If you think lowering the number of threads would yield higher NPS, I would try it otherwise I don't see what it would bring in "comparability"acepoint_de wrote: ↑Sat Feb 27, 2021 11:38 am @Ckappe: thx for your benchmarks. Is there a certain reason that you only published the output of 16 threads but not 1, 2 and 4 (which would at least give some comparability)?
Ciao
acepoint
I think its the same rationale as you did not test with 16 threads on your 8core M1... right...
It's sort of like comparing a 6-geared sports-car car with a 4-gear one to what the top-speed only on gear one is.
Benching cores could be fine for tuning purposes and optimization within the SAME architecture but provide little comparability between completely different architectures imhop.
Just for refence though, I did a completely meaningless 8 thread bench.
Avg cFish 8-thread values:
Asus Duo 15 - 4900H
"Cfish-18022021 AVX2.exe bench 10 8" (default=mixed)
===========================
Total time (ms) : 2,434
Nodes searched : 56,803,468
Nodes/second : 23,433,774
So the conclusion is that going from 8-16 threads yields about 40% on a 4900H for cFish. But it is kind of irrelevant for comparing different architected CPUs like the M1 imop.
Asus Duo 15 - 4900H
"Cfish-18022021 AVX2.exe bench 10 16" default=mixed)
==========================
Total time (ms) : 7,209
Nodes searched : 233,263,127
Nodes/second : 32,357,210