[610.57.04 / GB203M] RTX 5080 Max-Q Laptop (ASUS ROG Zephyrus G14 GA403WW) randomly drops off the PCIe bus in Hybrid mode: pciehp "Link Down / Card not present", Xid 79 only logged later

Open
  • Is something wrong?
#1,415 2 comments 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

Nobody has claimed this yet.

Assessment

Difficulty
5/5
Estimated time
Over a week
Newbie friendliness
5/100
Issue type
Bug
Clarity
Needs clarification
Activity status
Active
Tech stack
c, linux
Domain
operating-systems

Research direction

The reporter links #1151, #1173 and #1283 as similar Xid 79 / PCIe link-down reports; read those threads first, then the pciehp "Link Down / Card not present" sequence and the attached redacted nvidia-bug-report gist. The drop is intermittent and not reproducible on demand, so "done" means explaining why root port 0000:00:03.1 loses the endpoint, likely by correlating logs across affected boards and trying options such as pcie_aspm=off. No source file, test or candidate fix is named anywhere in the report.

Written by the indexing model from the issue text.

Description

NVIDIA Open GPU Kernel Modules Version

610.57.04 (nvidia-open-dkms 610.57.04-1, Arch extra)

Please confirm this issue does not happen with the proprietary driver (of the same version).
  • Not tested. RTX 50-series (Blackwell) only supports the open kernel modules, so there is no proprietary build to compare against.
Operating System and Version

Omarchy (Arch Linux based), kernel 7.2.5-3-omarchy, Hyprland (Wayland), GCC 16.2.1

Hardware: GPU

NVIDIA GeForce RTX 5080 Max-Q / Mobile (GB203M / GN22-X9, 10de:2c59, subsystem ASUS 1043:39c8) at PCI 0000:64:00.0, behind root port 0000:00:03.1.
Laptop: ASUS ROG Zephyrus G14 GA403WW, BIOS GA403WW.310 (2025-11-21; fwupd shows no newer firmware on LVFS).
Hybrid graphics (supergfxd mode Hybrid); AMD iGPU (amdgpu) drives the internal display. Plugged into AC during the 2026-10-04 event (not known for 2026-09-24).

Describe the bug

The dGPU randomly disappears from the PCIe bus during normal desktop use. After that the NVIDIA driver is stuck and every NVIDIA-using app hangs (e.g. Steam's window never appears). Only a full power-off recovers it (a Restart hung on the boot logo the previous time).

Timeline of the most recent event (2026-10-04, boot at 08:00:44, plugged in, no suspend/resume in this boot):

10:11:17 pcieport 0000:00:03.1: pciehp: Slot(0-2): Link Down
10:11:17 pcieport 0000:00:03.1: pciehp: Slot(0-2): Card not present
10:11:17 [drm] [nvidia-drm] [GPU ID 0x00006400] Removing device
10:11:17 [drm] [nvidia-drm] [GPU ID 0x00006400] Unloading driver
10:11:22 nvidia-modeset: ERROR: GPU:0: Error while waiting for GPU progress: 0x0000ca7d:0 2:0:4048:4040
         (repeats every 5 s, 495+ times, until reboot)
10:43:08 NVRM: Xid (PCI:0000:64:00): 79, pid=15518, name=lutris, GPU has fallen off the bus.
10:43:08 NVRM: Xid (PCI:0000:64:00): 154, GPU recovery action changed from 0x0 (None) to 0x1 (PF FLR)
10:43:08 NVRM: GPU0 _threadNodeCheckTimeout: API_GPU_ATTACHED_SANITY_CHECK failed!
10:43:08 NVRM: GPU0 nvCheckOkFailedNoLog: Check failed: GPU lost from the bus [NV_ERR_GPU_IS_LOST] (0x0000000F)

Notes:

  • The Xid 79 is only logged ~32 minutes AFTER the link went down, when a process (lutris) next touched the GPU. The kernel's pciehp message is the real first sign.
  • No AER/DPC messages anywhere: firmware reports _OSC: platform does not support [SHPCHotplug AER LTR DPC].
  • After the drop: lspci -vv -s 64:00.0 shows "!!! Unknown header type 7f"; root port 00:03.1 shows DLActive-, LnkSta: Speed 2.5GT/s (endpoint max 32GT/s); sysfs power/runtime_status still says active (D0); nvidia-smi fails with "Unable to determine the device handle for GPU0: Unknown Error". nvidia-bug-report.sh itself hangs (it calls nvidia-smi), so the attached report was cut off at that step.
  • power/control was on (runtime PM disabled by a udev rule while on AC), so this is not a runtime-D3 transition.
  • Possibly unrelated boot messages that may be worth a look: PlatformRequestHandler failed to get target temp / platform power mode from SBIOS (NV_ERR_INVALID_DATA) and subdeviceCtrlCmdKGrGetNumTpcsForGpc: Incorrect GPC-Idx provided = 6.

Earlier occurrence on the same machine, 2026-09-24: pciehp: Link Down / Card not present at 21:02:43, 46 minutes after boot, no sleep, no display plug/unplug (AC state unknown), followed by the same repeating "Error while waiting for GPU progress". On 2026-09-23 right after a resume from suspend, Xid 31 (GMMU page fault) and Xid 154 ("OS Reboot" needed) were logged. On 2026-10-02 the journal also shows an unrelated-looking Xid 13 (Graphics Exception) from steamwebhelper. Suspend/resume services (nvidia-suspend/resume) are enabled and a deliberate suspend test passed on 2026-09-20.

To Reproduce

Not reproducible on demand. It has happened twice in 10 days (2026-09-24 and 2026-10-04), 46 min and about 2h 10m after boot, with no known trigger.

Bug Incidence

Intermittent. 2 confirmed full drop-offs in about 10 days of normal use.

nvidia-bug-report.log.gz

Redacted report (hostname, username, IPs, MACs, GPU UUID, filesystem UUID and PCIe serial removed): https://gist.github.com/adamlrutledge/80a3b645c9d217f48cdfb2c0c25b0970

It was captured while the GPU was still dead, but the script was interrupted when nvidia-smi hung, so it is incomplete (the kernel/journal logs and system info are there).

More Info

Things already tried that did NOT prevent it: keeping the GPU out of runtime suspend while on AC (power/control=on), NVreg_PreserveVideoMemoryAllocations=0 with nvidia-suspend/resume services enabled. No kernel boot parameters are set (ASPM left at default; the root port reports ASPM not supported).

Looks related to (but not identical to) these open reports:

  • #1151: RTX 5080 (desktop, GB203), random Xid 79 with no precursor, 595.71.05
  • #1173: RTX 5080 Laptop, Xid 79 after HDMI hotplug, 610.43.02
  • #1283: ASUS Zephyrus G16, dGPU silently removed from the PCI bus in D0, 610.57.04 (there the driver is never notified; here nvidia-drm does log "Removing device" at the time of the link-down)

Happy to run extra debugging or test kernel/driver options (e.g. pcie_aspm=off) if you tell me what would help.

Dominant language
C
Stars
17.5k
Forks
1.9k
No merged PRs in 30d

Getting set up

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from NVIDIA/open-gpu-kernel-modules

All issues in NVIDIA/open-gpu-kernel-modules

Similar issues

More C issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.