D-PE-OE-01 practice question 8 of 10

On a Dell PowerEdge server equipped with NVIDIA GPUs, one of the installed GPUs shows a continuous Increase in ECC Correctable Errors whenever the…

On a Dell PowerEdge server equipped with NVIDIA GPUs, one of the installed GPUs shows a continuous Increase in ECC Correctable Errors whenever the system is under load. Running dcgmi diag -r 3 returns PASS for all diagnostic tests, and a full GPU memory BIST detects no memory cell faults. The IDRAC System Event Log repeatedly reports: ’ Correctable Memory Warning increased” for the same GPU. The system appears healthy, and no PCIe or XID fatal errors are recorded. What action should be taken first?

Answer options

A: Reduce the GPU memory clock to mitigate temperature-dependent memory errors.

B: Reseat the GPU to eliminate any potential connector or contact issues.

C: Update the PCIe switch firmware to prevent bus-related memory parity errors.

D: Inspect the airflow and thermal conditions for the PCIe slot where the GPU is installed. Correct answer

Correct Answer:

D. Inspect the airflow and thermal conditions for the PCIe slot where the GPU is installed.

Explanation

A continuous increase in correctable ECC (Error Correction Code) errors specifically under load strongly suggests a thermal issue. High operating temperatures can cause memory instability, leading to a higher rate of soft errors that are correctable by ECC. While diagnostics pass at idle or under light load, the thermal stress of a heavy workload reveals the problem. Inspecting airflow, fan operation, and potential obstructions for the specific PCIe slot is the most logical first step to diagnose a heat-related problem before attempting more invasive actions.

Why Incorrect

A. Reducing the clock speed is a mitigation strategy, not a primary diagnostic step. The root cause, likely thermal, should be investigated first.
B. Reseating the GPU is a valid troubleshooting step, but it's less likely to be the cause of errors that only appear consistently under thermal load.
C. A PCIe switch firmware issue would likely cause more severe bus-related errors (like PCIe fatal errors), not just isolated GPU memory ECCs.

References

1. Dell PowerEdge R750xa Installation and Service Manual, "Troubleshooting GPUs" section. While not explicitly stating this scenario, it emphasizes ensuring proper cooling and airflow as a primary step for GPU health. (Reference: Dell PowerEdge R750xa Installation and Service Manual, p. 115, "GPU thermal considerations").

2. NVIDIA Data Center GPU Manager (DCGM) Documentation, "Diagnostics". The documentation notes that environmental factors like temperature can impact GPU health and performance, which can manifest as increased ECC counts. (Reference: NVIDIA DCGM User Guide, Section 3.2 "Health Monitoring").

Keep practicing

Take the free D-PE-OE-01 practice test

Ten exam-style questions with answers and explanations, plus the exam facts and study guides.

Start the free practice test