NVIDIA vs. AMD Linux Gaming Performance For End Of May 2021 Drivers

[Linuxxx]

I'm fairly certain the 5900X has cores split 6 and 6 between the CCX's, so that both are equal to each other.

Agree with the rest of what you said though. Most of the latency issues Zen 2 had are solved or at least mitigated with Zen 3. That's one of the ways they improved gaming performance by so much this generation. And if RPCS3 was ok with Zen 2 8 core CPUs they are certainly going to be fine with Zen3 up to 16 cores due to the doubled size of each CCX.

Still doesn't answer the question why the 6800 performed the way it did.

If the "amdgpu" kernel module keeps bouncing around between the two CCXs of the 5900X because of the unnecessary interrupt distribution introduced by the "irqbalance" daemon, do you really think that this won't have any impact on the frame-time & FPS numbers?

[smitty3268]

Still doesn't answer the question why the 6800 performed the way it did.

If the "amdgpu" kernel module keeps bouncing around between the two CCXs of the 5900X because of the unnecessary interrupt distribution introduced by the "irqbalance" daemon, do you really think that this won't have any impact on the frame-time & FPS numbers?

It wouldn't surprise me if there is a bug with the 6800 driver code. It certainly wouldn't be the first time. That said, sure, I don't know why it's slow. Maybe it's the irqbalance thing you mentioned.

I was only responding that in general, Zen 2 had known performance issues with games when going past 8 cores, and that has largely been mitigated with Zen 3 designs.

[cynical]

I'm fairly certain the 5900X has cores split 6 and 6 between the CCX's, so that both are equal to each other.

Oh yes you are correct about this.

I only brought up the RPCS3 example because you claimed that inter-CCX latency can't be that bad, while you now stand corrected by acknowledging yourself that the 5900X can have a worsening effect on gaming results.

Well it’s not even the latency that is really the problem in gaming, although it might add something. It’s that the 5800X can boost higher than the 5900X on all cores because fewer cores means less heat when you do it, so you can sustain higher boost clocks for longer.

I’m not sure why the 6800 result is off, but could it be a regression in the driver code?