RADV vs. AMDVLK vs. Radeon Software Vulkan Driver Performance - October 2018 Linux Gaming

[phoronix]

Phoronix: RADV vs. AMDVLK vs. Radeon Software Vulkan Driver Performance - October 2018 Linux Gaming

Given AMD's weekly updating of the public AMDVLK Vulkan driver source tree as their official open-source Vulkan Linux driver while RADV continues to be maintained as the popular open-source Radeon Vulkan driver within the Mesa source tree, here is a fresh look at how those competing drivers perform. Additionally there are the results from Radeon Software / AMDGPU-PRO using its closed-source Vulkan driver that is derived from the same sources as AMDVLK but built against AMD's proprietary shader compiler.

http://www.phoronix.com/vr.php?view=26932

[andre30correia]

Good results with RADV, no need to use the others in most case, I only miss something here, using stable drivers of RADV maybe they are no crashes with stable mesa and llvm

[valici]

That sweet closed source shader compiler.

[hakzsam]

phoronix The RoTR hang on Vega is unexpected. I can't reproduce it locally. Can you give me more info? like mesa/llvm sha1? This is likely a recent regression . Thanks!

[Strunkenbold]

LLVM is definitely missing some serious optimizations for those Vega cards. Wondering if the same optimizations that boost Vulkan performance could give a boost to OpenGL applications too.
But highest priority should be to fix those hangs with Tomb Raider first. There are also some serious issues with GPU hangs under DXVK.

[Michael]

phoronix The RoTR hang on Vega is unexpected. I can't reproduce it locally. Can you give me more info? like mesa/llvm sha1? This is likely a recent regression . Thanks!

I don't have the SHA1 handy as unfortunately the Oibaf PPA build doesn't get it included in the version string and that box is running some other tests at the moment, but have you tried Mesa 18.3 specifically against Linux 4.19+? From some other tests, it seems to happen just on Linux 4.19 (and -next I think still) with RADV.

[shmerl]

radv for Vega has some room for improvement still.

[humbug]

Proprietary driver rocks performance wise. In general Vega has a reputation of not delivering on Linux (compared to it's windows performance). Seems like the proprietary driver goes some way in alleviating that.

Those extremely bad vkmark numbers for RadV on vega: is it some application specific quirk or is it exposing some weakness in RadV?

[M@yeulC]

I wonder if RADV could use the proprietary shader compiler, as this would make some interesting comparisons possible.

[duby229]

I want to comment that Mesa development has taken years to come this far. And as such in comparison to earlier generations, Vega's performance and compatibility is actually really f-ing good. These performance numbers are likely due to a combination of compiler optimizations and binary specific optimizations. In some scenario's Vega on Radv may never achieve the same raw FPS as the proprietary driver. Looking at past history for precedence reminds me of the VLIW-4 architecture cards performance with the R600g driver. Really good drivers for really good cards but they never really did achieve their performance potential. The same thing may well be true for Vega, I don't know.