Question Gaming on a Threadripper 7995WX

[igor_kavinski]

The best one is the last one.

Higher avg and min fps even though the core frequency is pretty low compared to 7950X3D.

 

[iamgenius]

The threadripper managed to perform better in some games. If you get a threadripper and obtain a good overclock out of it, you will get better single core performance and the gap will be even smaller. It is not like you are giving up on gaming if you get a 7000 series threadripper.

 

[inquiss]

The best one is the last one.

Higher avg and min fps even though the core frequency is pretty low compared to 7950X3D.

You clocks are different between the two so hard to say what's causing the FPS increase

 

[igor_kavinski]

You clocks are different between the two so hard to say what's causing the FPS increase

TR is running at lower clocks. Possibly the game threads are properly being scheduled on a single CCD. If there is some sort of L3 cache "sharing" scheme between the CCDs, the game could get frequent data from the other CCD's caches instead of going out to RAM.

 

[DAPUNISHER]

With the proliferation of fake benchmarks on Youtube, and this channel having no street cred = sus.

 

[igor_kavinski]

With the proliferation of fake benchmarks on Youtube, and this channel having no street cred = sus.

Looks legit to me since most games are getting hurt by running on the 7995WX.

I did gaming tests on my Zen 2 Epyc 128 thread ES.

Performance drops once all threads come into the picture.

But if I keep only two cores per CCD enabled for a total of 16 cores without SMT, there's a good uplift.

FFXV benchmark gained something like 1500 points at 1080p lowest settings.

 

[inquiss]

TR is running at lower clocks. Possibly the game threads are properly being scheduled on a single CCD. If there is some sort of L3 cache "sharing" scheme between the CCDs, the game could get frequent data from the other CCD's caches instead of going out to RAM.

I wasn't very clear, the GPU clocks were different. The desktop processor had fixed, lower clocks whereas the threadripper's GPU had variable faster clocks.

 

[igor_kavinski]

The desktop processor had fixed, lower clocks whereas the threadripper's GPU had variable faster clocks.

Maybe in some other games. I focused mainly on Forza coz that's where I noticed higher fps. I looked at the core clocks on both sides for a while. 7950X3D was hitting almost 4800 Mhz while TR stayed mostly at 3900 MHz with a bit of quick and limited time boosting to 4400 MHz. I think it's the massive combined cache across CCDs that did the trick for Forza.

 

[inquiss]

Maybe in some other games. I focused mainly on Forza coz that's where I noticed higher fps. I looked at the core clocks on both sides for a while. 7950X3D was hitting almost 4800 Mhz while TR stayed mostly at 3900 MHz with a bit of quick and limited time boosting to 4400 MHz. I think it's the massive combined cache across CCDs that did the trick for Forza.

Look at the GPU clocks

 

[igor_kavinski]

Look at the GPU clocks

Ah. That does look anomalous. Why would a 4090 clock higher with TR?

 

[inquiss]

Ah. That does look anomalous. Why would a 4090 clock higher with TR?

It has no reason to. It's not a fair gaming test.