STAC Report: STAC-A2 (derivatives risk) on multi-chip Cascade Lake AP

2.67x the space efficiency and 1.25x the energy efficiency of a system using the single-chip Cascade Lake

11 September 2019

STAC recently performed STAC-A2 benchmark tests on a solution based on 2 x Intel Xeon Platinum 9242 (“Cascade Lake-AP”) CPUs. This high-end Xeon fits two 24-core dies into each socket, yielding 96 total physical cores in a 2-socket 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).

This stack consisted of the STAC-A2 Pack for Intel Parallel Studio XE (Rev M) with 2 x Intel Xeon Platinum 9242 processors and Red Hat Enterprise Linux 7.7 in an Intel Server System S9200WK SDP. The server had patches applied to mitigate Spectre & Meltdown security vulnerabilities.

• Compared to all publicly disclosed STAC-A2 results to date, this solution set a new record for space efficiency (STAC-A2.β2.HPORTFOLIO.SPACE_EFF).
• This beats the previous record, also held by Intel (SUT ID INTC181012), by 8.5%

• Compared to the most recently benchmarked solution using GPUs (SUT ID NVDA181105), this system:
  • Had over 1.8x the space efficiency (STAC-A2.β2.HPORTFOLIO.SPACE_EFF)
  • Was within 20% of the energy efficiency (STAC-A2.β2.HPORTFOLIO.ENERG_EFF)

• Compared to a similar 2-socket system using Intel Xeon Platinum 8280 (“Cascade Lake”) processors (SUT ID INTC190402), this system had:
  • 2.67x the space efficiency (STAC-A2.β2.HPORTFOLIO.SPACE_EFF)
  • 1.84x the throughput (STAC-A2.β2.HPORTFOLIO.SPEED)
  • 1.25x the energy efficiency (STAC-A2.β2.HPORTFOLIO.ENERG_EFF)
  • 1.58x the speed in warm runs of the large problem size and 1.21x the speed in cold runs (STAC-A2.β2.GREEKS.10-100k-1260TIME.COLD and STAC-A2.β2.GREEKS.10-100k-1260TIME.WARM, respectively)
  • 1.38x the speed in warm runs of the baseline problem size (STAC-A2.β2.GREEKS.TIME.WARM)
  • 13% more asset capacity (119 assets vs 105 assets) (STAC-A2.β2.GREEKS.MAX_ASSETS

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.