How to Install NVIDIA Drivers and CUDA 13.1 on Ubuntu 24.04

Setting up a GPU server on Ubuntu 24.04 becomes much easier when the installation process is done in the right order. Instead of using the standalone .run installer, a cleaner and more reliable approach is to use NVIDIA’s official APT repository. This makes package management, dependency handling, updates, and version control much easier over time.

This tutorial covers the complete setup process for installing NVIDIA Drivers, CUDA 13.1, and Docker GPU support with NVIDIA Container Toolkit on Ubuntu 24.04. It is especially useful for anyone working with AI training, inference workloads, CUDA development, rendering, or production GPU server deployments.

The guide explains how to first confirm that the NVIDIA GPU is detected properly, then install the required kernel headers, add NVIDIA’s official Ubuntu 24.04 repository, and install the driver. After that, it walks through installing either the CUDA runtime or the full CUDA 13.1 toolkit depending on the server’s use case. It also covers configuring the CUDA path, validating the installation using nvidia-smi and nvcc, and enabling Docker GPU support if needed.

Another useful part of the tutorial is that it highlights common mistakes that can cause issues during setup. These include installing the rolling cuda package instead of a pinned version, mixing APT installs with older runfile installs, adding unnecessary environment variables, and skipping reboot steps after installation.

For anyone planning to deploy a clean and maintainable GPU server on Ubuntu 24.04, this tutorial should be a useful reference.

Read the full tutorial here: https://www.servers99.com/tutorials/howto/install-nvidia-drivers-cuda-ubuntu-24-04/