The Performance Impact Of Genoa-X's 3D V-Cache With The AMD EPYC 9684X
Those wishing to go through 130+ benchmarks in total for the AMD EPYC 9684X Genoa-X processor with 3D V-Cache on/off can see this result file for all my data in full.
Across that wide range of diverse Linux workloads, the EPYC 9684X was on average boosted by around 12% when utilizing the 3D V-Cache.
The CPU power consumption of the EPYC 9684X was 255 Watts on average across all of the tests when 3D V-Cache was enabled by default or dropped to just 250 Watts when 3D V-Cache was disabled. (The strange CPU peak power consumption numbers is likely a race condition as also seen at times with the Intel RAPL/PowerCap interface too.) Roughly a 5 Watt power cost to the 3D V-Cache on the EPYC 9684X isn't bad at all considering the significant speed-ups for many workloads.
The CPU peak frequency showed no major difference when 3D V-Cache was enabled.
The CPU core temperature was also the same to only about two degrees higher when 3D V-Cache was in use. Not surprising given the only ~5 Watt power difference observed during the testing process.
In my benchmarking of the AMD EPYC 9684X it continues to impress and shows terrific uplift in numerous workloads and scientific domains. From today's testing it's nice to see the direct correlation of 3D V-Cache on the performance and confirming its minimal impact on power consumption.
If you enjoyed this article consider joining Phoronix Premium to view this site ad-free, multi-page articles on a single page, and other benefits. PayPal or Stripe tips are also graciously accepted. Thanks for your support.