In this blog post, I’ll guide you through the process of installing proper NVIDIA drivers and CUDA on Ubuntu 23.10.
Scenario
Let’s first understand the situation you might be facing:
- You own an NVIDIA GPU and need to install CUDA.
- You’re facing issues with missing NVIDIA drivers.
- The
additional driveroption in Ubuntu is displaying errors likepc-client-error-quark.
Figure: The Additional Drivers interface may display errors for all available NVIDIA drivers (both proprietary and open). This article provides a solution to resolve this issue using the terminal.
The Solution: A Step-by-Step Guide
Here’s a step-by-step guide to help you easily install recommended NVIDIA driver1. After that you can install by CUDA using nvidia-cuda-toolkit2 (not official).
Open a terminal and follow the steps.
Step 1: Upgrade Ubuntu distro and remove unnecessary packages
sudo apt dist-upgrade
sudo apt autoremove
Step 2: Check Ubuntu drivers for Nvidia and install recommended driver
ubuntu-drivers devices
Figure: Check the available NVIDIA drivers with one “recommended” text.
One of them is marked as recommended. Install that driver:
sudo apt install nvidia-driver-XXX
# Example
sudo apt install nvidia-driver-550
Step 3: Restart the machine
sudo reboot
After reboot if you check System Details from Settings > About > System Details, it will show the NVIDIA GPU:
Figure: After installation of proper GPU driver the system details show the GPU model
Step 4: Install CUDA toolkit
With the NVIDIA driver installed, proceed to install CUDA:
sudo apt-get install nvidia-cuda-toolkit
Confirm the GPU driver installation with nvidia-smi and check the NVCC compiler version with nvcc --version:
nvidia-smi
nvcc --version
Figure: Displaying the NVIDIA GPU and NVCC compiler version after driver and CUDA toolkit installation.
Conclusion
Follow these simple steps to install the proper NVIDIA GPU driver, even if you encounter errors with the Additional Drivers setting.
While the NVIDIA documentation3 provides an alternative installation procedure, nvidia-cuda-toolkit offers a straightforward solution.
The same approach may work in other Ubuntu versions as well Ubuntu 23.10.
References:
Advertisement




