NCP-AII Question 14
Select 4You are managing an NVIDIA DGX system running a deep learning workload. The system unexpectedly crashes, and upon inspection, you find that one of the GPUs is not being recognized by the system. What steps should you take to identify the faulty component?
- A
Check the system logs and look for GPU-related error messages, such as ECC memory errors or PCIe issues.
- B
Run the 'nvidia-smi' command to verify the status of all GPUs in the system.
- C
Immediately replace the suspected GPU without further diagnostics to minimize downtime.
- D
Inspect the power supply units (PSUs) to ensure they are providing adequate power to the system.
- E
Verify if the GPU driver is properly installed and up to date.
Show answer and explanation
Correct answers: A, B, D, E
Explanation
To identify the faulty component in a DGX system, it is essential to take a systematic approach. Checking logs, using diagnostic tools like 'nvidia-smi', inspecting power supplies, and verifying GPU drivers are all critical steps in diagnosing GPU-related issues. Replacing hardware without proper diagnostics could lead to unnecessary downtime and costs.
- A. Correct.
System logs can provide critical information about GPU errors, such as ECC memory faults or PCIe connectivity issues, which help in identifying faulty GPUs.
- B. Correct.
The 'nvidia-smi' command is a crucial diagnostic tool to check the status and health of GPUs, including utilization, memory errors, and GPU presence.
- C. Incorrect.
Replacing the GPU without further diagnostics can lead to unnecessary hardware replacements or missing the root cause of the issue. Diagnostics should always be performed first.
- D. Correct.
Faulty or insufficient power supply can cause GPUs to malfunction or not be recognized. Inspecting the PSUs ensures that power delivery to the GPUs is stable.
- E. Correct.
Outdated or improperly installed drivers can cause GPUs to be unrecognized or function incorrectly. Verifying the driver status is a necessary diagnostic step.