Power8 processors pack over twice the punch of their x86 counterparts when it comes to running financial applications.
Those were the findings presented by IBM to the benchmarking organisation STAC earlier this month. The STAC Summit in London attracted banking tech experts from all over Europe who heard how a Power8-based system set new performance records for financial workloads.
Tests were undertaken using STAC-A2 benchmarks which, according to STAC, give a more accurate view of workload performance compared to micro benchmarks or simple code loops. They were carried out with a model S824 using two 3.52 GHz processor cards and 1TB of DRAM running Red Hat Enterprise Linux version 7.
The S284 achieved 2.3 times the performance of a comparable Intel server with two 2.30GHz Xeon E5-2699 v3 Haswell processors. In fact, it also performed 1.7 times better than the x86 server that did best in the tests – a box with two Intel Xeon E5-2699 v3 processors with an Intel Xeon Phi 7120A coprocessor.
Furthermore, the S824 set new records for path scaling and asset capacity. Compared to a server with four 2.80 GHz Xeon E7-4890 v2 Ivy Bridge processors, it delivered 2.1 times the throughput and delivered a 16% increase in asset capacity.
In a blog post about the results, Sumit Gupta, an IBM HPC and OpenPower exec, pointed out that Power8 processors are based on high performance, multi-threaded cores. Each core of the S824 server runs up to eight simultaneous threads at 3.5 GHz.
Gupta said: “Power System S824 also has a very high bandwidth memory interface that runs at 192 GB/s per socket which is almost three times the speed of a typical x86 processor. These factors along with a balanced system structure including a large internal 8MB per core L3 are the primary reasons why financial computing workloads run significantly faster on Power8-based systems than alternatives.”
The solution stack used in the test included a STAC-A2 Pack for Linux on Power Systems, provided by IBM. This used IBM XL, a suite for C/C++ developers that included the C++ Compiler, Mathematical Acceleration Subsystem libraries (MASS), and the Engineering and Scientific Subroutine Library (ESSL).
There was just one fly in the ointment. According to earlier benchmark tests, the S824 could be beaten by a server from American manufacturer Supermicro containing two 3.0GHz 10-core Intel Xeon E5-2690 v2 processors and an Nvidia K80 GPU accelerator. Even then, it would only lag performance-wise by 10%.