"GPU has fallen off the bus", Xid 79, RTX 3070

Graphics / Linux Linux
,
1

The GPU falls of the bus at a random time in a demanding application (Cyberpunk 2077). The only way to get rid of the black screen is to coldboot, and that’s not even guaranteed - sometimes multiple coldboots are needed to get the GPU up and running.

Hardware:

  • CPU: AMD Ryzen 7 5800X 8-Core
  • GPU: GeForce RTX 3070
  • Motherboard: Asus Prime b450m-k ii
  • OS Type: Pop!_os

Tried & failed methods:

  • Different nvidia-drivers (nvidia-driver-560 & nvidia-driver-550-server via apt)
  • Undervolted the GPU (180 W power limit & 1000-1800 MHz clock lock)
  • Checked hardware for problems (No loose connection or cable)

Here is the relevant snippet of journalctl:

Oct 08 00:27:10 kotias kernel: snd_hda_intel 0000:09:00.1: AMD-Vi: Event logged [IO_PAGE_FAULT domain=0x0011 address=0xda82de64 flags=0x0020]
Oct 08 00:27:10 kotias kernel: snd_hda_intel 0000:09:00.1: AMD-Vi: Event logged [IO_PAGE_FAULT domain=0x0011 address=0xda82de00 flags=0x0020]
Oct 08 00:27:10 kotias kernel: NVRM: GPU at PCI:0000:09:00: GPU-efe74fdc-6c45-4e43-4220-47aee04ee805
Oct 08 00:27:10 kotias kernel: NVRM: Xid (PCI:0000:09:00): 79, pid=‘’, name=, GPU has fallen off the bus.
Oct 08 00:27:10 kotias kernel: NVRM: GPU 0000:09:00.0: GPU has fallen off the bus.
Oct 08 00:27:10 kotias kernel: snd_hda_intel 0000:09:00.1: AMD-Vi: Event logged [IO_PAGE_FAULT domain=0x0011 address=0xda82de00 flags=0x0020]
Oct 08 00:27:10 kotias kernel: snd_hda_intel 0000:09:00.1: AMD-Vi: Event logged [IO_PAGE_FAULT domain=0x0011 address=0xda82de00 flags=0x0020]
Oct 08 00:27:10 kotias kernel: snd_hda_intel 0000:09:00.1: AMD-Vi: Event logged [IO_PAGE_FAULT domain=0x0011 address=0xda82de00 flags=0x0020]
Oct 08 00:27:10 kotias kernel: NVRM: A GPU crash dump has been created. If possible, please run NVRM: nvidia-bug-report.sh as root to collect this data before NVRM: the NVIDIA kernel module is unloaded.
Oct 08 00:27:10 kotias kernel: snd_hda_intel 0000:09:00.1: AMD-Vi: Event logged [IO_PAGE_FAULT domain=0x0011 address=0xda82de00 flags=0x0020]
Oct 08 00:27:10 kotias /usr/libexec/gdm-x-session[10656]: (standard_in) 1: syntax error
Oct 08 00:27:10 kotias kernel: nvidia-modeset: ERROR: GPU:0: Failed to query display engine channel state: 0x0000c67e:6:0:0x0000000f

6

If you have already tried everything, such as updating or downgrading the GPU driver, updating the BIOS and chipset, and adjusting BIOS settings without success, here’s another troubleshooting step you can try:

Test the PCIe Power Cable:

  • While your computer is powered on and idle, gently touch the PCIe power cable connected to your GPU.
  • If nothing happens, try this again while the GPU is under stress (e.g., during a stress test or while gaming).
  • If your GPU loses connection (e.g., the display goes blank or the system crashes), it’s likely a power delivery issue.

Fix the PCIe Power Cable Pins:

  • Power off your computer.
  • Unplug the PCIe power cable from your GPU.
  • Using a needle, carefully narrow the female pins inside the PCIe power connector to ensure a tighter fit.
  • Additionally, check for any pins that appear darker than the others. These are usually the ones that need attention and may require narrowing.
7

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.