This is the second of three posts following on from the tests with Sub NUMA Clustering where I was able to run some benchmarks on new hardware at UTS as part of their commissioning testing programme. I was super excited at this set, though as it’s the first time I’ve had my hands on a new AMD EPYC, for which there has been a lot of hype about their price/performance point. For these tests I was on an AMD EPYC 7551P running RedHat EL 7.x with the system default MPI and GCC (4.8.5). It was the standard motor bike simpleFoam case with 24M cells.
The runtime and relative run time graphs, as shown in Figure 1and Figure 2, respectively are not particularly interesting as the results appear pretty similar to every other OpenFOAM benchmark on quality hardware. That is, going from one to two cores is pretty linear but the returns drop off as the core count increases.
Figure 2 – Relative run time for the motor bike benchmark on an AMD EPYC 7551P.