Mesa GL Thread Testing With Pentium + Core i7 & RX 580 + R9 Fury

[phoronix]

Phoronix: Mesa GL Thread Testing With Pentium + Core i7 & RX 580 + R9 Fury

With Mesa's GL threading support ready for wider testing and the developers pursuing per-application enabling of this driver-agnostic Mesa OpenGL multi-threading work, here are some benchmarks of mesa_glthread when using a Pentium and Core i7 CPUs as well as a Radeon RX 580 and R9 Fury.

http://www.phoronix.com/vr.php?view=24930

[Kendji]

Wonder how Vega will perform compared to the Fury

[mphuZ]

Wonder how Vega will perform compared to the Fury

Learn only at the end of August..

[M@GOid]

While doing my tests I noted that games taxing only one core, are the ones beneficing the most from this. So that is Aspyr ports and old VP ports, the truck simulators, etc. But there are exceptions.

[aufkrawall]

Dying Light and Hitman would be interesting, their OpenGL ports are very CPU limited here with the Nvidia driver.

[Michael]

Dying Light and Hitman would be interesting, their OpenGL ports are very CPU limited here with the Nvidia driver.

Dying Light doesn't have a benchmark mode AFAIK and HITMAN has only been segv'ing for me the past month or two.

[aufkrawall]

Thx for the feedback. I was also hoping that users who read my post will thest those games with GL threading and provide feedback to Mesa, which doesn't necessarily mean that tests must be automated.

[boxie]

While doing my tests I noted that games taxing only one core, are the ones beneficing the most from this. So that is Aspyr ports and old VP ports, the truck simulators, etc. But there are exceptions.

if that is the case, then some sort of instrumentation on mesa (so we know fps etc) and a cpu usage monitor for the owning thread or entire system could allow it to be flipped on automatically

[schmidtbag]

Thanks for these tests. It seems for the most part this proved my theory wrong, that better CPUs may have an impact on the performance. Though there were some tests where the i7 seemed to perform distinctly different, I wouldn't say it was enough of a difference to suggest hardware differences are significant enough.

[bosjc]

Sounds like it should have a runtime flag like NVIDIA that developers (and/or users) can add to their launch scripts if it helps in a certain game.