hmm not the clocks bro. i mean not entirely.
its losing out to 3700x because the architecture on ipc is linked to caching which reduces latency of the infinity fab.

same-thing clock to clock a 3950x has better ipc..

for ryzen vs ryzen i would run test on low ram like 2400/2666 to show the core advantage.

the architecture for clock speed is pretty interesting. it does not scale like intel.

This post has been edited by cstkl1: May 15 2020, 06:44 PM

Yeah not entirely la but the clock helps heaps.

Latency between CCX and such, so many things involved but on my side I'm not totally clear what affects what most. There's also the cache.

Interesting idea, the thing is I always run at 3200 CL16 because it's the most common kit.

For other tests you'll see that I do 3733 CL16 (soon to be 3733 CL14) - 3733 with 1:1 IFB matching at 1866.

Everyone's usage experience varies so much. Going from 2666 CL15 to say 3733 CL16 - the difference can range from literally nothing to being super duper far apart.

the reason i suggest not to up the memory clock cause the higher tier cpu benefits its more with its larger cache.
its also very apparent on single threaded IPC improvement clock for clock say a 3700x vs 3950x.. the 3950x has like 5% ipc improvement.

it also the same thing with HEDT cpu from intel.. the higher tiered cpu with larger l3 cache has a mega increase on the RAM bandwidth...


This post has been edited by cstkl1: May 18 2020, 07:15 PM