Initial Benchmarks Of The AMD EPYC 7601 On Ubuntu Linux

Written by Michael Larabel in Processors on 14 September 2017 at 03:00 PM EDT. Page 3 of 5. 41 Comments.

First up is the NAS Parallel Benchmark tests. The "EP" being, of course, for embarrassingly parallel. We see by the large margins how much better the EPYC 7601 does over the older Xeons now that the core count race is back on. The dual Xeon Gold 6138 system understandably comes out ahead given it has 80 threads versus 64 on the EPYC 7601. While this dual Xeon Gold system has 1.25x the thread count of the EPYC system, its EP.C performance is only better by 1.12x.

With NPB's Lower-Upper Gauss-Seidel solver, the EPYC 7601 manages to run faster than the Xeon Gold 6138 system even with the lower core/thread count but with a slightly higher base clock speed of 2.2GHz vs. 2.0GHz and a higher L3 cache size of 60MB vs. 27.5MB per socket. The EPYC 7601 also costs around $1000 USD less than two Xeon Gold 6138 processors.

When increasing the class size of the LU solver, the EPYC 7601 receded behind the Xeon Gold 6138 server but when forcing the NUMA interleaving policy, the performance was about tied with the Intel platform.

Parboil's OpenMP benchmarks benefit greatly from the NUMA interleaving policy and allow it to compete with or outperform the dual Xeon Gold 6138 system from Tyan.

With the Rodinia OpenMP tests too, the EPYC 7601 server shows strong competition against the dual Xeon Gold configuration and with the NUMA interleaving policy being forced, in some tests is significantly faster than the out-of-the-box configuration.


Related Articles