GPU crash after switching lights on/off or plugging in USB

Hey,
For a few years I've had this issue on Linux only. When turning on/off lams in the apartment, or plugging in/out a USB the screens sometimes go completely black, ie the driver crashed. Sounds still come from the speakers. Only fix is to hard reboot. Sometimes its enough to just touch a random cable

Have switched the PSU, have bought 2x power surge protectors at no avail.
I guess it must be an issue with the AMD Vega 56 or its corresponding driver. Have reinstalled and tried several driver versions over the years, but always the same problem.

Google tells me there are others with issues that the screen goes black and driver crash, but none that mention in situations like mine - powering random stuff on/off.

Any help is much appreciated.

Note. Just switched one of my monitors to a much larger one, and the problem is now more frequent than previously

Hello,

After reading your comment, i don't really see the connection to only linux when turning on and off whatever light bulb in the apartment, unless there is a huge issue with the Electrical circuit and the Grounding of it.

Might be the case that one of those gets into a short, influencing everything else.

Is like wearing an armor suit with a king cobra under it. Is quite important where are placed.
Check the power cable that goes to the monitor, or the socket where is connected, change them. Is quite possible if the monitor is turned off, the GPU is not able to establish a "reconnection" with the monitor and acts as there was nothing connected to it.
DPMS options ... ?

If the consumption is higher that the electrical circuit/cable is capable to deliver, then is quite possible to get heated up; that leads to inductance, and can act as short circuit without blowing the fuse, but rather one of your appliances/monitors/pc ...

Thank you for your reply.

I guess the reason for Linux only is due to the display driver not restoring itself when crashed.

I've used several different monitors (always dual screen), with several cables of different lengths.
Also it cant be the fuse/circuit as the same problem lasted over three different apartments (i move around a lot). Have been unable to find anything in the logs unfortunately.

Have you tried buying a new PC? Your issue sounds... absurd, to say the least.

I have not, this PC serves me very well between the crashes, cant motivate a replacement

Just enabled amdgpu.gpu_recovery=1, lets see how that works.

Seems to resolved, switching the lights on/off always triggered the crash recently.

This issue had to do with the amdgpu driver and inability to recover, in combination with specific kernels.
Tried with kernel 5.4, 5.6 and it failed on both (where internet reports it should work with 5.4, but it doesn't). Upgraded to 5.7 and set amdgpu parameters to
options amdgpu amdgpu.gpu_recovery=1 amdgpu.noretry=0 amdgpu.lockup_timeout=1000

and now it works (so far)

references:
https://bbs.archlinux.org/viewtopic.php?id=254283

Edit:
Not resolved....

Based on a second hand laptop experience:
Check the Power lead plugged into the PC. The 3 pin 'Male' plugs (at the PC end) are manufactured with two different lengths of internal connectors. If the power cord is not the original then it is possible it has been substituted with one that requires longer pins in the PC socket, that is the plug section and socket do not make reliable contact when joined.

1 Like

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.

Forum kindly sponsored by