This is a modal window.
Beginning of dialog window. Escape will cancel and close the window.
End of dialog window.
Take a deep dive into the NVIDIA Nsight Compute SOL (Speed Of Light) section, one of the first and most important collections of data within Nsight Compute. SOL analysis reveals how your code performs, and device utilization compared to relevant maximums. Learn how SOL analysis helps you find and resolve limitations to the performance of your code.
Key takeaways include:
◻️ Capture SOL data using the Nsight Compute command line or GUI tool. SOL is the starting point for Nsight Compute reporting, offering insights into device utilization and performance bottlenecks.
◻️ Navigate through various metrics, including Compute (SM) throughput and memory utilization.
◻️ Explore kernel duration, SM activity, memory traffic, and more, to help find the best opportunities for code optimization.
◻️ Analyze the transition from a 64-bit to a 32-bit floating-point variant in the sample code. Trace a significant reduction in kernel duration and the shift toward more "balanced" code. Understand how the change in computation affects hardware utilization and overall GPU performance.