Trace an Ollama slowdown through the GPU stack
Investigate driver visibility, service logs and GPU placement in order when a previously fast local model becomes slow.
In brief
- Ollama’s troubleshooting guide starts with the server logs.
- Write down the last working versions and any recent driver, kernel or container changes.
What to check
Ollama’s troubleshooting guide starts with the server logs. On Linux, use the systemd journal for a native service or container logs for Docker. Check whether the host detects the accelerator and, for containers, whether the workload can see it too. Then repeat a small known-good request and inspect which processor is serving the model.
Put it to work
Write down the last working versions and any recent driver, kernel or container changes. Compare a short-context request with the workload that became slow. If only the larger request degrades, inspect memory pressure as well as device discovery. Change one setting at a time and retest, so the fix can be reproduced on the next restart or upgrade.
Isolate the change with a small known-good prompt
Start the same model and prompt that previously worked. Record whether the host can see the card and whether Ollama reports GPU memory for the active model. Compare the run with a smaller context allocation. If the small case uses the GPU and the larger case spills into system memory, investigate capacity first. If both cases lose the device, inspect driver discovery and service logs.
Keep the repair test narrow
Record the exact error, driver and service versions, and the time of the failure. Change one layer at a time, then repeat the same prompt and record its processor placement and timing. Check behavior after a service restart and a reboot before calling the incident resolved. Keep the known-good combination handy so a future driver upgrade can be assessed against a stable baseline.
Sources and verification
Based on primary documentation reviewed on 3 November 2026. Practical test procedures are editorial guidance. Performance depends on the model, hardware and configuration.
- Troubleshooting Ollama · Primary document
- Hardware support Ollama · Primary document
- List running models Ollama · Primary document
Prepared with AI assistance from the linked sources. This is not a claim of independent field reporting or human review. How we publish · Report a correction
Keep up with homelab
Explore the rest of the local Ollama guide series.
RSS requires an email newsletter opt-in.
