NVIDIA Driver, CUDA 11.4, cuDNN v8.2.4 installation on Ubuntu 20.04

Uninstall previous versions

• First of all remove all related versions of NVIDIA driver, CUDA and cuDNN by executing the following commands.

sudo apt-get --purge remove "*cuda*" "*cublas*" "*cufft*" "*cufile*" "*curand*" \
 "*cusolver*" "*cusparse*" "*gds-tools*" "*npp*" "*nvjpeg*" "nsight*" "*nvvm*"
sudo apt-get --purge remove "*nvidia*" "libxnvctrl*"
sudo rm -rf /usr/local/cuda*
sudo apt autoremove

Installation of NVIDIA Drivers, CUDA and cuDNN

NVIDIA Drivers

• Check driver version for the CUDA version you are going to install. https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html

• After confirming the suitable version, go to this page and download appropriate driver for your GPU: https://www.nvidia.com/download/index.aspx?lang=en-us
  In my case, for example, I downloaded NVIDIA-Linux-x86_64-470.82.01.run
  

  Info: To install a specific nvidia driver version go to https://www.nvidia.it/Download/Find.aspx?
  

  Run the following command:
  

  sudo sh NVIDIA-Linux-x86_64-470.82.01.run --no-x-check

  Replace NVIDIA-Linux-* with the file you downloaded.

• You just need to approve and accept the prompts. Installation is pretty straightforward.

Install CUDA matching NVIDIA Driver version

• In this tutorial, I am going to install CUDA 11.4.3 (Although 11.5 is the latest one at the time of writing this tutorial).

wget https://developer.download.nvidia.com/compute/cuda/11.4.3/local_installers/cuda_11.4.3_470.82.01_linux.run

• After running the below command, make sure that you uncheck NVIDIA Driver option (as shown below) as we have already installed it in the previous step.

  Info: To install a specific version of the cuda toolkits go to https://developer.nvidia.com/cuda-toolkit-archive
  

  Note: Some versions not officially supported ubuntu 20.04 due to lack of some prerequisites, you can override the install-time prerequisite checks by running the installer with the flag. Remember that the prerequisites will still be required to use the NVIDIA CUDA Toolkit -override

  sudo sh cuda_11.4.3_470.82.01_linux.run

  Replace cuda_* with the file you downloaded.

• After successful installation, you will get a warning message (as shown below) that it did not install CUDA Driver, just ignore it as we had installed it separately before.

Add CUDA path to the bashrc file

• For programs to be able to recognize the location of CUDA, we must add PATH to the bashrc file. Execute the following commands to add CUDA path.

echo 'export PATH=/usr/local/cuda-11.4/bin:$PATH' >> ~/.bashrc
echo 'export LD_LIBRARY_PATH=/usr/local/cuda-11.4/lib64:$LD_LIBRARY_PATH' >> ~/.bashrc
source ~/.bashrc
sudo ldconfig

Installation of cuDNN matching CUDA version

• Installation of cuDNN is highly recommended to speed up calculations further. You must have an NVIDIA account to download cuDNN. Please create your account before downloading.

• Download cuDNN from here: https://developer.nvidia.com/rdp/cudnn-archive

- In this tutorial, I am going to install `cuDNN v8.2.4` for `CUDA 11.4`.

• There are several ways to install cuDNN but the simplest is to download the compressed file and extract to relevant CUDA directory.

tar -xzvf ${CUDNN_TAR_FILE}

sudo cp -P cuda/include/cudnn.h /usr/local/cuda-11.4/include
sudo cp -P cuda/lib64/libcudnn* /usr/local/cuda-11.4/lib64/
sudo chmod a+r /usr/local/cuda-11.4/lib64/libcudnn*

Verify Installation

• If you followed the above steps correctly, NVIDIA drivers, CUDA and cuDNN should have installed correctly.

• Execute nvidia-smi and nvcc --version to see if everything is working as expected and the CUDA versions match for them.

nvidia-smi

nvcc --version

Issues

Feel free to open issues if you face any problem. Thank you!