NVIDIA Jetpack Reference Guide/Version index/Jetpack4: Difference between revisions

From RidgeRun Developer Wiki
No edit summary
No edit summary
Line 19: Line 19:


== New features introduced ==
== New features introduced ==
Jetpack 4.2 is the first production release of Jetpack 4. The key features of this release include updates to the Operating System (OS), TensorRT, cuDNN, CUDA, Multimedia API, Computer Vision, and Developer Tools.
=== OS ===
The main changes includes:
* Makes use of the Jetson Linux L4T 32.1, the production release for Jetson AGX Xavier, Jetson TX2, and Jetson Nano
* Use the Linux Kernel version 4.9
* Add support for CSI virtual channed for Jetson AGX Xavier and Jetson TX2
* RDMA support on Jetson AGX Xavier
* PCIe endpoint mode on Jetson AGX Xavier
* DRAM ECC Page Black Listing (PBL) on Jetson TX2i
* OEM fusing tools updated to support Jetson AGX Xavier
* New Jetson.GPIO Python library
* Added support for more USB wireless adapters
=== TensorRT ===
The main new features include:
* TRT Python API: Enables all the TRT Python samples
* TF-TRT included in monthly TensorFlow for Jetpack releases
* DLA support for FP16 -- AlexNet, GoogleNet, ResNet-50, and LeNet for MNIST networks have been validated
* Support for both HMMA (FP16) and IMMA (INT8) on iGPU
* Fine grain control for scheduling layers to execute on either iGPU or DLA
* New Caffe SSD and INT8 API samples
=== cuDNN ===
The main new features are:
* Supports IMMA Programmability
* Grouped convolution performance improvements for NHWC input/output and FP16 and FP32 compute
* Strided convolution now supported by FFT tiling algorithms
* Performance improvements for dilated convolutions and convolutions computed with Winograd transform
=== CUDA ===
The new features include:
* Supports IMMA Programmability. Direct use of these instructions dramatically reduces kernel execution time and kernel launch latency.
* Support for CUDA-Vulkan interoperability
=== Multimedia API ===
The key new features are:
* New MM API sample showing how to exceed 16 stream decoding limit from upstream V4L2.
* nv3dvideosink is an optimized, GPU-accelerated alternative to eglsink.
* 10-bit decoding support in GstV4l2 decode.
* 12-bit decoding support in GstV4l2 decode (Only for Jetson AGX Xavier and Jetson TX2 series).
=== Computer Vision ===
The main updates include:
* OpenCV: Now features GPU acceleration for real-time operation
=== Developer Tools ===
The new highlights are:
* Nsight Eclipse 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
** Improved correlation highlighting on the timeline
** Improved performance and responsiveness of the timeline
** Added support for launching profiler as superuser (root)
* NVIDIA Nsight Graphics
** First introduction of this tool


== Software components and versions ==
== Software components and versions ==

Revision as of 20:00, 28 August 2024






Supported platforms

The last minor version of JetPack 4 (4.6.5), supports the following Jetson boards:

New features introduced

Jetpack 4.2 is the first production release of Jetpack 4. The key features of this release include updates to the Operating System (OS), TensorRT, cuDNN, CUDA, Multimedia API, Computer Vision, and Developer Tools.

OS

The main changes includes:

  • Makes use of the Jetson Linux L4T 32.1, the production release for Jetson AGX Xavier, Jetson TX2, and Jetson Nano
  • Use the Linux Kernel version 4.9
  • Add support for CSI virtual channed for Jetson AGX Xavier and Jetson TX2
  • RDMA support on Jetson AGX Xavier
  • PCIe endpoint mode on Jetson AGX Xavier
  • DRAM ECC Page Black Listing (PBL) on Jetson TX2i
  • OEM fusing tools updated to support Jetson AGX Xavier
  • New Jetson.GPIO Python library
  • Added support for more USB wireless adapters

TensorRT

The main new features include:

  • TRT Python API: Enables all the TRT Python samples
  • TF-TRT included in monthly TensorFlow for Jetpack releases
  • DLA support for FP16 -- AlexNet, GoogleNet, ResNet-50, and LeNet for MNIST networks have been validated
  • Support for both HMMA (FP16) and IMMA (INT8) on iGPU
  • Fine grain control for scheduling layers to execute on either iGPU or DLA
  • New Caffe SSD and INT8 API samples

cuDNN

The main new features are:

  • Supports IMMA Programmability
  • Grouped convolution performance improvements for NHWC input/output and FP16 and FP32 compute
  • Strided convolution now supported by FFT tiling algorithms
  • Performance improvements for dilated convolutions and convolutions computed with Winograd transform

CUDA

The new features include:

  • Supports IMMA Programmability. Direct use of these instructions dramatically reduces kernel execution time and kernel launch latency.
  • Support for CUDA-Vulkan interoperability

Multimedia API

The key new features are:

  • New MM API sample showing how to exceed 16 stream decoding limit from upstream V4L2.
  • nv3dvideosink is an optimized, GPU-accelerated alternative to eglsink.
  • 10-bit decoding support in GstV4l2 decode.
  • 12-bit decoding support in GstV4l2 decode (Only for Jetson AGX Xavier and Jetson TX2 series).

Computer Vision

The main updates include:

  • OpenCV: Now features GPU acceleration for real-time operation

Developer Tools

The new highlights are:

  • Nsight Eclipse 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
    • Improved correlation highlighting on the timeline
    • Improved performance and responsiveness of the timeline
    • Added support for launching profiler as superuser (root)
  • NVIDIA Nsight Graphics
    • First introduction of this tool

Software components and versions

Libraries

Developr Tools

Jetson Linux