“The performance of the EPYC 7002 and 7003 series CPUs is not as good as the E5 in terms of plotting speed. I tested them under the same network conditions, with identical memory and hard drives, and found the EPYC’s speed to be significantly slower than expected. For plotting a 7TB hard drive, the E5 can complete 5% in a day, while the EPYC only completes 1%. This is unexpected, especially since my EPYC has 64 cores and the E5 only has 18 cores. Given this, should the team consider optimizing the EPYC CPU? I’ve seen many EPYC users comment that its speed is too slow.”
I do not think there is anything special about Epyc that is unoptimized. There are many generations of E5 processors as well, so I’m not sure how to compare.
But otherwise it might depend on other hardware differences, also there were threads where pinning farmer to run on a particular NUMA node was helpful on those bigger machines.