AMDVLK vs. RADV Radeon Vulkan Driver Performance For Linux Gaming On Ubuntu 19.04

[Michael]

Hey Michael, I don't know if you or anyone cares about more steam play benchmarks, but if so then HITMAN 2 might fit the bill. Seems to work good (spent 60 hours so far) and has a built in benchmark mode.

Last I checked I had it working with Steam Play but the benchmark mode had some issue on exiting. Was a month or two back so will check when time allows to see if can get it working correctly.

[ryao]

It's the same codebase, but not the same codegen backend.

As far as I can tell, the purpose of amdvlk was to be able to use the same SPIR-V compiler that they use on Windows for unified development. Why would they go through the trouble of being able to use the same compiler only to make a different one?

[aufkrawall]

As far as I can tell, the purpose of amdvlk was to be able to use the same SPIR-V compiler that they use on Windows for unified development. Why would they go through the trouble of being able to use the same compiler only to make a different one?

Afair they want to replace the closed compiler at some point, which would turn the Windows driver into the open amdvlk.
Just doesn't look like this would happen any time soon, given the questionable state and development.

[valici]

Congrats RADV developers, the few you are.
Although i would want Amd and Radv devel forces combined.

[smitty3268]

As far as I can tell, the purpose of amdvlk was to be able to use the same SPIR-V compiler that they use on Windows for unified development. Why would they go through the trouble of being able to use the same compiler only to make a different one?

The front-end side is the same, but they don't intend to ever open source the backend. That's why amdvlk switched over to using LLVM, based on a fork of the llvm support present for radeonsi.

I believe they wanted to replace the backend with the new LLVM based one everywhere eventually, but that's obviously a long way from happening at this point.

[Teggs]

Interesting that some people saw Google write 'working with AMD' and 'open-source driver', and assumed RADV. Others (like me) assumed AMDVLK. I wonder if anyone is allowed to tell us which is correct?

Disappointing results in the benchmark here, though. Not having developer/publisher input hurts AMDVLK greatly. It will never be considered the primary driver for gaming if so many games won't run on it from day 1, no matter that it shines sometimes.

[fuzz]

Interesting that some people saw Google write 'working with AMD' and 'open-source driver', and assumed RADV. Others (like me) assumed AMDVLK. I wonder if anyone is allowed to tell us which is correct?

Both have points, with a Google engineer working on RADV while AMD probably wants to support AMDVLK.

[boxie]

Both have points, with a Google engineer working on RADV while AMD probably wants to support AMDVLK.

Why not both?

[fuzz]

Why not both?

We can only speculate ¯\_(ツ)_/¯

[LeJimster]

Why not both?

They definitely could do a per-game thing. But let's be honest, which driver is more stable, open and well supported with dxvk? If you were going to tell developers to target Linux you would choose RADV every time.

If google have any driver issues, they will be in a better place to fix them with RADV then AMDVLK.