STAC Report: STAC-A2 (derivatives risk) on 8 x Ampere GPUs in an NVIDIA DGX A100

NVIDIA DGX A100 system with new GPUs sets three records

19 October 2020

STAC recently performed STAC-A2 Benchmark tests on a stack consisting of CUDA 11 with 8 x NVIDIA A100 (Ampere) SXM 40 GiB GPUs in a DGX A100 server.

The STAC Report is now available here.

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).

NVIDIA wished to highlight several results from this report:

  • Compared to all publicly reported solutions to date, this GPU-based solution set three records:
    • the highest maximum paths (STAC-A2.β2.GREEKS.MAX_PATHS)
    • the highest maximum assets (STAC-A2.β2.GREEKS.MAX_ASSETS)
    • the fastest warm time in the baseline Greeks benchmark (STAC-A2.β2.GREEKS.TIME.WARM)
  • Compared to the best results for any non-NVIDIA-based single server solution, this solution had:
    • 2.15x the speed in the baseline Greeks benchmark (STAC-A2.β2.GREEKS.TIME.WARM vs. SUT ID INTC181012)
    • 1.78x the speed in the large Greeks benchmark (STAC-A2.β2.GREEKS.10-100k-1260.TIME.WARM vs. SUT ID INTC190903)
    • 14% higher energy efficiency (STAC-A2.β2.HPORTFOLIO.ENERG_EFF vs. SUT ID INTC190903)
  • Compared to the previous best results for NVIDIA-based solutions, this solution:
    • had 61% higher maximum paths (STAC-A2.β2.GREEKS.MAX_PATHS vs. SUT ID NVDA181105)
    • was 54% faster in the large Greeks benchmark (STAC-A2.β2.GREEKS.10-100k-1260.TIME.WARM vs. SUT ID NVDA181105)

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