NVIDIA JetPack SDK is the most comprehensive solution for building end-to-end accelerated AI applications. JetPack provides a full development environment for hardware-accelerated AI-at-the-edge development. All Jetson modules and developer kits are supported by JetPack.

JetPack includes Jetson Linux with bootloader, Linux kernel, Ubuntu desktop environment, and a complete set of libraries for acceleration of GPU computing, multimedia, graphics, and computer vision. It also includes samples, documentation, and developer tools for both host computer and developer kit, and supports higher level SDKs such as DeepStream for streaming video analytics, Isaac for robotics, and Riva for conversational AI.

JetPack 5.0.2

JetPack 5.0.2 is a production quality release and replaces JetPack 5.0/JetPack 5.0.1 Developer Preview releases which were meant for development only. JetPack 5.0.2 includes Jetson Linux 35.1 BSP with Linux Kernel 5.10, an Ubuntu 20.04 based root file system, a UEFI based bootloader, and OP-TEE as Trusted Execution Environment. JetPack 5.0.2 includes the latest compute stack on Jetson with CUDA 11.4, TensorRT 8.4.1, cuDNN 8.4.1 See highlights below for the full list of features.

This release supports Jetson AGX Orin 32 GB production module and Jetson AGX Orin Developer Kit. It also supports Jetson AGX Xavier series and Jetson Xavier NX series modules, as well as Jetson AGX Xavier Developer Kit and Jetson Xavier NX Developer Kit.

Starting from this release, CUDA can be upgraded to newer versions without upgrading JetPack. Read the CUDA section in the highlights below for details.

Also starting with this release, NVIDIA SDK Manager offers the option of either a full installation of JetPack or the installation of JetPack runtime components only. The JetPack runtime installation does not include samples or documentation and takes only half the storage as a full JetPack installation.

Installing JetPack


Debian Package Method

JetPack can be installed or upgraded using a Debian package management tool running on Jetson. From JetPack 5.0.2 onwards, NVIDIA also hosts runtime and development debian meta packages for all JetPack components. The runtime packages do not include samples and documentations while the development packages include these and are intended for development. Users can install full JetPack or only runtime JetPack components over Jetson Linux.

Please refer to JetPack documentation for instructions on the list of debian packages we host. Also Refer to the JetPack documentation for instructions on how to install JetPack 5.0.2 on top of Jetson Linux or upgrade JetPack from JetPack 5.0/5.0.1 Developer Preview releases

SD Card Image Method

NVIDIA SDK Manager Method

Key Features in JetPack

Jetson Linux

NVIDIA Jetson Linux 35.1 provides the Linux Kernel 5.10, UEFI based bootloader, Ubuntu 20.04 based root file system, NVIDIA drivers, necessary firmwares, toolchain and more.

JetPack 5.0.2 includes Jetson Linux 35.1 which adds following highlights on top of Jetson Linux 34.1/34.1.1: (Please refer to release notes for additional details)

  • Adds Support for Jetson AGX Orin 32GB production module
  • Enables Jetson AGX Orin Developer Kit to emulate performance of Jetson AGX Orin 64GB, Jetson AGX Orin 32GB, Jetson Orin NX 16GB and Jetson Orin NX 8GB modules. Refer to the emulation section in release notes for detail.
  • Security
  • Over The Air Updates:
  • Bootloader:
    • UEFI support for USB and NVMe
    • UEFI support for 4K Display
  • Power:
    • Low Power SC7 state
    • Support for Power Estimator for Jetson AGX Orin and Jetson Xavier NX series
  • Display:
  • Kernel:
    • Real Time Kernel Patches
  • Sensor Processing Engine (SPE):
    • Sources for SPE firmware
    • Support for I2C and Inter VM Communication (IVC)
  • Check Multimedia and Camera section below for highlights
  • Jetson Linux Sources are now available on GIT in addition to the Jetson Linux page. Refer to section "Downloading Jetson Linux Sources" in release notes for detail.

1Signing (PKC) boot firmware upto UEFI supported for Jetson AGX Orin. Signing and Encrypting (PKC + SBK) boot firmware upto UEFI supported for Jetson AGX Xavier and Jetson Xavier NX

TensorRT

TensorRT is a high performance deep learning inference runtime for image classification, segmentation, and object detection neural networks. TensorRT is built on CUDA, NVIDIA’s parallel programming model, and enables you to optimize inference for all deep learning frameworks. It includes a deep learning inference optimizer and runtime that delivers low latency and high-throughput for deep learning inference applications.

JetPack 5.0.2 includes TensorRT 8.4.1

cuDNN

CUDA Deep Neural Network library provides high-performance primitives for deep learning frameworks. It provides highly tuned implementations for standard routines such as forward and backward convolution, pooling, normalization, and activation layers.

JetPack 5.0.2 includes cuDNN 8.4.1

CUDA

CUDA Toolkit provides a comprehensive development environment for C and C++ developers building GPU-accelerated applications. The toolkit includes a compiler for NVIDIA GPUs, math libraries, and tools for debugging and optimizing the performance of your applications.

JetPack 5.0.2 includes CUDA 11.4.14

Starting with JetPack 5.0.2, upgrade to latest and greatest CUDA releases from CUDA 11.8 onwards without the need to update Jetson Linux other JetPack components. Refer to instructions in the CUDA documentation on how to get the latest CUDA on JetPack.

Multimedia API

The Jetson Multimedia API package provides low level APIs for flexible application development.

Camera application API: libargus offers a low-level frame-synchronous API for camera applications, with per frame camera parameter control, multiple (including synchronized) camera support, and EGL stream outputs. RAW output CSI cameras needing ISP can be used with either libargus or GStreamer plugin. In either case, the V4L2 media-controller sensor driver API is used.

Sensor driver API: V4L2 API enables video decode, encode, format conversion and scaling functionality. V4L2 for encode opens up many features like bit rate control, quality presets, low latency encode, temporal tradeoff, motion vector maps, and more.

JetPack 5.0.2 Camera highlights include:

  • Argus support for YUV444 and Linear RGB888 output format for Jetson AGX Xavier and Jetson Xavier NX.
  • HDR Support for Jetson AGX Orin. Digital Overlap (DOL)1 Mode and Piecewise Linear (PWL) are supported.
  • Support for Error Resiliency on Jetson AGX Orin
  • New Samples:
    • argus_demosaicOutput to demonstrate the CVOutput capability.
    • argus_rawBayerOutput to demonstrate raw capture using argus with options available to enable/disable 3A/ISP to converge sensor exposure settings.
    • argus_userAlternatingAutoExposure to demonstrate the captures using alternating exposure.
    • argus_yuvOneshot now supports YUV444 format along with YUV420 format.

JetPack 5.0.2 Multimedia highlights include:

  • AV1 encoding/decoding on Jetson AGX Orin
  • Support for nvJPEG load balancing to use 2 nvJPEG engines in Jetson AGX Orin, providing 2X the encoding\decoding performance compared to JetPack 5.0/5.0.1 Developer Preview releases.
  • Pre-Processing Enhancements (PPE) module. This module provides a framework for various video encode enhancement algorithms. Currently in this release Temporal Adaptive Quantization (TAQ) is supported.
  • New NVUtils Multimedia buffer management and buffer transformation API2

1Further IQ improvements will be included in a future release

2nvbuf_utils is deprecated. Refer to migration guide to migrate from nvbuf_utils to NVUtils.

Computer Vision

VPI (Vision Programing Interface) is a software library that provides Computer Vision / Image Processing algorithms implemented on multiple hardware accelerators found on Jetson such as PVA (Programmable Vision Accelerator), GPU, NVDEC(NVIDIA Decoder), NVENC (NVIDIA Encoder), VIC (Video Image Compositor) and so on.

OpenCV is an open source library for computer vision, image processing and machine learning.

JetPack 5.0.2 includes VPI 2.1 with the following highlights:

  • New algorithms:
    • Fast Keypoint Detection on GPU and CPU backends
    • Image Channel Extraction on GPU and CPU backends
    • Median Filter on GPU and CPU backends
  • Unsigned Int8 support for Gaussian Pyramid on PVA backend
  • Unsigned/signed Int8 and signed Int16 support for KLT Feature Tracker on PVA backend

JetPack 5.0.2 includes OpenCV 4.5.4

Graphics

JetPack 5.0.2 includes the following graphics libraries:

  • New: Support for Vulkan® 1.3 (including the Roadmap 2022 Profile).
  • New: Support for Vulkan® SC 1.0
    • Vulkan SC is a low-level, deterministic, robust API that is based on Vulkan 1.2. This API enables state-of-the-art GPU-accelerated graphics and computation that can be deployed in safety-critical systems and that are certified to meet industry functional safety standards. Refer to https://www.khronos.org/vulkansc/ for more information.
    • Vulkan SC can also be invaluable for real-time non safety critical embedded applications. Vulkan SC increases determinism and reduces application size by shifting preparation of the run-time application environment either offline, or into application setup, as much as possible. This includes offline compilation of graphics pipelines that define how the GPU processes data, together with static memory allocation, that together enable detailed GPU control that can be rigorously specified and tested.
    • Vulkan SC 1.0 is evolved from Vulkan 1.2 and includes: the removal of runtime functionality that is not needed in safety-critical markets, an updated design to provide predictable execution times and results, and clarifications to remove potential ambiguity in its operation. For more details see https://www.khronos.org/blog/vulkan-sc-overview
    • Note: Jetson support for Vulkan SC is not safety certified.
    New: Support for OpenWF™ Display 1.0
    • OpenWF Display is a Khronos API for low overhead interaction with the native display driver on Jetson and allows interaction with Vulkan SC to display images.
    • Note: Jetson support for OpenWF Display is not safety certified.
  • New: Support for complete GBM API
    • Generic Buffer Management (GBM) is an API that provides a mechanism for allocating buffers for graphics rendering. The handle it creates can be used to pass directly to Weston/Wayland for rendering or initialize EGL and to create render target buffers.
    • Earlier releases used to support subset of GBM API called miniGBM. From JP 5.0.2 onwards, all GBM APIs are supported.
  • Support for OpenGL® 4.6
  • Support for OpenGL® ES 3.2
  • Support for EGL® 1.5

JetPack 5.0.2 includes OpenCV 4.5.4

Nsight Developer Tools

CUDA Toolkit provides a comprehensive development environment for C and C++ developers building high-performance GPU-accelerated applications with CUDA libraries. The toolkit includes Nsight Visual Studio Code Edition, Nsight Eclipse Plugins, debugging and profiling tools including Nsight Compute, and a toolchain for cross-compiling applications

NVIDIA Nsight Systems is a low overhead system-wide profiling tool, providing the insights developers need to analyze and optimize software performance.

NVIDIA Nsight Graphics is a standalone application for debugging and profiling graphics applications.

NVIDIA Nsight Deep Learning Designer is an integrated development environment that helps developers efficiently design and develop deep neural networks for in-app inference.

JetPack 5.0.2 includes NVIDIA Nsight Systems v2022.3

JetPack 5.0.2 includes NVIDIA Nsight Graphics 2022.3

JetPack 5.0.2 includes NVIDIA Nsight Deep Learning Designer 2022.1

Nsight System, Nsight Graphics, and Nsight Compute are all supported on Jetson Orin modules to assist development for autonomous machines.

Refer to release notes for more details.

Supported SDKs and Tools

NVIDIA DeepStream SDK is a complete analytics toolkit for AI-based multi-sensor processing and video and audio understanding.

A new DeepStream release supporting JetPack 5.0.2 is coming soon!

NVIDIA Triton™ Inference Server simplifies deployment of AI models at scale. Triton Inference Server is open source and supports deployment of trained AI models from NVIDIA TensorRT, TensorFlow and ONNX Runtime on Jetson. On Jetson, Triton Inference Server is provided as a shared library for direct integration with C API.

PowerEstimator is a webapp that simplifies creation of custom power mode profiles and estimates Jetson module power consumption.

JetPack 5.0.2 supports PowerEstimator for Jetson AGX Orin and Jetson Xavier NX modules

Cloud Native

Jetson brings Cloud-Native to the edge and enables technologies like containers and container orchestration. NVIDIA JetPack includes NVIDIA Container Runtime with Docker integration, enabling GPU accelerated containerized applications on Jetson platform.

NVIDIA hosts several container images for Jetson on NVIDIA NGC. Some are suitable for software development with samples and documentation and others are suitable for production software deployment, containing only runtime components. Find more information and a list of all container images at the Cloud-Native on Jetson page.

JetPack 5.0.2 Cloud Native highlights include:

  • New Containers on NGC:
    • JetPack container with all JetPack components like CUDA, cuDNN, TensorRT, VPI, etc within the container.
    • Cross Compilation Container which makes it simpler to cross compile JetPack and Jetson Linux components on a x86 host machine running Linux.
    • Jetson Linux Flashing Container which includes all the prerequisites required to flash Jetson modules and developer kits from a x86 host machine running Linux.

Security

NVIDIA Jetson modules include various security features including Hardware Root of Trust, Secure Boot, Hardware Cryptographic Acceleration, Trusted Execution Environment, Disk and Memory Encryption, Physical Attack Protection and more. Learn about the security features by jumping to the security section of the Jetson Linux Developer guide.

JetPack 5.0.2 Security highlights include:

  • Secure Boot1
  • Disk Encryption
  • Secure storage in OP-TEE using normal world (REE) file system

1Signing boot firmware upto UEFI supported for Jetson AGX Orin. Signing and Encrypting boot firmware upto UEFI supported for Jetson AGX Xavier and Jetson Xavier NX

Functional Safety

NVIDIA Jetson approach to Functional Safety is to give access to the hardware error diagnostics foundation that can be used in the context of safety-related system design. Jetson Safety Extension Package (JSEP) provides error diagnostic and error reporting framework for implementing safety functions and achieving functional safety standard compliance. Learn more.

JetPack 5.0.2 does not support JSEP.

JETPACK DOCUMENTATION AND WEBINARS

Documentation

Webinars

Support

    Visit NVIDIA Jetson forum for all your questions and issues. NVIDIA has a dedicated team of Jetson experts to support you on the forum.


For older versions of JetPack, please visit the JetPack Archive.