The results are in, and GPUs are still the fastest solution on the planet for financial risk management.
This is according to the latest STAC-A2 audited test results. STAC-A2 is a risk management benchmark created by leading global banks working with the Securities Technology Analysis Center (STAC) in order to allow assessment of financial compute solutions. These new results come from a partnership between HPE, NVIDIA, and Red Hat.
The HPE ProLiant XL270d Gen9 server enables a dense node strategy for financial High Performance Computing, allowing up to 8 NVIDIA GPUs to efficiently populate a single HPC server. Red Hat worked with both HPE and NVIDIA to enable support for this cutting edge new hardware in the underlying operating system. Using Red Hat Enterprise Linux 7, a single HPE ProLiant XL270d Gen9 server with 8 NVIDIA Tesla V100 GPUs was benchmarked and the results are stunning :
- 8.9x the next best throughput
- 6.2x the next best time in warm runs of the Greeks benchmark
- 2.7x the next best energy efficiency
- 1.9x the next best maximum basket size
These comparisons are to the previous best of all publicly reported results to date on non-NVIDIA GPU based architectures by STAC (in all these cases a single server with 2 Intel® Xeon® Platinum 8180 Processors with 56 CPU cores total, but without GPUs).
With this type of dense node, a compute farm for portfolio risk calculations as modeled by the STAC-A2 benchmark can be realized with far fewer nodes achieving higher performance with lower operational costs.
Compared to all publicly reported results to date, this solution set new records in most performance and efficiency benchmarks.
This dense node solution allows an exciting strategy of scaling-up with NVIDIA GPU in a single HPE server, rather than scaling out to many more servers in the datacenter, thus enabling the highest performance at the lowest operating cost, and smallest datacenter footprint.