Y'all like Linux? Of course you do, that's why you're here! Wendell gives you the scoop on the more powerful half of AMD's Ryzen 9000 offerings, and makes a ...
From my experince AMD drivers are pretty close, I’d even say slightly better on GNU/Linux, definitely more stable and consistent. For Nvidia, yeah they are bad at supporting GNU/Linux. Improved a lot through the years but still not there. For Intel, well not exactly an option for gaming, at least not the integrated GPUs I have used so far, but still better than in Windows in a similar way as in AMD case.
P.S.
Another great thing with libre/opensource GNU/Linux drivers: When you report a bug with Mesa3D drivers the bug is quite quickly fixed, especially when you can provide them with backtrace and/or Vulkan/OpenGL API trace. Doing a bisect of source code commits amd identifying the commit that introduced a regression also help a great deal. Good luck doing the same with closed/Windows drivers: you can wait for years and no fix.
Two problems:
Benchmarks are not real world performance.
Video card drivers are still not up to snuff as they are with windows, to the best of my knowledge. (don’t kill me if i’m wrong!)
It depends: if you’re talking rasterization performance, then yeah, they’re pretty much on parity.
If you’re talking about all the other things a video card does that’s not strictly make-pixel-light-up, then it’s very much a mixed bag.
From my experince AMD drivers are pretty close, I’d even say slightly better on GNU/Linux, definitely more stable and consistent. For Nvidia, yeah they are bad at supporting GNU/Linux. Improved a lot through the years but still not there. For Intel, well not exactly an option for gaming, at least not the integrated GPUs I have used so far, but still better than in Windows in a similar way as in AMD case.
P.S. Another great thing with libre/opensource GNU/Linux drivers: When you report a bug with Mesa3D drivers the bug is quite quickly fixed, especially when you can provide them with backtrace and/or Vulkan/OpenGL API trace. Doing a bisect of source code commits amd identifying the commit that introduced a regression also help a great deal. Good luck doing the same with closed/Windows drivers: you can wait for years and no fix.