1
Agile Condor: Scalable High Performance Embedded Computing Architecture Objectives Enable scalable real-time processing for intelligence surveillance and reconnaissance (ISR) missions Fast technical refresh rate Design to support Group 3-5 DoD unmanned aerial systems (UAS) Weigh less than 60 pounds Consume less than 500 Watts of power Deliver over 15 GFLOPS/Watt (enabled by GPUs) Develop a rugged pod-based external payload Operations Concept Overview Vision Increase warfighter effectiveness by enhancing remotely piloted systems capabilities and expanding their capacity to create effects in the battlespace. Base on open industry standards Pod and Chassis Details ISR missions face elevated demands in order to effectively operate in contested and challenged environments Agile Condor will support a rapid response to unexpected challenges Future: Secure Bio-Inspired Computing Flight Tests Agile Condor predecessor (6U VPX HPEC) flight tested in July 2015 Greater system resilience, adaptability, autonomy and intelligence Oct. 2016 planned flight test for Agile Condor 12.8 GFLOPS/Watt (90% from GPUs) Embedded real-time experiments performed Mark Barnell, Courtney Raymond Air Force Research Laboratory, Rome, NY [email protected], [email protected] DJ Isereau, Chris Capraro SRC, Inc., North Syracuse, NY [email protected], [email protected] 14 slot 3U VPX conduction cooled chassis Existing pod baseline flight certified MIL-STD-8591 design for aircraft interface and aerodynamics Ambient air cooling thermal management Agile Condor brings distributed high-performance computing closer to the sensor DISTRIBUTION A. Approved for public release; distribution unlimited (88ABW-2016-0194) Patent applied for Latest generation Intel® Core™ i7 SBC 2 - NVIDIA® Maxwell™ GM107 GPUs CUDA Compute Capability v5.0 1096 GFLOPS each Delivers 88% of total processing COTS upgradable Presented at IEEE HPEC 2015 CONTACT NAME Mark Barnell: [email protected] POSTER P6292 CATEGORY: AEROSPACE & DEFENSE - AD01

Agile Condor: Scalable High Performance Embedded Computing ...on-demand.gputechconf.com/gtc/2016/posters/GTC_2016_Areospace_… · Agile Condor: Scalable High Performance Embedded

Embed Size (px)

Citation preview

Page 1: Agile Condor: Scalable High Performance Embedded Computing ...on-demand.gputechconf.com/gtc/2016/posters/GTC_2016_Areospace_… · Agile Condor: Scalable High Performance Embedded

Agile Condor: Scalable High Performance Embedded Computing Architecture

Objectives • Enable scalable real-time processing for

intelligence surveillance and reconnaissance (ISR) missions

• Fast technical refresh rate

• Design to support Group 3-5 DoD unmanned aerial systems (UAS)

• Weigh less than 60 pounds

• Consume less than 500 Watts of power

• Deliver over 15 GFLOPS/Watt (enabled by GPUs)

• Develop a rugged pod-based external payload

Operations Concept Overview

Vision Increase warfighter effectiveness by enhancing remotely piloted systems capabilities and expanding their capacity to create effects in the battlespace.

• Base on open industry standards

Pod and Chassis Details

• ISR missions face elevated demands in order to effectively operate in contested and challenged environments

• Agile Condor will support a rapid response to unexpected challenges

Future: Secure Bio-Inspired Computing Flight Tests • Agile Condor predecessor (6U VPX HPEC)

flight tested in July 2015

Greater system resilience, adaptability, autonomy and intelligence

• Oct. 2016 planned flight test for Agile Condor

• 12.8 GFLOPS/Watt (90% from GPUs)

• Embedded real-time experiments performed

Mark Barnell, Courtney Raymond Air Force Research Laboratory, Rome, NY

[email protected], [email protected]

DJ Isereau, Chris Capraro SRC, Inc., North Syracuse, NY

[email protected], [email protected]

14 slot 3U VPX conduction cooled chassis

• Existing pod baseline flight certified • MIL-STD-8591 design for aircraft interface and

aerodynamics • Ambient air cooling thermal management

Agile Condor brings distributed high-performance computing closer to the sensor

DISTRIBUTION A. Approved for public release; distribution unlimited (88ABW-2016-0194)

• Patent applied for

Latest generation Intel® Core™ i7 SBC 2 - NVIDIA® Maxwell™ GM107 GPUs • CUDA Compute Capability v5.0 • 1096 GFLOPS each • Delivers 88% of total processing

COTS upgradable

Presented at IEEE HPEC 2015

contact name

Mark Barnell: [email protected]

P6292

category: aerosPace & Defense - aD01