Owning an AMD GPU has been the largest source of stress, trouble and annoyance for me since switching to Linux full time about 5 years ago. To be more specific, the amdgpu kernel driver maintainers have delivered some of the most tone-deaf, dismissive, and patronizing corporate-speak I’ve ever seen in open source. And being open source shouldn't shield them from criticism. This is hardware people paid for, which included driver support on Windows _and_ Linux. The fact that the driver is open source doesn't mean I can waive it off as "oh they're doing it for free or very little monetary benefit and it's a community driven thing". This only passes for drivers that are a community effort, not for drivers that you pay for which happen to be open source. At this point, I'm quite confident my next GPU will be Nvidia and I will be using the nvk+nouveau/nova stack. I'm sure it will be far worse than the AMD experience for a while but at least it won't feel like every maintainer engages in corporate speak for "your GPU is 2 generations old, so your bug report mean jackshit to us" even after bisecting the exact commit that introduced the bug. The bug in question here is https://gitlab.freedesktop.org/drm/amd/-/issues/4559 which only affects RDNA1 and RDNA2, caused by a workaround for another bug that was introduced while enabling a feature that is only applicable on RDNA3. Not only is this a clear kernel regression policy violation, the driver has been left in this state for 2 whole release cycles now while a band-aid fix is in place for Steamdeck but not for any other ASIC, I guess because regular customers are worthless while Valve is not.