My experience with my chia plotting (r9 3900x and i7-5820k)
Chia only uses more than 1 thread in phase 1. So 60% of the time it is entirely single thread performance (including memory latency) limited or disk i/o. And in the multithreaded phase one, it doesn’t scale much past 3-4 threads. Both not great for a very “wide” beast TRipper.
- All core load frequency matters. My clocks dip the more I load up cpu on on AMD.
1b. Per core IPC also. 10900 has way higher IPC and all core speeds than the any Tripper atm( new ones will crush it once released). - intel 10xxx and older have low memory latency. (11th gen actually regressed on memory latency). Not sure if this is an issue. But a ryzen benefits from faster ram(and therefore faster infinity fabric)
- 10900 has ample cores for the multithreaded phase 1…and they are fast.
So all in all per plot speed TR sucks compared to a 10900. But if you throwing enough plotting nvme and ram on a TR… It will have more plots per day with more parallel plotting jobs. Despite the per plot speed difference.