Event

<p>Relying only on nvidia-smi is like measuring highway usage by checking if any car is present, not how many lanes are full. </p> <p>This talk reveals the metrics nvidia-smi doesn't show and introduces open source tools that expose actual GPU efficiency metrics.</p> <p>We'll cover:</p> <ol> <li>Why GPU Utilization is not same as GPU Efficiency.</li> <li>Deep dive into relevant key metrics: SM metrics, Tensor Core metrics, Memory metrics explained.</li> <li>Practical gpu profiling and monitoring setup.</li> <li>Identifying bottlenecks in inference workloads.</li> </ol> <p>Attendees will leave understanding how to identify underutilized GPU and discover real optimization opportunities across inference workloads.</p>