How to install multi-version CUDA on Jetson AGX Orin?

[kzz]

I have an Jetson AGX Orin Developer Kit, with Ubuntu 22.04, Jetpack 6.1 and CUDA Toolkit 12.6. For Some projects I need CUDA Toolkit 11.8, so I want to install CUDA Toolkit 11.8.

I have referenced many workflows on the web, most of which don’t work or even destroy the original environment.

Perhaps the most authoritative tutorial is [CUDA Toolkit 11.8 Downloads | NVIDIA Developer], but

1. this workflow only provide package for Ubuntu 20.04, but my OS is Ubuntu 22.04
2. when I execute sudo apt-get -y install cuda, it still installs CUDA Toolkit 12.6. Besides, it makes my origin CUDA Toolkit disabled(can’ t access Unified Memory).

So who can tell me how to install multi-version CUDA on Jetson AGX Orin?

thanks a lot!

AastaLLL has provided the information about the compatibility of JetPack and CUDA

JetPack 5 / Ubuntu 20.04: v11.4-v12.2
JetPack 6 / Ubuntu 22.04: v12.2~]
(How to install multi-version CUDA on Jetson AGX Orin? - #4 by AastaLLL)

So installing CUDA v11.8 on JetPack 6 is impracticable.

[carolyuu]

Hi,
Here are some suggestions for the common issues:

1. Performance

Please run the below command before benchmarking deep learning use case:

$ sudo nvpmodel -m 0
$ sudo jetson_clocks

2. Installation

Installation guide of deep learning frameworks on Jetson:

• TensorFlow: Installing TensorFlow for Jetson Platform - NVIDIA Docs
• PyTorch: Installing PyTorch for Jetson Platform - NVIDIA Docs
  We also have containers that have frameworks preinstalled:
  Data Science, Machine Learning, AI, HPC Containers | NVIDIA NGC

3. Tutorial

Startup deep learning tutorial:

• Jetson-inference: Hello AI World guide to deploying deep-learning inference networks and deep vision primitives with TensorRT and NVIDIA Jetson
• TensorRT sample: Jetson/L4T/TRT Customized Example - eLinux.org

4. Report issue

If these suggestions don’t help and you want to report an issue to us, please attach the model, command/step, and the customized app (if any) with us to reproduce locally.

Thanks!

[AastaLLL]

Hi,

Sorry that CUDA 11.8 cannot work on the JetPack 6 environment.
Please find below for the details:

JetPack 5 / Ubuntu 20.04: v11.4-v12.2
JetPack 6 / Ubuntu 22.04: v12.2~

Thanks.

[kzz]

thanks but if I want to install CUDA v12.2 on my device described before, what should I do?

thanks!

[whitesscott]

You could try a dusty-nv/jetson-containers docker container with cuda11.8 and another with cuda 12.2 to see if that would work for you.

[AastaLLL]

Hi,

CUDA 12.2 is the default version of JetPack 6.0.
It is also the latest version that can be upgraded for the JetPack 5.x users.

We only support upgrading the library to a newer version (forward compatibility).
To downgrade CUDA for an older version might cause some unexpected problems.

Thanks.