Skip to content

Remark Jetson Nano / 22.04 / CUDA / GCC 8.x #163

Closed

Description

@nmaas87

Dear all,
thanks for the awesome repo.
Just a remark for people coming after me: I built an image for the 4 GB / B01 / 300er revision Jetson Nano with Ubuntu 22.04 (because 24.04 is not supported with Jetpack 32.x/latest for Jetson Nano). Just a remark: Ubuntu 22.04 uses GCC 11.4.x, while nvcc needs GCC 8.x, meaning if you build with Ubuntu 22.04, you won't be able to use nvcc to compile your own CUDA code on the hardware, unless you get GCC 8 installed on the system. I have not tried to run existing CUDA kernels on the Jetson, but found it to be noteworthy. Never the less - awesome that you got it working and slimed down, would have loved to see 24.04 and above working with Jetson Nano, but looks like this hardware will be filling up landslides soon as Nvidia does not want to keep on supporting their hardware, but just keep on selling new stuff. There is probably a leason to be learned...
Thanks guys! 👍🏻

Activity

  1. mrcmunir commented on Dec 23, 2025

    @mrcmunir
    Contributor

    (because 24.04 is not supported with Jetpack 32.x/latest for Jetson Nano). Just a remark: Ubuntu 22.04 uses GCC 11.4.x, while nvcc needs GCC 8.x, meaning if you build with Ubuntu 22.04, you won't be able to use nvcc to compile your own CUDA code on the hardware, unless you get GCC 8 installed on the system. I have not tried to run existing CUDA kernels on the Jetson, but found it to be noteworthy. Never the less - awesome that you got it working and slimed down, would have loved to see 24.04 and above working with Jetson Nano, but looks like this hardware will be filling up landslides soon as Nvidia does not want to keep on supporting their hardware, but just keep on selling new stuff. There is probably a leason to be learned... Thanks guys! 👍🏻

    you can use Ubuntu 24.04 with some tweaks from nintendo switch repos
    https://github.com/theofficialgman/l4t-debs
    https://github.com/theofficialgman/l4t-debs-source
    Scripts image they are used from raspberry pi image.
    https://github.com/theofficialgman/l4t-image-buildscripts

    For Cuda, you're right, you need to have GCC 8 or lower installed on your system. What I do is have two or more versions on the system and then use sudo update-alternatives to point to two different versions in g++ or gcc, allowing me to switch between them as needed.

  2. Repository owner locked and limited conversation to collaborators on Dec 23, 2025
  3. converted this issue into a discussion #166 on Dec 23, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions