How to Install NVIDIA Drivers on Ubuntu 26.04 LTS (Resolute Raccoon)

You’ve got an NVIDIA GPU. You installed Ubuntu 26.04 LTS. And now your desktop feels… sluggish. The resolution is off. Animations stutter. And forget about gaming or any CUDA-based machine learning work.

That’s because Ubuntu ships with the open-source nouveau driver by default. It works for basic display output, but it’s nowhere near capable of unlocking your GPU’s full potential. You need the proprietary NVIDIA driver.

Installing NVIDIA drivers on Ubuntu 26.04 LTS (Resolute Raccoon) has gotten significantly easier, but there are a few 26.04-specific gotchas you need to know:

1. Secure Boot will block unsigned kernel modules
2. GNOME 50 is Wayland-only — NVIDIA’s Wayland support is now excellent, but requires driver 545+ (and 595 is the latest)
3. The open kernel modules (nvidia-open) are now the default for Turing (RTX 20-series) and newer GPUs

This guide walks through every method, from the dead-simple one-liner to the manual CUDA repo approach.

Before You Start: Disable Secure Boot (Or Enroll MOK)

This is the #1 cause of failed NVIDIA installations on Ubuntu 26.04.

NVIDIA’s kernel modules need to be signed to load with Secure Boot enabled. If you skip this step, you’ll boot into a black screen or fall back to software rendering.

Option A: Disable Secure Boot in your BIOS/UEFI settings (easiest, works every time)

Option B: Enroll a Machine Owner Key (MOK) during installation (see the troubleshooting section below)

Method 1: ubuntu-drivers CLI (Recommende)

Ubuntu 26.04 now includes recent NVIDIA drivers (up to 595) in its restricted repository. This is the simplest and most reliable method.

Step 1: Enable the restricted repository

First, check if restricted is already enabled:

grep -r "restricted" /etc/apt/sources.list /etc/apt/sources.list.d/

If not:

sudo add-apt-repository restricted
sudo apt update

Step 2: Find the right driver

List all available NVIDIA drivers for your GPU:

ubuntu-drivers list

On Ubuntu 26.04, you might need to include DKMS packages to see the latest 595 driver:

ubuntu-drivers list --include-dkms

Step 3: Install

The auto-install picks the best driver for your hardware:

sudo ubuntu-drivers autoinstall

Or target a specific version (e.g., the latest nvidia-driver-595):

sudo ubuntu-drivers install nvidia:595

For server/headless or compute-only setups:

sudo ubuntu-drivers install --gpgpu nvidia:595-server

Step 4: Reboot

sudo reboot

Step 5: Verify

nvidia-smi

You should see your GPU model, driver version, and CUDA version. A successful output looks like:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 595.xx    Driver Version: 595.xx    CUDA Version: 12.x          |
+-----------------------------------------------------------------------------+

Method 2: NVIDIA’s Official CUDA Repository (CUDA Users)

If you need the absolute latest driver or want CUDA toolkit packages alongside the driver, use NVIDIA’s official repo.

Step 1: Download and install the CUDA keyring

curl -sSo cuda-keyring.deb https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2604/x86_64/cuda-keyring_1.1-1_all.deb
sudo dpkg -i cuda-keyring.deb
rm -f cuda-keyring.deb

Step 2: Install the open kernel module driver

For Turing (RTX 20-series) and newer GPUs:

sudo apt update
sudo apt install -y nvidia-open

For older GPUs (GTX 10-series, Maxwell, Kepler):

sudo apt install -y cuda-drivers

Step 3: Reboot and verify

sudo reboot
nvidia-smi

Why Open Kernel Modules?

Starting with Ubuntu 26.04, NVIDIA and Canonical both recommend the open kernel modules (nvidia-open) for any GPU based on the Turing architecture or newer. These are NVIDIA’s official open-source GPU kernel modules, providing:

• Identical performance to the proprietary kernel modules
• Better debugging and community contributions
• Faster integration into Ubuntu’s kernel updates

For older GPUs (Pascal, Maxwell, Kepler), you still need the traditional proprietary modules (nvidia-driver-xxx).

Troubleshooting

“Secure Boot is enabled” error during install

If you kept Secure Boot enabled, the installer will prompt you to set a MOK enrollment password. This is not optional.

1. Type a password when prompted (remember it!)
2. Complete the installation
3. Reboot — you’ll see a blue MOK Manager screen
4. Select Enroll MOK → Continue → Yes
5. Enter the password you set earlier
6. Reboot again

Your modules will now be signed and loaded.

nvidia-smi returns “command not found”

The driver didn’t install correctly. Check:

dpkg -l | grep nvidia

If no packages are listed, go back and try again. Ensure restricted repo is enabled.

Black screen after reboot on GNOME 50

Ubuntu 26.04 uses Wayland-only GNOME 50. NVIDIA’s Wayland support is excellent on driver 545+, but if you’re on an older driver:

1. At the GRUB menu, press e and add nouveau.modeset=0 to the kernel command line
2. Boot, then install the correct driver using Method 1
3. After install, run sudo update-grub

Wayland vs. X11 on GNOME 50

Important: GNOME 50 has fully removed X11 from GDM. You cannot fall back to an X11 session anymore. But with NVIDIA driver 545+ (and especially 595), this is no longer a problem. Fractional scaling, variable refresh rate, and multi-monitor setups all work natively under Wayland.

If you need X11 for legacy application compatibility, you’ll need a display manager (like SDDM or LightDM) that still supports Xorg — but honestly, that’s a diminishing use case in 2026.

Final Thoughts

NVIDIA driver installation on Ubuntu 26.04 LTS is the smoothest it’s ever been, provided you handle Secure Boot upfront. The ubuntu-drivers tool is the recommended path: it queries your hardware, picks the correct driver, and integrates with your system’s update mechanism. The CUDA repository method is only necessary if you need the very latest driver or plan to use CUDA toolkits directly.

If you have a Turing or newer GPU, go with the open kernel modules — they’re the future and they work great. If you’re on older hardware, the proprietary modules still serve you well.

After the reboot, run nvidia-smi and enjoy that warm feeling of seeing your GPU fully recognized. Your system is now ready for gaming, rendering, or AI workloads. Whatever you throw at it.