NVIDIA AI Infrastructure and Operations Operations and Monitoring NCA-AIIO
Quiz 7 Question 10 of 20

A system administrator scales an image classification model training job from two GPUs to eight GPUs. However, the total training time remains nearly identical, and monitoring logs reveal that while the CPU cores are pinned near 100% utilization, the GPU utilization has dropped significantly. What is the most likely root cause of this scaling failure?

Select an answer to reveal the explanation.