New Jetson software, modules and pricing

NVIDIA is pleased to announce several exciting new Jetson product updates for production modules, developer kits, and the software stack used across the entire Jetson family. Jetson developer kits are intended for rapid prototyping and development, Jetson modules are production grade and used for final products, and NVIDIA JetPack SDK is the comprehensive software solution for both developing and deploying AI applications.

See the following updates being announced today:

  • JetPack 4.2.1 - New features include support for NVIDIA Container Runtime for Docker, NGC, FreeRTOS, and beta support for INT8 in Xavier DLA. This version of JetPack enables installation and use of the latest DeepStream and Isaac SDKs. Available in July 2019.
  • DeepStream SDK 4.0 - Delivers a complete streaming analytics toolkit that now runs across all NVIDIA GPU platforms from Jetson to Tesla. It features turnkey integration with Azure IoT, hardware acceleration for encode/decode and optical flow, and support for multiple heterogeneous cameras. Available in July 2019.
  • Isaac 2019.2 – New capabilities include a GPU-driven machine learning and computer vision acceleration, time synchronization for sensor fusion, new Gems for model-based pose estimation and feature detection, path segmentation and mono camera depth estimation and a navigation stack for the Isaac SIM environment. Available in July 2019.
  • Announcing Jetson AGX Xavier 8GB, a new lower-power and lower-price Jetson AGX Xavier module. It delivers up to 20 trillion operations per second (TOPS) of AI performance while consuming only 20W or less. It is fully hardware and software compatible with the existing Jetson AGX Xavier, and is available from distributors worldwide for $599 (1KU+ Qty) starting August 2019.
  • Other updates for Jetson production modules:
    • The original Jetson AGX Xavier module is now available at a new price of $899 (1KU+ Qty).
    • The new Jetson TX2 4GB module is now available for order and is priced at $249 (1KU+ Qty).
    • Jetson Nano modules will be available in August 2019 (update: now available from Arrow)
  • Increasing developer access to the Jetson platform with updated pricing for Jetson developer kits:

For more info about these Jetson product updates: https://www.nvidia.com/en-us/autonomous-machines/embedded-systems/

Thanks for the update Dusty!

For those of us that have a product in the pipeline for the Jetson module (using the new pinout), is there any chance of getting a pre-production module?

The next units available will be what ships from distributors for general access, we are trying to pull them in and get them shipped from the factory as soon as possible. As always there is a lead time to some components. Thanks for your patience and design with Nano!

I’m wondering what the best strategy is for keeping my Jetson Nano up to date? Some of this is that I am not sure which components are ones I am interested in. I’m using my Nano as a desktop running ubuntu 18.04. I update the ubuntu distribution using apt on an ongoing basis.

  • I am assuming that the JetPack has a new Linux kernel which includes new or improved drivers for the hardware?

  • Is there a reason I would want to download the latest Nano SD image? The issue is that I have installed multiple Ubuntu packages on my system and would have to re-install them.

  • If I download and use the Nvidia SDK Manager do I end up with the equivalent of starting with the new SD image, but keeping my currently installed Ubuntu packages.

I’m hoping you can just provide me with a list like

  • use apt update to keep the Ubuntu up to date
  • download and use the SDK manager every time JetPack is updated
    Thanks!

The latest SD card image would be based on the new JetPack 4.2.1, so it would contain the latest versions of the components included in JetPack 4.2.1 along with any enhancements and bug fixes.

For Nano, flashing with the SDK Manager is essentially equivalent to using the SD card image, with the exception that the SDK Manager allows you to select which NVIDIA components get installed (like CUDA toolkit, cuDNN, TensorRT, VisionWorks, ect). Like using the latest SD card image, flashing with SDK Manager will reset the filesystem, so you will want to backup any files that you want to save.

In the future we’ll be moving to OTA updates for the NVIDIA components, so a reset of the filesystem won’t be needed then, however for JetPack 4.2/4.2.1 this will still be the case.

Thanks, that was just the help I needed. Just to be clear, what I read was: with SDK Manager will reflash the root file system: all users will have to be installed, any tweaks to ubuntu will need to be re-tweaked, etc.

Any time frame for moving to OTA?
Thanks!

That’s correct, user packages and user configurations will need to be re-installed.

OTA is still in development so we don’t have a concrete release schedule for it yet, but we are planning by end of the year.

There’s a note that a future release of JetPack will support TX1. Any idea when that will be available? I would love to get my TX1 off the shelf :)

Not to mention the TX1 is still twice the performance of the Nano …

Yes, you will be able to install JetPack 4.2.1 to TX1.

@dusty That’s great news! My 2 Nanos are busy with edge computing experiments, but I have a lot of NLP work waiting for the Tx1 to do.

Hi dusty,

Any updates for DeepStream?
Especially Deepstream for Nano !

Thanks,

Hi ddpx2, the latest release timeframe for DeepStream 4.0 is tentatively later next week, stay tuned - thanks.

Hi Dusty,

Thanks for your update!

Hello, is the Deepstream 4 requires to be run on Jetpack 4.2.1? Couldn’t it be run on previous Jetpack 4.2 image on Nano?

Hi abushuvom, it is built against the latest versions of TensorRT, cuDNN, and CUDA that are included with JetPack 4.2.1, so JetPack 4.2.1 will be the supported version.

Update on JetPack 4.2.1 - we’re working on fixing a couple of last bugs to finalize the release, and regrettably the schedule has slipped slightly. JetPack 4.2.1 includes a raft of great features for all our Jetson products, and we look forward to sharing it soon - stay tuned!

“We regret to admit that mistakes were made.”

;-)

Ship it when it’s done. The yells when it’s wrong are much louder than the yells when it’s late.

@dusty_nv sorry again, I am very confused about the images.

  • How do I check jetpack version btw. I have image from 18th march. But newest on download page is from 31th may? What's the difference between those and is it affected by apt-get upgrades?
  • I won't be able to upgrade current image to 4.2.1 by any means?
  • Will 4.21 still be on kernel 4.9 with same ubuntu?
  • I noticed SDImage in Nvidia's nano course is marked DLI? What's it special for? I assume it's same 4.2 jetpack with some more packages, for example the one with jupyter notebook?
  • Any way to upgrade my standard 4.2 to follow the course?
  • Will the course still be relevant for 4.2.1? To do the course, would you suggest waiting for 4.2.1, do with special DLI image or upgrade normal one?
  • Thanks

    The newer JetPack 4.2 image from 5/31 has an updated bootloader – see this thread for more info: https://devtalk.nvidia.com/default/topic/1055264/l4t-r32-1-updated-u-boot-bootloader-for-jetson-nano/

    Upgrading to JetPack 4.2.1, you will need to re-flash SD card, so you will want to backup your files first. In the future, we are moving to OTA updates so re-flashing won’t be necessary.

    The DLI image is JetPack 4.2 plus PyTorch, Jupyter server, the notebooks, and other dependencies.