Skip to main content
Thoroughly test your machines before listing them on Runpod. A burn test runs your GPUs under sustained load to verify stability and catch potential hardware issues.

​
Run a GPU burn test

1

Stop the Runpod agent

Stop the Runpod agent to prevent the machine from accepting jobs during testing:
sudo systemctl stop runpod
2

Run the burn test

Run a GPU burn test for at least 48 hours (172800 seconds) to verify stability under load:
docker run --gpus all --rm jorghi21/gpu-burn-test 172800
Monitor GPU temperatures and watch for any errors during the test.
3

Verify CPU, memory, and storage

Use stress-ng to test other system components:
# Install stress-ng
sudo apt install stress-ng

# Run a 1-hour stress test on CPU, memory, and disk
stress-ng --cpu 0 --vm 2 --hdd 2 --timeout 3600
4

Restart the Runpod agent

After all tests pass, start the Runpod agent again:
sudo systemctl start runpod
5

Self-rent your machine

Go to your machine dashboard and self-rent the machine. Test it with popular templates to verify everything works correctly with real workloads.

​
What to watch for

  • GPU temperature: Sustained temperatures above 85°C may indicate cooling issues.
  • Memory errors: Any GPU memory errors during the burn test require investigation.
  • System stability: Crashes, freezes, or unexpected reboots indicate hardware problems.
  • Performance degradation: Significant performance drops over time may indicate thermal throttling.