AMDGPU Reset Recovery To Be Flipped On By Default For Newer Radeon GPUs

[phoronix]

Phoronix: AMDGPU Reset Recovery To Be Flipped On By Default For Newer Radeon GPUs

The AMDGPU DRM Linux kernel driver has offered GPU reset recovery for a while now in case of hangs that can be toggled by a module parameter, but the default behavior for the next kernel release is slated to change where it will be enabled by default for the newer Radeon GPUs...

http://www.phoronix.com/scan.php?page=news_item&px=AMDGPU-GPU-Reset-Recover-Def

[TemplarGR]

Well, till this spring i had gpu hangs that needed the reset button treatment on my Tonga gpu very often, but to be frank the situation has vastly improved with the latest Kernel/Mesa. I get the feeling that things are far more stable now. Still this reset recovery feature seems like a nice thing to have.

[Brisse]

Only hangs I've had recently is with DXVK due to me still being on LLVM6 rather than LLVM7. A few kernel versions back hangs were common but they seem to have gone away.

[Med_]

This is good news. I consistently get hangs with games. Typically once every few hours. I do not bother reporting them as I cannot reproduce on demand and the bug tracker is full of them with similar logs ([drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx timeout, last signaled seq=145101, last emitted seq=145103). I have activated the option, we will see whether that at least prevents the power button treatment.

[equeim]

I sometimes have random hangs with my TAHITI (GCN 1.0) card. Although I suspect that this is hardware issue, since it happens with both radeon and amdgpu drivers. The only difference is that with amdgpu it lets me login via ssh and get dmesg logs, while with radeon it hangs completely.

[Otus]

Is there anything similar for other vendors? My work laptop with Intel+Nvidia hangs almost every week.

[boxie]

This is good news. I consistently get hangs with games. Typically once every few hours. I do not bother reporting them as I cannot reproduce on demand and the bug tracker is full of them with similar logs ([drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx timeout, last signaled seq=145101, last emitted seq=145103). I have activated the option, we will see whether that at least prevents the power button treatment.

I have found the only reason my GPU hangs (vega64) is because it gets too hot. happens under both windows and linux - so, I have got my fan profiles cranked up. I can get many hour gaming sessions in, albeit with more fan noise.

[puleglot]

This is good news. I consistently get hangs with games. Typically once every few hours. I do not bother reporting them as I cannot reproduce on demand and the bug tracker is full of them with similar logs ([drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx timeout, last signaled seq=145101, last emitted seq=145103). I have activated the option, we will see whether that at least prevents the power button treatment.

What GPU?

[Med_]

I have found the only reason my GPU hangs (vega64) is because it gets too hot. happens under both windows and linux - so, I have got my fan profiles cranked up. I can get many hour gaming sessions in, albeit with more fan noise.

It would be worth a try. How do you alter that with AMDGPU?

[Med_]

What GPU?

RX480.