NVIDIA Xid 79: GPU Has Fallen Off the Bus
Preserve evidence, isolate the affected GPU and investigate PCIe accessibility without assuming a single root cause.
Problem
A GPU becomes inaccessible and the driver reports Xid 79. Jobs may fail and the device may no longer be usable through the normal management interface.
Conclusion
NVIDIA describes Xid 79 as the driver finding the GPU inaccessible over PCI Express. It is a symptom, not proof of one failed component. Drain affected work and capture diagnostics before recovery removes useful evidence.
Environment
Record GPU identifiers, PCI bus addresses, driver and kernel versions, motherboard or server model, power configuration, recent changes and incident time. No specific hardware incident has been reproduced for this article.
Symptoms
Look for the Xid code and its timestamp in kernel logs. Correlate it with scheduler failures and device visibility. We do not provide fabricated example logs.
Cause
The error identifies lost access to the GPU. Establishing why requires platform evidence; a model parameter change alone is not an adequate root-cause explanation.
Solution
Collect read-only diagnostics while the system remains available.
nvidia-smi -q
journalctl -k -b --no-pager
lspci -nn
Preserve the incident window, including preceding PCIe or platform errors. Stop scheduling new work on the affected node. Follow the server vendor's maintenance procedure for physical inspection or power-related checks. Arrange recovery in a maintenance window and involve support when the problem recurs.
Verification
Benchmark data pending real-world testing.
For this fault, recovery evidence matters more than throughput: confirm device enumeration, management access and a controlled workload after recovery. Monitor for recurrence and retain the original incident bundle.
Caveats
A successful reboot does not establish the root cause. Do not reseat or service energized hardware. Review diagnostic bundles for sensitive paths and identifiers before sharing them.