Mesa RADV vs. AMDVLK Radeon Vulkan Performance For July 2021

[phoronix]

Phoronix: Mesa RADV vs. AMDVLK Radeon Vulkan Performance For July 2021

It's been a while since last looking at the performance of AMD's official "AMDVLK" open-source Linux Vulkan driver against that of the popular Mesa "RADV" Radeon Vulkan driver. But here are some fresh benchmarks for those interested while using the latest-generation Radeon RX 6800 XT graphics card paired with the in-development Linux 5.14 kernel across testing both Vulkan drivers.

https://www.phoronix.com/vr.php?view=30360

[perpetually high]

[aufkrawall]

amdvlk probably regressed (it does so hard and often). What a pathetic display of AMD, I expect them to be stubborn and unteachable and to continue with business as usual...

[torbido]

Wow! You know what is mostly wrong about AMDVLK, after you install it, it is used by default, and you have to use this variable to use RADV:

AMD_VULKAN_ICD=RADV

[user1]

I've already experienced the exact same brokenness of AMDVLK almost a year ago, when I last tried it on my RX 580. Radv was performing ~25% better, or more in almost all games. Seems this brokenness has now expanded to the latest hardware. This just confirms that AMDVLK has turned into nothing more but a Stadia driver (since Google Stadia uses it). I mean looking at its github Issues page, there are so many serious issues still open and the devs barely fix them.

[perpetually high]

amdvlk probably regressed (it does so hard and often). What a pathetic display of AMD, I expect them to be stubborn and unteachable and to continue with business as usual...

They're not completely independent from what I remember. bridgman explain AMD team works closely with open source developers. AMDVLK is not meant for gamers, it's for workstation and enterprise. Am I wrong? If I am, please correct me.

I'm glad RADV and AMDVLK exists because it gives people the choice to use RADV for 90% of games and AMDVLK for the 10% where it's optimized for particular games. I don't see what the big deal is.

[StarterX4]

Even tho it sucks at games, it may be more stable in enterprise segment?

[tomas]

Is it true that AMDVLK is used for Stadia instead of RADV? If so, why???
It does not make any sense. All benchmarks done by Phoronix during the last years have shown RADV to be superior to AMDVLK. By a large margin.

[mphuZ]

Are you sure that you have no problems with your system? This regression looks wild. Or is it only with UltraHD?

I don't consider tests with Proton, since neither the developers nor Valve themselves have lifted a finger to fix errors with AMDVLK.
But the test in the Tomb Rider on low and high settings looks strange.

bridgman, what do you say about this?

[mphuZ]

Is it true that AMDVLK is used for Stadia instead of RADV?

Yes

It does not make any sense. All benchmarks done by Phoronix during the last years have shown RADV to be superior to AMDVLK. By a large margin.

I'll tell you a little secret: there are no problems with AMDVLK when developing native games on Stadia. Just like in Windows for many years.
Therefore, developers do not focus on Phoronix tests, because 80% of the games from the test are launched with Proton.