Cantech Knowledge Base

Your Go-To Hosting Resource

  1. Linux
  2. How to Install NVIDIA CUDA Toolkit on Ubuntu 22.04?

How to Install NVIDIA CUDA Toolkit on Ubuntu 22.04?

We’ve got you covered! If you are about to dive into a world of machine learning, AI development, or any form of GPU-accelerated computing, considering and installing the NVIDIA CUDA Toolkit should be one of the first tasks on your to-do list. You can consider this a quick instruction for installing the CUDA Toolkit on Ubuntu 22.04 as well as some important suggestions to get you and your system set up and running as soon as possible.

What is CUDA?

CUDA (Compute Unified Device Architecture) is a parallel computing platform and programming model developed by NVIDIA. CUDA enables developers to use a parallel computing approach to general-purpose computing on graphics processing units (GPGPU) and speed up workloads versus processing on a CPU.

Hence, its applications include:

  • Deep Learning and Neural Networks
  • Scientific Simulations
  • Big Data Analytics
  • Video Processing

Prerequisites

Before we install CUDA on Ubuntu 22.04, here’s what you need:

  • A CUDA-capable NVIDIA GPU (check list at official CUDA GPUs)
  • Ubuntu 22.04 LTS (clean installation preferred)
  • NVIDIA driver version 535 or newer
  • Internet access to download installation packages
  • Basic knowledge of using the Linux terminal

Step-by-Step: Install NVIDIA CUDA Toolkit on Ubuntu 22.04

Step 1: Remove Conflicting Packages

To avoid any issues during installation, remove old or conflicting NVIDIA and CUDA packages:

sudo apt-get --purge remove '*cublas*' 'cuda*' 'nsight*' 'nvidia*'

sudo apt-get autoremove -y

sudo apt-get clean

It ensures that the new CUDA installation starts from a clean slate.

Step 2: Install the Appropriate NVIDIA Driver

First, add the graphics driver PPA and install the latest driver:

sudo add-apt-repository ppa:graphics-drivers/ppa

sudo apt update

sudo apt install nvidia-driver-535 -y

Reboot your machine:

sudo reboot

After rebooting, confirm installation with:

nvidia-smi

You should see your GPU listed along with driver information.

Step 3: Download CUDA Toolkit from NVIDIA

Visit the official NVIDIA CUDA Toolkit download page and select:

  • OS: Linux
  • Architecture: x86_64
  • Distribution: Ubuntu
  • Version: 22.04
  • Installer Type: deb (local)

Then, run the following:

wget https://developer.download.nvidia.com/compute/cuda/12.2.0/local_installers/cuda-repo-ubuntu2204-12-2-local_12.2.0-535.54.03-1_amd64.deb

sudo dpkg -i cuda-repo-ubuntu2204-12-2-local_12.2.0-535.54.03-1_amd64.deb

Step 4: Add GPG Key and Update Repositories

To authenticate the repository:

sudo cp /var/cuda-repo-ubuntu2204-12-2-local/cuda-216F19BD-keyring.gpg /usr/share/keyrings/

sudo apt-get update

Step 5: Install the CUDA Toolkit

sudo apt-get -y install cuda

It installs CUDA, cuDNN, and other supporting packages.

Step 6: Set Environment Variables

Add the following lines to your .bashrc file to make the CUDA Toolkit available to your shell:

echo 'export PATH=/usr/local/cuda-12.2/bin:$PATH' >> ~/.bashrc

echo 'export LD_LIBRARY_PATH=/usr/local/cuda-12.2/lib64:$LD_LIBRARY_PATH' >> ~/.bashrc

source ~/.bashrc

You can also update /etc/environment for system-wide access.

Verify Installation

Check CUDA Compiler (nvcc)

nvcc --version

Expected output includes CUDA release version and build info.

Confirm GPU Status

nvidia-smi

It shows active GPU processes, memory usage, and driver versions.

Run CUDA Sample Projects

git clone https://github.com/NVIDIA/cuda-samples.git

cd cuda-samples/Samples/1_Utilities/deviceQuery

make
./deviceQuery

If everything works, you should see the output ending with Result = PASS.

Optional: Install CUDA via Conda

Using Conda is a great way to isolate your CUDA environment:

conda create -n cuda-env

conda activate cuda-env

conda install -c nvidia cuda

Use this approach if you want to avoid modifying your system-wide configuration.

In Conclusion

Installing the NVIDIA CUDA Toolkit on Ubuntu 22.04 enables you to utilize the power of GPU computing, for example, training your next deep learning model, running a computationally heavy simulation, or processing large datasets.

The most important point is to take your time, to take this guide in sequence, step by step and check each step before going to the next one. After you have made it through the installation and checked if CUDA is installed properly, you can begin developing high-performance GPU-accelerated applications.

Frequently Asked Questions (FAQs)

1. What is the latest CUDA version for Ubuntu 22.04?

Currently, CUDA 12.8 is the latest stable release for Ubuntu 22.04.

2. Can I install multiple versions of CUDA?

Yes, multiple instances of CUDA can be installed. However, be sure to manage environment variables carefully to avoid version conflicts.

3. Should I use the .run file instead?

The .run file offers more customization but can be harder to manage. The .deb (local) method is safer and more user-friendly.

4. What if nvidia-smi doesn’t show the GPU?

Double-check your NVIDIA driver installation and ensure Secure Boot is disabled (BIOS setting).

5. Do I need cuDNN?

If you want to use TensorFlow or PyTorch with CUDA, you should also install cuDNN for optimal performance.

May 17, 2025