AMD EPYC 9754 Benchmarks For The 128-Core Bergamo

Even more benchmarks can be found on this OpenBenchmarking.org page for my launch day testing. When taking the geometric mean of all results tested successfully across all of the processors, here is the breakdown:

For all these workloads that can scale well, the EPYC 9754 2P was about 20% faster than the EPYC 9654 / 9684X processors. The 20% uplift is great as while it's going from 96 to 128 cores (33%), most workloads don't scale linearly especially when reaching very high thread counts and there is also the matter of Bergamo featuring the denser Zen 4C cores. The EPYC 9754 also has a lower base clock at 2.25GHz compared to 2.4GHz with the EPYC 9654 and a maximum boost clock of 3.1GHz compared to 3.7GHz with the EPYC 9654. Or when forcing all the EPYC CPUs to 400 Watt and power determinism mode, it was around a 13% difference. The EPYC 9754 Bergamo processor easily blew past the 60-core Xeon Platinum 8490H Sapphire Rapids processors in these benchmarks.

Across all of these benchmarks carried out, the EPYC 9754 2P on average had a 385 Watt power draw... In comparison the EPYC 9654 2P had a 447 Watt average and the EPYC 9684X 2P had a 464 Watt average. And need we mention the Xeon Platinum 8490H 60-core processor consuming even more power with a 568 Watt average. The EPYC 9754 power consumption results surpassed my expectations in frankly not expecting Zen 4C to deliver such power efficiency improvements while still performing so well.

AMD EPYC Bergamo is an extremely promising new option for cloud service providers or those wanting dense container/VM deployments while interested in reducing operational costs with leading power efficiency. Beyond VMs/containers, the 128 cores / 256 threads per socket showed to provide benefit as well for those having baremetal workloads like CPU-based rendering/ray-tracing that scale well along with numerous MPI and OpenMP scientific code-bases for maximizing the performance and leading power efficiency. With the AMD EPYC 9754 able to drop-in to existing AMD Socket SP5 server motherboards is icing on the cake with more EPYC 9004 series platforms and motherboards continuing to reach market and able to leverage that maturing ecosystem of hardware.

Running Bergamo under Linux was also a great and trouble-free experience with leveraging all of the common AMD Zen 4 enablement work already carried out in months prior to last year's Genoa launch. So both the hardware and software ecosystem are in good shape for Bergamo for those wishing to take on up to 512 threads per (2P) server.

Next year Intel will have Sierra Forest as their first E-core server solution providing up to 144 cores per socket. However, with being an E-core design will presumably lack HT (SMT) and AVX-512 support found with Bergamo. It should be an interesting battle next year as well as ultimately seeing how Bergamo competes with AmpereOne once those processors actually become available and there is independent performance analysis of those Arm server CPUs.

Those are the initial AMD EPYC 9754 "Bergamo" benchmarks I was able to prepare in time for launch day. In the days ahead are several more AMD Bergamo (and Genoa-X) follow-up articles including a look at the SMT impact on power and performance for the EPYC 9754 among other interesting benchmark pieces. Thanks to AMD and Intel for providing the processor samples for testing and benchmarking at Phoronix. If you didn't already, also check out today's AMD EPYC 9684X Genoa-X review.