The Nsight suite of Developer Tools provide insightful tracing, debugging, profiling, and other analyses to optimize your complex computational applications across NVIDIA GPUs, and CPUs including x86, Arm, and Power architectures.
Unlocking the Power of GPU Profiler and Debugger: Nsight Systems 2021.5 and Nsight Compute 2021.3
NVIDIA Nsight Systems is a performance analysis tool designed to visualize, analyze and optimize programming models, and tune to scale efficiently across any quantity, or size, of CPUs and GPUs; from workstations to supercomputers.
Nsight Systems 2021.5 highlights include:
- Statistics now available in graphical user interface (GUI).
- Multireport view with horizontal and vertical layouts to aid investigations across server nodes, VMs, containers, ranks, and processes (coming soon).
- Expert system now includes GPU utilization analysis for OpenGL and DX12.
- NVIDIA NIC InfiniBand metrics sampling (experimental).
- DirectX12 memory operations and warnings.
- DXGI/DX12/Vulkan API calls correlation to WDDM queue packets.
- Windows 11 support.
NVIDIA Nsight Compute 2021.3 released new features for measuring and modeling occupancy, source and assembly code correlation, and a hierarchical roofline model to identify bottlenecks caused by accessing cache memory.
- Occupancy Calculator – Helps you understand the hardware resource utilization of your kernels, and model how adjustments could impact occupancy.
- Command line source page – enables accessing the information from the Source page in the GUI directly from the command line. By using the
--page sourceflag, you can see the lines of source, PTX, or assembly and the collected metrics for those lines output on the command line. This feature gives additional flexibility when it comes to analyzing the collected data as well as scripting and post-processing results for further reporting and analysis.
- Hierarchical Roofline – The Roofline chart now supports a hierarchical roofline, which represents additional levels in the memory hierarchy, in addition to device memory. You can now see if developed kernels have bottlenecks related to cache memory.
There are additional improvements including more configurable baseline comparisons, access to source-level information from the CLI, and additional SSH functionality.
What’s New for Gaming and Graphics Developers: Nsight Graphics 2021.5, Nsight Perf SDK, and Nsight Aftermath SDK
NVIDIA Nsight Graphics is a powerful tool that enables you to debug and profile applications that use Direct3D (11, 12, DXR), Vulkan (1.2, Vulkan Ray Tracing), and OpenGL. It provides the ability to export frames for later analysis, as well as GPU Trace, a powerful profiler that enables you to visualize GPU low-level metrics.
This latest Nsight Graphics 2021.5 release extends support for multiple APIs with the following updates:
- Full Windows 11 support for both API Capture and Tracing.
- Acceleration Structure Viewer: Bounding Volume Overlap Analysis.
- You are able to specify the continual refresh flag through the Nsight API.
- Support for Linux NGX.
NVIDIA Nsight Perf SDK is a graphics profiling toolbox for DirectX, Vulkan, and OpenGL, enabling you to collect GPU performance metrics directly from your application.
- Simplified APIs for HTML report generation.
- Lower-level range profiling APIs, with utility libraries providing ease-of-use.
- All of the preceding, usable in D3D11, D3D12, OpenGL, and Vulkan.
- Samples illustrating use cases in 3D, Compute, and Ray Tracing.
NVIDIA Nsight Aftermath SDK is a simple library you integrate into your D3D12 or Vulkan game’s crash reporter to generate GPU “mini-dumps” when a TDR or exception occurs.
- Support for Windows 11 for Nsight Aftermath available with Nsight Graphics 2021.5.
- Debug GPU exceptions through a detailed “mini-dump.”
- Contains the state of GPU pipeline subunits at exception time.
- Captures all active warps and current PCs.
- Map warp locations back to original HLSL/GLSL source code.
- Use markers to pinpoint the exception location in the API call stream.
- Helpful for debugging GPU exceptions during development, while in QA, or from deployed applications.
Efficient Model Design for In-App DL Inference: Nsight Deep Learning Designer 2021.2
Nsight Deep Learning Designer, is the first in class, IDE tool for developers who want to incorporate high-performance DL-based features into their applications.
It enables in-depth analysis of end-to-end DL workflows for efficient model design.
This release includes new features:
- Inference performance profiling with GPU metrics and Tensor Core Utilization.
- Analyze the model visually with the Channel Inspector.
- Compatible with PyTorch.
- Specialized Analysis Operators: Noise and Mix, Affine, Linear Blend, Resize, Selector, and many more.
Learn more details about Nsight DL Designer in this post.
Easy Software Development with IDE support: Nsight Visual Studio 2021.3, Nsight Visual Studio Code Edition 2021.1 and Nsight Eclipse
NVIDIA Nsight Visual Studio Edition is an application development environment that brings GPU computing into Microsoft Visual Studio IDE.
Nsight Visual Studio 2021.3 release provides Windows 11 support for full-fledged GPU kernel debugging and code inspection for bottlenecks, system utilization, and throughput improvements.
Nsight Visual Studio is included in the CUDA® toolkit, release 11.5, with bug fixes and performance improvements.
Nsight Visual Studio Code Edition is an application development environment for heterogeneous platforms that brings CUDA® development for GPUs into Microsoft Visual Studio Code.
Nsight Visual Studio Code Edition 2021.1 release includes features like IntelliSense support for smart CUDA code completion, debug CPU and GPU code in single session, remote development for cluster environments, and more.
Nsight Eclipse Edition is a full-featured IDE powered by the Eclipse platform that provides an all-in-one integrated environment to edit, build, debug, and profile CUDA-C applications. Some key highlights include the ability to provide seamless CPU and CUDA debugging, native Eclipse plug-in, and docker container support.
Getting Started with NVIDIA Nsight DevTools
Nsight Compute: Download, Documentation, Web Page, [GTC Session] Understanding CUDA Application Behavior, Performance, and Optimization Just Got Easier with the Latest Developer Tools, [NVIDIA DLI] Optimizing CUDA Machine Learning Codes with Nsight Profiling Tools, [Demo Video] Guided Analysis with Nsight Compute, Forum.
Nsight Systems: Download, Documentation, Web Page, DevNews, [GTC Session] Understanding CUDA Application Behavior, Performance, and Optimization Just Got Easier with the Latest Developer Tools, Forum.
Nsight Graphics: Download, Documentation, Web Page, DevNews, [GTC Session] Leveraging NVIDIA Graphics DevTools for High-performance Ray-tracing Applications, [NVIDIA DLI] Developer Tools Fundamentals for Ray Tracing Using NVIDIA Nsight Graphics and NVIDIA Nsight Systems, Forum.
Nsight Perf SDK: Download, Documentation, Web Page, Forum.
Nsight Aftermath: Download, Documentation (included in download package), Web Page, Forum.
Nsight Deep Learning Designer: Download, Documentation, Web Page, DevBlog, [GTC Session] Optimize Neural Networks for Quality and Performance with Nsight DL Designer, Efficient model design for in-app inferencing with Nsight DL Designer, Forum.
Nsight Visual Studio Edition: Download, Documentation, Web Page, Forum.
Nsight Visual Studio Code Edition: Download, Documentation, Web Page, [GTC Session] It’s Alive! CUDA in Visual Studio Code, Demo video: Nsight Visual Studio Code Edition, Forum.
Nsight Eclipse Edition: Download (part of CUDA toolkit installer), Documentation, Web Page, Forum.
Sign up for the Developer Newsletter to stay informed about new announcements and releases!
Check out Nsight DevTool day sessions and DLI courses at GTC to learn more!