Install CUDA 8.0 and CuDNN 6.0


I need to install CUDA 8.0 and CuDNN 6.0 on a Jetson TX2. I don’t have physical access, only SSH, and I can’t flash Jetpack.

I have tried the following:

  1. Install JetPack 3.1 in an Ubuntu VM
  2. Go into the jetpack_download folder and copy all the arm64 packages to the Jetson filesystem via SSH: ``` scp jetpack_download/*arm64.deb jetson0:jetpack_download/ ```
  3. SSH into the Jetson node and install all the copied packages:
    ssh jetson0
    cd jetpack_download/
    sudo dpkg -i *
  4. Install CUDA 8.0:
    sudo apt update
    sudo apt install cuda-8-0

But I get an error saying that a package is missing:

$ sudo apt install cuda-8-0
Reading package lists... Done
Building dependency tree
Reading state information... Done
Some packages could not be installed. This may mean that you have
requested an impossible situation or if you are using the unstable
distribution that some required packages have not yet been created
or been moved out of Incoming.
The following information may help to resolve the situation:

The following packages have unmet dependencies:
 cuda-8-0 : Depends: cuda-runtime-8-0 (>= 8.0.84) but it is not installable
E: Unable to correct problems, you have held broken packages.

The package cuda-runtime-8-0 is indeed missing from the local CUDA repository. How can I make this installation work?

Hi s2wFHwZ,

I can’t answer this due to the VM is not supported and suggested, may other developers share experience with you.

The VM is not really important, all I did on it was install Jetpack to recover the .deb packages from the jetpack_download folder. My problem is the installation of CUDA 8.0 on the Jetson TX2. I assumed that there would be a cleaner way of doing it than my current method.

Unless the “.deb” files for CUDA 8 are coming from the same release flashed to the Jetson it won’t work, e.g., CUDA 8 belongs with a particular release, CUDA 9 with a different release, and CUDA 10 with yet another release.

It’s the same release, with CUDA 8.0

I found that I could install cuda-toolkit-8-0 instead of cuda-8-0; this works and installs everything I need.