AMD's OpenGL Linux Driver Already Scored A Nice Performance Win For 2024

[phoronix]

Phoronix: AMD's OpenGL Linux Driver Already Scored A Nice Performance Win For 2024

While much of the focus by graphics vendors these days is on their Vulkan driver support/performance and less so about OpenGL in 2024, AMD's open-source RadeonSI Gallium3D driver for Linux systems is still showing no signs of slowing down and still scoring more performance victories...

https://www.phoronix.com/news/RadeonSI-First-Win-2024

[user1]

Shouldn't RADV also benefit from the winsys improvements since it supposedly reuses RadeonSi's winsys? Or was it hard forked from RadeonSi early when RADV development has started?

Btw, typo:

and "massively creases" the CPU cache footprint

[risho]

with vulkan and zink what is the point in still putting time and energy into native opengl?

[Eirikr1848]

OpenGL?! Minecraft improvements?! Minecraft on Steam Deck improvements down the line?

with vulkan and zink what is the point in still putting time and energy into native opengl?

Counterpoint: with Proton: what's the point in using Vulkan or Zink or OpenGL?

Ultimately you can wrap a wrapper with a wrapper but native implementations will likely continue to command a performance lead for some time.

[plandream]

Excellent news. Marek does it again.

[risho]

Counterpoint: with Proton: what's the point in using Vulkan or Zink or OpenGL?

Ultimately you can wrap a wrapper with a wrapper but native implementations will likely continue to command a performance lead for some time.

OpenGL is mostly stuff that is old and doesn't need optimal performance. Either way seeing as how dxvk and vkd3d are running directx on vulkan and running through windows apis clearly wrappers upon wrappers is not a problem. If they would just switch to zink then they could just do what they are already doing: focus on making vulkan as good as possible and then picking up everything else for free. it would probably save a lot of wasted developer time.

[plandream]

We are not there yet. We still need performant and conforming opengl drivers.

[user1]

OpenGL is mostly stuff that is old and doesn't need optimal performance. Either way seeing as how dxvk and vkd3d are running directx on vulkan and running through windows apis clearly wrappers upon wrappers is not a problem. If they would just switch to zink then they could just do what they are already doing: focus on making vulkan as good as possible and then picking up everything else for free. it would probably save a lot of wasted developer time.

There is still some OpenGL stuff that while not as demanding as today's AAA games, still on the more demanding side. There's plenty of low end hardware that benefits from the optimizations of a native GL driver.

Also, in case of AMD, RadeonSi is developed by AMD, while RADV is not, so RADV devs sometimes need to rely on the work done in RadeonSi. For example, for stuff like early new hardware support.

[You-]

with vulkan and zink what is the point in still putting time and energy into native opengl?

For new drivers it may not be worth it. But with existing drivers, there is a different question of benefits because the work on the OpenGL driver has already been done and is just being tweaked.

[Ember2528]

with vulkan and zink what is the point in still putting time and energy into native opengl?

Really, there is still plenty of demanding OpenGL software. First one off the top of my head is OpenMW. At default settings and without mods you are mostly fine performance wise on mid-high end hardware. But, say, you want to crank the view distance up and go to one or the heavier areas of Tamriel Rebuilt like the Mephalan Vales and you'll struggle to get 15FPS even on a 7900XTX