–2000 50–2000 50–2000
1000-2000 1000-2000 1000-2000 1000-2000 1000-2000 1000-2000 1000-2000 1000-2000 1000-2000
0.57 0.51 0.46 0.4 0.37 0.31 0.29 0.26 0.23
0.25 0.25 0.25 0.12 0.12 0.12 0.06 0.06 0.06
[1] Failures during the first 4000 hours of operation (~1 year's use at 50% duty cycle). Early failures are associated with defects.
[2] Long term reliability rate applies for the specified lifetime of the IC.
[3] While the overall IC failure rate does not change with time, as the number of transistors per chip increases [from ORTC], the relative failure rate per transistor must decrease
Reliability requirements vary with different applications. For many mainstream customers it will be sufficient to hold current reliability levels steady during this period of rapid technological change. However, other customers would like reliability levels to be improved. Degradation of current reliability levels is not acceptable. Reliability requirements are for the packaged device
A reliability qualification can always be attempted with available knowledge. The better the knowledge the less risk in the qualification and vice versa. Yellow coloring indicates some risk. Striped indicates a greater risk (due to changed and possible new failure modes). Finally, red indicates an unspecified solution (e.g., what technology will be used for post-Cu) for which the
Technology Requirements tables], the failure rate per m of interconnect must decrease. Even more important for
2020 2021 2022
50–2000 50–2000 50–2000
50–2000 50–2000 50–2000
1000-2000 1000-2000 1000-2000
0.2
0.03
0.18 0.16
0.03 0.03