7

The official packages do not seem up-to-date (https://packages.ubuntu.com/search?keywords=nvidia-cuda-toolkit) but using the latest cuda seems to matter when using newer hardware.

Nvidia is making packages available, but there is no guarantee that they will work well or will integrate with Ubuntu's management of nvidia drivers.

Is there an effort to create an ubuntu package in a PPA so we do not have to install manually and work out conflicts individually? (the package would handle that)

lgautier
  • 215

2 Answers2

10

  1. Open the terminal and remove any NVIDIA traces that you may have on your system.

    sudo rm /etc/apt/sources.list.d/cuda*
sudo apt remove --autoremove nvidia-cuda-toolkit
sudo apt-get remove --autoremove nvidia-*

  2. Setup the correct CUDA PPA on your system.

    export DISTRO=ubuntu2004
export ARCHITECTURE=x86_64
sudo apt update
sudo add-apt-repository ppa:graphics-drivers
sudo apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/${DISTRO}/${ARCHITECTURE}/7fa2af80.pub 
sudo bash -c "echo deb\ http://developer.download.nvidia.com/compute/cuda/repos/${DISTRO}/${ARCHITECTURE}/\ / > /etc/apt/sources.list.d/cuda.list"  
sudo bash -c "echo deb\ http://developer.download.nvidia.com/compute/machine-learning/repos/${DISTRO}/${ARCHITECTURE}\ / > /etc/apt/sources.list.d/cuda_learn.list"

  3. Install CUDA 11.1 packages.

    sudo apt update
sudo apt -y install cuda-11-1
sudo apt install libcudnn8

  4. Specify the PATH to CUDA in the '.profile' file. Open the file by running sudo nano ~/.profile and add the following lines at the end of the file:

    # set PATH for cuda 11.1 installation
if [ -d "/usr/local/cuda-11.1/bin/" ]; then
    export PATH=/usr/local/cuda-11.1/bin${PATH:+:${PATH}}
    export LD_LIBRARY_PATH=/usr/local/cuda-11.1/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}
fi

    Use the keyboard combination Ctrl + O and after that press Enter to save the file to its current location. Use the keyboard combination Ctrl + X to exit nano.

  5. Restart with sudo reboot and check the versions of the installation.

    CUDA:

    nvcc  --version

    NVIDIA driver:

    nvidia-smi

    libcudnn:

    /sbin/ldconfig -N -v $(sed ‘s/:/ /’ <<< $LD_LIBRARY_PATH) 2>/dev/null | grep libcudnn
karel
  • 114,770
0

Thanks karel for the answer. In step 2 I faced a GPG error: "public key is not available".

I followed the comment of jtran1999 in the nvidia issue 1632 to proporly add the key and the CUDA PPA. Maybe it'll help others with the same error:

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-ubuntu2004.pin
sudo mv cuda-ubuntu2004.pin /etc/apt/preferences.d/cuda-repository-pin-600
sudo apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/3bf863cc.pub
sudo add-apt-repository "deb https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/ /"
sudo apt-get update

Then continued for step 3 and everything worked fine!!

Cristina Bazzano
  • 1