1

I am developping an OpenGL graphics program in Arch Linux. Everything was working fine, except I recently attempted to render the Sponza scene.

The program launches and runs fine, however my GPU starts making a screeching noise (I imagine it's the fans turning too quickly) and after about a minute the entire system crashes and reboots.

Checking VRAM shows that there is still at leas 4GB of VRAM available when the program crashes, so I don't think this is a running out of memory problem.

The other potential problem I can think off is my power supply running out of power, however I have a 750B evga power supply and my GPU is only consuming 119W out of 200W at the time of the crash (inspected with nvidia-smi).

Given that the system reboots when the error happens I am not sure which log to inspect to determine the caus eof the error.

  • First of all, deal with overheating of your GPU. – ajeh Jun 18 '18 at 17:09
  • The last reported temperature of the GPU is 50 C, which is within normal parameters accoridng to the spec – Makogan Jun 18 '18 at 18:45
  • You wrote that fans are turning too fast and making noise. That's overheating. Whatever is reporting GPU temperature is out to lunch, as most sensor programs under *NIX are. Unless of course you've done lots of research and adjusted constants using an external thermometer. – ajeh Jun 18 '18 at 20:10
  • The other reason why it's unlikely to be overheating, is because closing the program just before the expected crash time (about a minute) prevents rebooting – Makogan Jun 18 '18 at 20:38
  • Then how do you explain high fan RPMs? The built in sensor in the GPU cannot be tricked - it's overheating. If you close the program, you avoid overheating the GPU. I develop in OpenCL and have learned from experience that Linux sensor monitoring lies 99% of the time. – ajeh Jun 18 '18 at 21:01
1

Since you're on Arch Linux, you can try journalctl -k -b -1 this should show all kernel logs from the last boot.

Also, I think this might be a problem with your PSU. You mentioned you have 4GB VRAM, judging from that, you have a powerful system, and 200W is nowhere near enough for that.

  • My PSU is a 750B evga, the 200W is the maximum load of the GPU – Makogan Jun 18 '18 at 16:32
  • Oh, my mistake here. In this case, check journalctl, since this is likely a software problem. – Marius Jun 18 '18 at 16:38
  • Power consumption by GPUs is highly overrated. Reality is different. I've tested many systems using a digital power meter, and all w/o exception used a fraction of "required" Watts. – ajeh Jun 18 '18 at 17:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.