AMDVLK Still Has Some Performance Advantages Over Mesa's Radeon RADV Vulkan Driver, But It's A Good Fight

[R41N3R]

RADV is one of the best graphic drivers I've ever used, quality always was high. Even if AMDVLK would be faster, I would prefer RADV any day, but very nice that it is mostly on par with AMDVLK and often delivers higher min. frames too :-)

[Danny3]

First image of this phoronix article.

As another thing, when I saw that image, I was wondering why why is it displayed with green color?
In my opinion only "Open source" deserves the green color, "Proprietary" I think it should be red.

[Space Heater]

First image of this phoronix article.

[tildearrow]

Why does it label AMDVLK's license as "proprietary"? I see the source code at https://github.com/GPUOpen-Drivers/AMDVLK and the license.txt says it is under the MIT license.

What am I missing?

Where? I don't see Proprietary on the GitHub page.

[Space Heater]

Why does it label AMDVLK's license as "proprietary"? I see the source code at https://github.com/GPUOpen-Drivers/AMDVLK and the license.txt says it is under the MIT license.

What am I missing?

[dragon321]

ermo RADV was created before AMD released their Vulkan driver as open source and because RADV was quite nice developed and better integrated with Mesa I think this is the reason why it's still developed.

[oibaf]

BTW the article text says "while the RADV Vulkan driver was using Mesa 19.2.0-devel via the Padoka PPA.", however the config shows it's using oibaf PPA:

[airlied]

Hats off to the RADV developers for continuing to refine their driver.

Historically, ISTR that RADV was sort of a stop-gap measure until AMD got their ducks in a row and began releasing the FLOSS AMDVLK driver. These days, what's the benefit of having two competing Vulkan drivers for AMD hardware though?

Couldn't an argument be made that the FLOSS graphics driver programming resources would be better spent collaborating on a single driver such as AMDVLK?

AMDVLK is open released, it isn't open developed. You can't particpate in the project at the near the same impact level as a truly open source development project like Mesa. So if you have resources to spend on a driver for Linux and you aren't AMD, radv is the better place to get a return on that investment.

The scales may change, and the vlk project may become open source developed over time, but currently it's not where development happens.

[ermo]

Hats off to the RADV developers for continuing to refine their driver.

Historically, ISTR that RADV was sort of a stop-gap measure until AMD got their ducks in a row and began releasing the FLOSS AMDVLK driver. These days, what's the benefit of having two competing Vulkan drivers for AMD hardware though?

Couldn't an argument be made that the FLOSS graphics driver programming resources would be better spent collaborating on a single driver such as AMDVLK?

[spykes]

Some benchmarks indeed shows AMDVLK ahead. But if we look closer at the results we notice than RADV has often higher minimal FPS than AMDVLK, which means less frame drop. I will always prefer the driver that gives me the stabler FPS.