STAC Report: STAC-A2 on 4 x NVIDIA Tesla P100 GPUs

Quad-GPU system sets multiple records in efficiency and performance.

November 10, 2016

STAC recently performed STAC-A2 Benchmark tests on a Supermicro server containing four NVIDIA Tesla P100 PCIe GPU accelerator cards. The STAC Report is now available here.

This stack consisted of the NVIDIA STAC-A2 Pack for CUDA (Rev C) driving four NVIDIA Tesla P100 PCIe GPU accelerator cards in a Supermicro SYS-2028GR-TRT server with dual Intel Xeon E5-2698 v3 processors and CentOS 7.2.
 
STAC-A2 is the technology benchmark standard based on financial market risk analysis. Designed by quants and technologists from some of the world's largest banks, STAC-A2 reports the performance, scaling, quality, and resource efficiency of any technology stack that is able to handle the workload (Monte Carlo estimation of Heston-based Greeks for a path-dependent, multi-asset option with early exercise).

This system set many records among publicly reported systems, including:

• Highest energy efficiency: 99,793 options/kWh (STAC-A2.β2.HPORTFOLIO.ENERG_EFF)
  • Over 2.9x that of the next best system (INTC161016)
• Highest space efficiency: 24.7 options/hour per cubic inch (STAC-A2.?2.HPORTFOLIO. SPACE_EFF)
  • 37% more than the next best (INTC161016)
• Fastest portfolio speed: 25.0 options per second (STAC-A2.β2.HPORTFOLIO.SPEED)
  • Over 5.3x that of the next best system (INTC161016)
• Fastest warm time in the baseline Greeks benchmark (STAC-A2.β2.GREEKS.TIME.WARM)
  • Over 4x that of the next best system (INTC161016) when using all 4 GPUs
  • Over 30% faster than the same system when using only 1 GPU

​For details, please see the report at the link above. Premium subscribers also have access to the code used in this project and the micro-detailed configuration information for the solution. To learn about subscription options, please contact us.