NVIDIA Nsight™ Systems is a system-wide performance analysis tool designed to visualize an application’s algorithms, identify optimization opportunities, and tune to scale efficiently across any quantity or size of CPUs and GPUs. It supports a wide range of NVIDIA platforms, from large servers to small systems-on-a-chip (SoCs), and provides valuable insights for optimizing AI, high-performance computing (HPC), pro-visualization, and gaming applications.

Features

• System-Wide Visualization: Visualizes unbiased, system-wide activity data on a unified timeline, allowing developers to investigate correlations, dependencies, activity, bottlenecks, and resource allocation.
• Low-Overhead Performance Analysis: Offers low-overhead performance analysis that visualizes hidden layers of events and metrics used for optimizations, including CPU parallelization, GPU optimization, system workload, CUDA libraries trace, network communications, and OS interactions.
• Multi-Node Profiling: Supports multi-node profiling to resolve performance limiters on the scale of data centers and clusters, automatically diagnosing performance limiters across many nodes simultaneously.
• GPU Metrics Sampling: Plots low-level input/output (IO) activity such as PCIe throughput, NVIDIA NVLink®, and DRAM activity, exposing SM utilization, Tensor Core activity, instruction throughput, and warp occupancy.
• Integration with JupyterLab: Allows profiling Python and other supported languages directly in JupyterLab, including detailed analysis with the full Nsight Systems GUI.
• Frame Stutter Detection: Automatically detects slow frames and local stutter frames, reporting CPU times per frame and API calls likely causing stutters.

Benefits

• Enhanced Performance: Identifies and optimizes performance bottlenecks, improving overall application efficiency.
• Detailed Insights: Provides comprehensive insights into GPU performance, from memory workload to source code efficiency.
• Flexibility: Supports various development environments and platforms, including desktop, mobile, cloud, and HPC.
• Ease of Use: Offers user-friendly interfaces and integration with popular development environments.
• Advanced Profiling: Enables detailed performance analysis and debugging for CUDA, graphics, and deep learning applications.