Free NVIDIA NCA-AIIO Actual Exam Questions - Question 6 Discussion
performance compared to running workloads on bare metal. Which factor is most likely contributing
to the performance degradation?
Probably B still makes the most sense here. Overcommitting GPU resources means the GPU scheduler has to juggle multiple workloads that all want heavy GPU time, which naturally slows down each one. The other options like networking or SSDs wouldn’t directly impact GPU compute speed. Even HA features mainly add resilience, not performance hits that big. Without proper passthrough or dedicated resources, the GPU just can’t keep up with multiple demanding workloads at once.
B, because sharing GPUs among too many VMs limits resources per VM.
It’s B because overcommitting GPU resources means the workload can’t get enough GPU power consistently, causing big slowdowns. The other options don’t directly throttle GPU compute like overcommitment does.
D imo, enabling high availability can add overhead and impact GPU performance in virtual setups, more so than options like networking or storage.
B imo, GPU sharing often leads to serious slowdowns.
B. Overcommitting GPU resources usually kills performance here.