NVIDIA JetPack SDK is the most comprehensive solution for building AI applications. Use the JetPack installer to flash your Jetson Developer Kit with the latest OS image, to install developer tools for both the host computer and developer kit, and to install the libraries and APIs, samples, and documentation needed to jumpstart your development environment.

NVIDIA JetPack is offered free of charge, but requires NVIDIA Developer Program membership. Not a member? Join the NVIDIA Developer Program here.

 Download *    Install Guide   Documentation 

Note: An Ubuntu host computer is required to run the JetPack installer. The installer can flash and update software on a target Jetson device, but it cannot not run directly on that device. When no target device is present, the installer can still be used to update software on the host computer.


4.0 Developer Preview Early Access

JetPack 4.0 Developer Preview EA includes support for the latest L4T BSP software packages for the Jetson AGX Xavier Developer Kit. Flash your Jetson Developer Kit with the latest BSP (L4T 31.0.1 for Jetson AGX Xavier) and install the latest software tools required to build and profile for applications on the Jetson Platform. This package contains CUDA toolkit for the host (Ubuntu) and target platform, the latest NVIDIA Developer Tools (NVIDIA Nsight Systems 2018.3 and NVIDIA Nsight Graphics 2018.4), VisionWorks 1.6, cuDNN v7.3, OpenCV 3.3.1, and TensorRT 5.0 RC.

Please follow the available guide to Download and Install JetPack. We encourage all users to report feedback, bugs and suggestions to improve the workflow experience of future versions of JetPack. Please use the Jetson forums. Also, a JetPack-specific feedback form is available here.


Release Highlights:

  • OS
    • L4T 31.0.1
      • Reference filesystem is now derived from Ubuntu 18.04
        • Unity desktop is kept as default for this release
  • Libraries and APIs
    • TensorRT 5.0 RC
      • Fine grain control for scheduling layers to execute on either iGPU or DLA
      • Support for both HMMA (FP16) and IMMA (INT8) on iGPU
      • "Early Access" DLA support for FP16
        • Supported networks for this release are AlexNet, googLeNet, and ResNet50
      • Capability to import quantized INT8 networks
    • cuDNN 7.3
      • Supports IMMA programmability
      • Grouped convolution performance improvements for NHWC input/output and FP16 and FP32 compute
    • CUDA 10
      • Supports IMMA programmability - direct use of these instructions dramatically reduces kernel execution time and kernel launch latency
      • Support for CUDA-Vulkan interoperability
  • Developer Tools
    • CUDA Tools
      • Nsight Eclipse plugins edition
        • True plugins can be installed in the developer's own Eclipse environment
        • Offers an all-in-one integrated environment to edit, build, and debug CUDA applications
    • NVIDIA Nsight Systems 2018.3
      • Profiling on Jetson AGX Xavier
      • Ability to trace cuDNN, cuBLAS, and OS runtime library API calls
    • NVIDIA Nsight Graphics 2018.4
      • Users can utilize the powerful Activities system to target their workflow for debugging and profiling
      • The Resources view allows you to edit the default shader used to display textures, and change the data display or highlight useful characteristics of the data
      • Improved editor capabilities including goto line and find/replace capabilities
  • JetPack Installer
    • Now supports Host computer running Ubuntu 18.04 or Ubuntu 16.04

Learn more about JetPack 4.0 Developer Preview EA

Early Access Notes

  • Power and Performance
    • Jetson AGX Xavier has been released to NVIDIA’s registered developers who have first access to this new product. In its current early access status, it is roughly 15x the performance of our previous Jetson TX2. This will continue to improve as Jetson AGX Xavier is tuned for better power and higher performance in future releases.
  • INT8 Support on DLA
    • Jetson AGX Xavier includes two Deep Learning Accelerators (DLA) can accelerate DL inference workloads with various data types. In this early access release, FP16 is supported. INT8 will be supported in a future release which should improve inference performance by at least 2x.
  • Vision Accelerator Support
    • Jetson AGX Xavier includes a 7-way VLIW Vision Accelerator (VA) for accelerating traditional computer vision workloads. Support for the VA will be included in a future release.
  • Fan speed optimized for quietness over temperature
    • The heat sink of the Jetson AGX Xavier Developer Kit may become hot to the touch (up to ~40C). This is expected behavior. The fan control has been optimized for noise over temperature. User-configurable settings will come in a future release.
  • Camera
    • The following camera features will be supported in a future release
      • Support for IMX185 sensor
      • Support for virtual channels (enables up to 16 cameras)
      • Support for SLVS-EC
  • TensorRT
    • When using DLA in this release, IExecutionContext::enqueue() blocks the calling CPU thread while DLA work is running. Asynchronous DLA execution in IExecutionContext::enqueue()will be supported in a future release. As a workaround for running multiple devices concurrently, execution contexts can be invoked from individual CPU threads.
    • For this release, the TensorRT builder writes a file named output.protobuf to its current working directory when compiling a network model for DLA. This means trtexec or any application using TensorRT with DLA will need write permissions to its current working directory. (E.g., this could be guaranteed by launching the application with sudo privileges.) This will be changed in the next release.
    • Python samples will be supported in a future release.
  • JetPack Installer
    • Please use the default download and install directories provided by JetPack. Changing the directories may cause an installation error.

Support

We encourage all users to report feedback, bugs and suggestions to improve the workflow experience of future releases of JetPack. A feedback form is available here.


To access older versions of JetPack, please visit JetPack Archive.