Learn how to use the Nsight Perf SDK to assess performance bottlenecks in your graphics applications running on NVIDIA GPUs. We'll walk through sample scenarios and explain how a few simple API calls can be added to applications or game engines with minimal overhead to running applications. These scenarios include application-level profiling, regression analysis, and GPU unit bottleneck investigation. Using generated HTML reports of sample applications that show per-GPU unit metrics, we'll show how easy it is to use these reports with minimal effort to gain insight into performance. We'll show usage patterns here as an example of ease of developer integration. You'll gain the background you need to improve your own NVIDIA GPU applications on all desktop GPUs that NVIDIA has supported, from our Volta architecture up to the latest Ampere family of GPUs.