1. Release Notes

1.1. JetPack 4.2

1.1.1. Release Highlights

OS

  • L4T 32.1

    • LTS Kernel 4.9

    • CSI virtual channel support for Jetson AGX Xavier and Jetson TX2

    • RDMA support on Jetson AGX Xavier

      GPUDirect RDMA enables a direct path for data exchange between the Jetson AGX Xavier GPU and a peer device connected via PCIe. Applications on the peer device can directly perform PCIe based Read/Writes on the Jetson memory without CPU processing.

    • PCIe endpoint mode on Jetson AGX Xavier

      PCIe endpoint mode enables Jetson AGX Xavier to act as the PCIe endpoint while connected to another computer system acting as the PCIe root port. The root port system CPU is able to access and write to memory on the Jetson AGX Xavier. NVIDIA provides Linux drivers that implement a virtual Ethernet link over PCIe.

    • DRAM ECC Page Black Listing (PBL) on Jetson TX2i

    • OEM fusing tools updated to support Jetson AGX Xavier

    • New Jetson.GPIO Python library

      • Jetson.GPIO provides an easy way to control GPIOs. The library has the same API as RPi.GPIO. See /opt/nvidia/jetson-gpio/doc/README.txt on your Jetson system for details.

      • Adafruit Blinka now uses Jetson.GPIO to support Jetson systems.

    • Added support for more USB wireless adapters

Libraries and APIs

  • TensorRT 5.0.6.3

    • TRT Python API

      • Enables all the TRT Python samples; e.g., YOLO.

    • 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 7.3.1.28

    • 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 10.166

    • Supports IMMA Programmability — direct use of these instructions dramatically reduces kernel execution time and kernel launch latency

    • Support for CUDA-Vulkan interoperability

  • MultiMedia API 32.1

    • New MM sample shows how to exceed 16 stream decoding limit from upstream V4L2

    • nv3dvideosink is a new renderer plugin for gstreamer framework. This optimized, GPU-accelerated sink is an alternative to eglsink for higher performance video playback at lower power.

    • 10-bit decoding support in GstV4l2 decode

    • 12-bit decoding support in GstV4l2 decode (only for Jetson AGX Xavier and Jetson TX2 series)

Developer Tools

  • NVIDIA Nsight Systems 2019.3 for application profiling across GPU and CPU.

    • Improved correlation highlighting on the timeline.

    • Improved performance and responsiveness of the timeline.

    • Added support for launching profiler as superuser (root).

  • NVIDIA Nsight Graphics 2018.7 for graphics application debugging and profiling.

    • Added support for Vulkan API debugging and Pixel History.

  • [NEW!] NVIDIA Nsight Compute 1.0 for CUDA kernel profiling.

    • NVIDIA Nsight Compute (bundled with CUDA Toolkit) is a new interactive kernel profiler for CUDA applications. It runs on your Linux host computer, provides detailed performance metrics for analysis, and enables results comparison between baselines and the current run. Nsight Compute can be extended with analysis scripts for post-processing results. Also included is a command line tool that can be run on your Jetson system, or on your Linux host computer. This first version of NVIDIA Nsight Compute supports Jetson AGX Xavier.

 

Roadmap Notes

  • TensorRT

    • INT8 support on DLA

      Jetson AGX Xavier includes two Deep Learning Accelerators (DLA) that can accelerate DL inference workloads with various data types. FP16 is supported by this release. INT8 will be supported in a future release, which should improve inference performance by 2x.

    • In this release, running two DLA and iGPU does not yield expected sum of performance. This will be fixed in a future release. Note that performance from running two Deep Learning Accelerators at the same time sums up as expected.

      • To achieve expected performance in this release, use spin-wait based synchronization via cudaEventDefault flag when creating the events. For trtexec, use the --useSpinWait option. These workarounds may decrease multi-process synchronization time at the cost of additional CPU usage.

  • 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.

Notices

Notice

THE INFORMATION IN THIS GUIDE AND ALL OTHER INFORMATION CONTAINED IN NVIDIA DOCUMENTATION REFERENCED IN THIS GUIDE IS PROVIDED “AS IS.” NVIDIA MAKES NO WARRANTIES, EXPRESSED, IMPLIED, STATUTORY, OR OTHERWISE WITH RESPECT TO THE INFORMATION FOR THE PRODUCT, AND EXPRESSLY DISCLAIMS ALL IMPLIED WARRANTIES OF NONINFRINGEMENT, MERCHANTABILITY, AND FITNESS FOR A PARTICULAR PURPOSE. Notwithstanding any damages that customer might incur for any reason whatsoever, NVIDIA’s aggregate and cumulative liability towards customer for the product described in this guide shall be limited in accordance with the NVIDIA terms and conditions of sale for the product.

THE NVIDIA PRODUCT DESCRIBED IN THIS GUIDE IS NOT FAULT TOLERANT AND IS NOT DESIGNED, MANUFACTURED OR INTENDED FOR USE IN CONNECTION WITH THE DESIGN, CONSTRUCTION, MAINTENANCE, AND/OR OPERATION OF ANY SYSTEM WHERE THE USE OR A FAILURE OF SUCH SYSTEM COULD RESULT IN A SITUATION THAT THREATENS THE SAFETY OF HUMAN LIFE OR SEVERE PHYSICAL HARM OR PROPERTY DAMAGE (INCLUDING, FOR EXAMPLE, USE IN CONNECTION WITH ANY NUCLEAR, AVIONICS, LIFE SUPPORT OR OTHER LIFE CRITICAL APPLICATION). NVIDIA EXPRESSLY DISCLAIMS ANY EXPRESS OR IMPLIED WARRANTY OF FITNESS FOR SUCH HIGH RISK USES. NVIDIA SHALL NOT BE LIABLE TO CUSTOMER OR ANY THIRD PARTY, IN WHOLE OR IN PART, FOR ANY CLAIMS OR DAMAGES ARISING FROM SUCH HIGH RISK USES.

NVIDIA makes no representation or warranty that the product described in this guide will be suitable for any specified use without further testing or modification. Testing of all parameters of each product is not necessarily performed by NVIDIA. It is customer’s sole responsibility to ensure the product is suitable and fit for the application planned by customer and to do the necessary testing for the application in order to avoid a default of the application or the product. Weaknesses in customer’s product designs may affect the quality and reliability of the NVIDIA product and may result in additional or different conditions and/or requirements beyond those contained in this guide. NVIDIA does not accept any liability related to any default, damage, costs or problem which may be based on or attributable to: (i) the use of the NVIDIA product in any manner that is contrary to this guide, or (ii) customer product designs.

Other than the right for customer to use the information in this guide with the product, no other license, either expressed or implied, is hereby granted by NVIDIA under this guide. Reproduction of information in this guide is permissible only if reproduction is approved by NVIDIA in writing, is reproduced without alteration, and is accompanied by all associated conditions, limitations, and notices.

Trademarks

NVIDIA, the NVIDIA logo, and cuBLAS, CUDA, cuDNN, cuFFT, cuSPARSE, DIGITS, DGX, DGX-1, DGX Station, GRID, Jetson, Kepler, NGX, NVIDIA GPU Cloud, Maxwell, NCCL, NVLink, Pascal, Tegra, TensorRT, Tesla and Volta are trademarks and/or registered trademarks of NVIDIA Corporation in the Unites States and other countries. Other company and product names may be trademarks of the respective companies with which they are associated.