JetPack 6.1 is now out and about in the world! This version of JetPack is for the NVIDIA Jetson Orin series, and will not work on earlier Jetsons. I’m especially interested in faster Argus library performance. Here’s the info lifted from the NVIDIA Jetson Developers Forum. Note that on the NVIDIA site there’s additional information and quick start install info:
We are pleased to announce the production release of JetPack 6.1. JetPack 6.1 packages Jetson Linux 36.4 with Linux Kernel 5.15 and Ubuntu 22.04 based root file system. The Jetson AI stack in JetPack 6.1 includes CUDA 12.6, cuDNN 9.3, TensorRT 10.3, VPI 3.2, DLA 3.1 and DLFW 24.0. This release supports all NVIDIA Jetson Orin modules and developer kits.
Highlights of JetPack 6.1:
- Upgraded compute stack with CUDA 12.6, TensorRT 10.3, cuDNN 9.3, VPI 3.2. DLA 3.1 and DLFW 24.0
- Camera: Enhanced Argus library, reducing CPU utilization by up to 40%
- Security: Support for Firmware-based Trusted Platform Module (fTPM) on the Orin platform. Refer to the security page for all security features.
- Software support for the PCN 211361 and fixes for various security vulnerabilities.
Installation:
You can install JetPack 6.1 with any of the methods below:
- SDK Manager: You can do a fresh install of JetPack 6.1 using SDK Manager.
- Debian Package: If you have JetPack 6 already installed on Jetson AGX Orin Developer Kit or Jetson Orin Nano Developer Kit, you can upgrade to JetPack 6.1 using APT. Refer to the steps here
- SD Card: If you are using Jetson Orin Nano Developer Kit, you can download the SD Card image from JetPack 6.1 page and use Balena Etcher to prepare the SD Card with JetPack 6.1
- Manual Flashing: If you prefer to install using the command line, you can flash Jetson device from a linux host by following steps here. Once Jetson Linux is flashed, you can install the compute stack using SDK Manager (using linux host) or by running “sudo apt update” followed by “sudo apt install nvidia-jetpack” on Jetson.
JetPack 6.1 Components:
- Jetson Linux 36.4
- CUDA 12.6
- TensorRT 10.3
- cuDNN 9.3
- VPI 3.2
- DLA 3.1
- DLFW 24.0
- Vulkan 1.3
- Nsight Systems v2024.5
- Nsight Graphics v2024.2
- Nsight Compute v2023.2 as part of the CUDA Toolkit
SDK Support:
- DeepStream 7.1 (coming in Q4 ‘24)
- Isaac ROS 3.0 (coming soon)
- Holoscan 2.6 (coming in Nov ‘24)
Containers: Containers will soon be available on JetPack 6.1. We will update the announcement as containers are made available.
JetPack 6 Resources:
Conclusion
As usual, make sure you have your backups made before upgrading. The best bits of the new release information are in the Release Notes, so make sure you read through those to find out what got fixed and new features added. Check it out!
5 Responses
I have come to the conclusion that this upgrade is too buggy to actually use. I realize that is a bit harsh and I need to explain. I upgrade my jetson to 6.1, After the upgrade, jtop shows the libraries have been upgraded, but it also says that the jetpack is not installed.
I teach a class with this jetson, and use nsight-compute. Nsight-compute consistely crashed on cuda examples that it had work fine with. I recompiled the examples, and discovered the recompiled code worked fine by itself, but nsight compute ran until the very end and then crashed. It offered to send a report to invidia. It did this for several cuda program that work fine by itself. (perhaps nsight compute also changes and requires some other compilation argument that was different?).
Is there a way to roll back the installation to the previous version?
I was fortunate that I as a instructor I have several versions of the jetson orin nano, and I was able to continue teaching with the older version of the jetpack.
The “JetPack not installed” issue is due to jtop not recognizing the new version. jtop needs to be upgraded. Because jtop (jetson-stats) is community maintained (not from NVIDIA), it usually takes a little while before the issue is resolved after a JetPack release. If needed, you can look at the source code and add the new release if really needed.
I don’t have enough experience with Nsight, but it probably is suffering the same type of issue and needs to be upgraded. You can check on the official NVIDIA forums to see what the NVIDIA engineers have to say.
I don’t know of any easy way to downgrade other than flashing the whole device. Sorry to hear you’re having issues.
Thank you for getting back to me this quickly. Jetpack 6.0 used nsight-compute version 2023.2.2. Jetpack 6.1 used version 2024.3.1. I will send a message off to nvidia to see if they can recommend a fix.
I wanted to add a comment. I filed a question about this on the Nvidia web site. On the same site they announced the Jetpack 6.1 and said that : JetPack 6.1 includes NVIDIA Nsight Compute v2023.2 as part of the CUDA Toolkit 12.2. The update that we installed had v2024.3.1. I am hoping to hear how I might role back nsight-compute on the jetson.
Interesting. That’s probably the source of the issues that you are experiencing. Hopefully NVIDIA can help you through the roll back process.