Upload
lamkhuong
View
231
Download
3
Embed Size (px)
Citation preview
Confidential
SK hynix Inc. October 2014
© 2014 SK hynix Inc. This material is proprietary of SK hynix Inc. and subject to change without notice.
HBM: Memory Solution for High Performance Processors Kevin Tran and June Ahn
Agenda
“Completion of HBM1 Qualification” Announcement
Landscape of memory solution challenges
Why HBM is the suitable solution
HBM Architecture Review
Conclusion
SK hynix achieved Customer Qualification Level samples in Sep’14
Achievement of HBM1 Qualification
SK hynix TSV chronicle SK hynix World-First HBM Products
‘08 4Gb Flash
‘10 4Gb DRAM DDP-WLP
‘11 16Gb DRAM
9MCP
‘11 16/ 32GB
DIMM
‘11 4hi KGSD
WIO
’13 5mKGSD
HBM (Bottom View) (Top View ) (Section View )
Worldwide first HBM provider
Customer Qualification Samples shipping today
Volume production begins Q1’15
HBM2 design wins in progress with major SoCs
in multiple markets
Agenda
“Completion of HBM1 Qualification” Announcement
Landscape of memory solution challenges
Why HBM is the suitable solution
HBM Architecture Review
Conclusion
All DRAMs expect to face immense bandwidth requirement
Challenges#1) Bandwidth
• DRAM transistor and tester limit over 10Gbps
• Trade off with power consumption
• Severe die overhead
Gb/s/pin IO
GDDR5 7.0 X32
DDR4 3.2 X16
LPDDR4 4.2 X32
Data Rate/Pin Challenges
Density has increased by 1000X over the past two decades, Latency has decreased only by 56%
Challenges#2) Density/Latency
Density Mode tRC(ns)
1992 4Mb Cache DRAM 110
2002 256Mb DDR 6X
2003 1Gb DDR2 5X
2012 4Gb DDR3 4X~5X
2014 8Gb DDR4 4X
• Capacity / Cost Limitation
• DRAM Scaling Challenges
• Severe die overhead increase
Density/Latency Challenges
Reducing power and increasing performance are always trade-offs
Challenges#3) Power Efficiency
Power
Perf
orm
ance
• Trade-off between performance and power
Nor
mal
ized
DDR2
56 =
100
%
Power Efficiency Challenges
System level board design challenges for # of DRAM
Challenges#4) Form Factor
TSV is a revolutionary technology enabling next generation memories
Revolutionary Changes with TSV memories
▼ High Speed
▼ Lower Power
▼ High Density
▼ Small Form Factor
Agenda
“Completion of HBM1 Qualification” Announcement
Landscape of memory solution challenges
Why HBM is the suitable solution
HBM Architecture Review
Conclusion
Each application has different memory requirement, but most common are high bandwidth and density based on real time random operation.
General Memory Requirements
Graphics Networking
HPC
Datacenter
Real time Based
Random Operation
Bandwidth Density Power
Bandwidth Power Latency
Density Bandwidth Power Latency
Bandwidth Latency Power
Lower speed/pin and Cio reduce power consumption by 42% compared to GDDR5
Low Power
Pseudo channel improves tFAW by 60% compared to DDR4
Low Latency
1GB HBM package size is smaller than 1 tablet of aspirin
Small Form Factor
DDR4
HBM
DDR3
X37 bigger
X44 bigger
SoC
HBM 1GB
HBM 1GB
HBM 1GB
HBM 1GB
vs.
PKG Size 2.5D stacking size
X1
SiP enhances memory channel conditions
Silicon in Package
Items DRAM (Off-chip)
SiP (2.5D)
Physical dimension very large small
Signal Distance long short
Signal Loading high low
Driver Strength large small
IO Speed /Pin high low
Termination need no
Power Consumption high Low
Long Distance Loading, Power , C&R ↑
Speed, tDV ↓
Memory Channel Conditions DRAM vs. SiP
DRAM SiP
Agenda
“Completion of HBM1 Qualification” Announcement
Landscape of memory solution challenges
Why HBM is the suitable solution
HBM Architecture Review
Conclusion
TSV is the underlying technology for 3D Stack (High Density / Small size PKG / High speed)
Through Silicon Via - TSV
Die Pad
Die #2
Die #1
Bond wire
PKG Pad
PKG Pad
TSV
Die Pad
< Wire bonding PKG > < TSV PKG >
System-in-Package implementation with KGSD*
HBM 2.5D SiP Structure
FBGA
PHY TSV DA ball
DRAM Slice
DRAM Slice
DRAM Slice
DRAM Slice
Interposer
SoC
KGSD
PHY
HBM in 2.5D SiP
Side
Mol
ding
Side
Mol
ding
* KGSD (Known Good Stacked Die)
3D Memory(HBM)
Base die Silicon die
PKG Substrate
(Source : esilicon)
HBM Overall specification
Base Die
Interposer
HBM1 2Gb Density per DRAM die
1Gbps speed /pin
128GB/s Bandwidth
4 Hi Stack (1GB)
HBM2 8Gb per DRAM die
2Gbps speed/pin
256GBps Bandwidth/Stack
4/8 Hi Stack (4GB/8GB)
x1024 IO
1.2V VDD
KGSD w/ μBump
4 Core DRAM + 1 Base logic die (Chip on Wafer)
HBM Architecture Overview
DA POWER/TSV PHY
Core Die 0
Core Die 1
Core Die 2
Core Die 3
Base Logic Die
128 I/O
Microbump array
Probe PAD(Microbump
depopulated)
128 I/O
1024 TSV I/O
B3 B1
B2 B0
B7 B5
B6 B4
B3 B1
B2 B0
B7 B5
B6 B4
Items Target # of Stack 4(Core) + 1(Base)
Ch./Slice 2
Total Ch. for KGSD 8
IO/Ch. 128
Total IO/KGSD 1024(=128 x 8)
Address/CMD Dual CMD
Data Rate DDR
[1] D.U Lee, SK hynix, ISSCC 2014
Each HBM die has 2channels 1 channel consists of 128 TSV I/O with 2n prefetch
HBM Core Architecture
[2] D.U Lee, SK hynix, ISSCC 2014
B0 B1
YCTRL YCTRL
B2 B3
XCTR
LXC
TRL
C
B0 B1
YCTRL YCTRL
B2 B3XC
TRL
XCTR
L
C
B4 B5
YCTRL YCTRL
B6 B7
XCTR
LXC
TRL
C
B4 B5
YCTRL YCTRL
B6 B7
XCTR
LXC
TRL
C
DWORD 032 I/O
DWORD 132 I/O AWORD DWORD 2
32 I/ODWORD 3
32 I/O
B0 B1
YCTRL YCTRL
B2 B3
XCTR
LXC
TRL
C
B0 B1
YCTRL YCTRL
B2 B3
XCTR
LXC
TRL
C
B4 B5
YCTRL YCTRL
B6 B7
XCTR
LXC
TRL
C
B4 B5
YCTRL YCTRL
B6 B7
XCTR
LXC
TRL
C
DWORD 032 I/O
DWORD 132 I/O AWORD DWORD 2
32 I/ODWORD 3
32 I/O
CH-Left CH-Right
1 bank : 2 sub-banks(64Mb) non-shared I/O between sub-banks
Base die consists of 3 Areas – PHY, TSV, Test Port Area
HBM Base Die Architecture
HBM ballout area 6,050x3,264 μm
[3] D.U Lee, SK hynix, ISSCC 2014
• HBM is comprised of 8ch (2Channel/die) with 128DQs per channel. • Each channel(CH) consists of 2 Pseudo Channel(PS). Only BL4 is supported.
Pseudo Channel Concept
CH-0
B0 B1
B2 B3
B4 B5
B6 B7
64I/O
B0 B1
B2 B3
B4 B5
B6 B7
64I/OADDCMD
PS-CH0 PS-CH1CH-1
B0 B1
B2 B3
B4 B5
B6 B7
64I/O
B0 B1
B2 B3
B4 B5
B6 B7
64I/OADDCMD
PS-CH0 PS-CH1
CH-2 CH-3CH-4 CH-5
CH-6 CH-7
(Note : Pseudo channel is only applicable to HBM2)
Each pseudo channel share AWORD, but has separated banks & independent 64 I/Os
Pseudo Channel Mode
Channel # CMD
Channel # ACT#-RD#
128DQ D0 128
D1 128
D2 128
D3 128
16Bank / CH
Channel # CMD
CH #_PC0 ACT#0-RD#0
64DQ D0 64
D1 64
D2 64
D3 64
CH #_PC1 ACT#1-RD#1
64DQ D0 64
D1 64
D2 64
D3 64
16Bank/PC1
16Bank/PC0
• For Legacy Mode, Each channel has 2KB page size • Restriction of Gapless Bank Activation by tFAW(4 activate window) tFAW=30ns > 4Bank*tRRD=16ns Lower efficiency of Band Width
• Suppose tCK=2ns, tFAW=30ns, tRRD=4ns
Restriction of tFAW in Legacy mode
Gap
Gap
Gap
• Pseudo channel has reduced page size compared to Legacy mode. : 2KB 1KB • Lower Active Power(IDD0) by 1K Page size • Define tEAW (1KB x 8 ACT) instead of tFAW (2KB x 4 ACT) • Bandwidth improvement by more Activations during tFAW
Benefit of Pseudo Channel
Memory Die
Logic Layer Host I/F + Memory I/F + Base Logic/IP Block
Base Die Customization – Future HBM Concept
27
Customization to meet various requirements
Overcome Memory Scaling
- Timing - Refresh
- Parallel-to-Serial(P2S)/S2P - JTAG, PMBIST - Configuration Registers - Error Handling - …
Agenda
“HBM1 Qualification Sample” Announcement
Landscape of memory solution challenges
Why HBM is the suitable solution
HBM Architecture Review
Conclusion
Complicated TSV Ecosystem requires close collaboration
among all stakeholders
TSV Ecosystem
OSATs
Set Makers
IP Enablers
Memory Vendors
SiP Assy’ & Test
SoC & Interposer Business Model for SiP
PHY & MC IP Controller IP
KGSD/KGD
Foundries
Conclusion
HBM enables new memory subsystem architecture for Next Generation high performance processors. SK hynix is the leader in HBM technology
SK hynix HBM1 Customer Qualification samples are shipping
HBM addresses the major memory requirements with lower power, lower
latency, and smaller form factor.
HBM2 Pseudo channel improves bus utilization and system performance
Thank You