X server random crash / frozen - 2080 (Ubuntu 16.04.5 - Driver 410.48)

Ever since installing the Nvidia 2080 i have been having this problem where X server will randomly freeze or crash and need to be manually killed / restarted in a different TTY. I have tried updating to Driver 415 which made no difference so reverted back to 410 and still having the same problem.

Seems to happen most often when going from normal view (window) to full screen on YouTube or VLC or any graphical website / app. but also appears to happen randomly when the GPU is not really being used as well like in notepad.

Every crash i always get the same Xid error in /var/log/syslog:

Dec  1 13:21:24 ubuntu-desktop kernel: [12777.323694] NVRM: GPU at PCI:0000:65:00: GPU-504ca980-96c3-b5b4-32e7-ebee1f7b139d
Dec  1 13:21:24 ubuntu-desktop kernel: [12777.323700] NVRM: GPU Board Serial Number: 0323618056898
Dec  1 13:21:24 ubuntu-desktop kernel: [12777.323705] NVRM: Xid (PCI:0000:65:00): 31, Ch 00000009, engmask 00000101, intr 00000000
Dec  1 13:21:24 ubuntu-desktop kernel: [12777.360748] NVRM: Xid (PCI:0000:65:00): 69, Class Error: ChId 0009, Class 0000902d, Offset 00000250, Data ffffffff, ErrorCode 0000000c
Dec  1 13:21:24 ubuntu-desktop kernel: [12777.379011] NVRM: Xid (PCI:0000:65:00): 69, Class Error: ChId 0009, Class 0000902d, Offset 00000220, Data ffffffff, ErrorCode 0000000c

Thanks
nvidia-bug-report.log.gz (1.03 MB)

Start by using cuda memtest to check for vmem errors:
https://sourceforge.net/projects/cudagpumemtest/
Installing cuda:

  • download the cuda .deb
  • add it to your system
  • don’t install cuda
  • instead, run sudo apt install cuda-toolkit-10-0