Christian König <ckoenig.leichtzumerken@xxxxxxxxx> writes: > Well none of the commits mentioned can affect radeon in any way. Radeon > simply doesn't use the scheduler. > > My suspicion is that the user is actually using amdgpu instead of > radeon. The switch potentially occurred accidentally, for example by > compiling amdgpu support for SI/CIK. Indeed, the lspci I originally posted does indicate amdgpu. What is the difference and should I switch it? If so, how? > Those amdgpu problems for older ASIC have already been worked on and > should be fixed by now. I just pulled v6.7-rc2 and it's still broken. I'll see if I can revert those 3 patches.