nvidia-driver

NVIDIA Driver + CUDA Toolkit Installer for Debian 11/12/13

This script installs the latest NVIDIA GPU drivers and the CUDA Toolkit on Debian 11 (Bullseye), Debian 12 (Bookworm), and Debian 13 (Trixie) — automatically and reliably.
It sets up the official NVIDIA repositories, installs drivers & libraries, and configures environment variables system-wide.

---

✨ Features

• ✅ Auto-detects Debian 11/12/13 and amd64/arm64 architecture
• ✅ Uses official NVIDIA CUDA APT repository
• ✅ Debian 13 fallback: tries debian13 repo first; if unavailable, falls back to debian12 with a warning
• ✅ Installs build tools: build-essential, dkms, linux-headers-$(uname -r)
• ✅ Installs driver + CUDA via the cuda meta-package
• ✅ Sets CUDA env vars via /etc/profile.d/cuda.sh
• ✅ Optionally blacklists nouveau to avoid conflicts
• ✅ Warns if Secure Boot is enabled

🧭 Debian 13 (Trixie) Notes

The script first tries NVIDIA’s debian13 repo. If it’s not available yet, it falls back to debian12. This often works, but compatibility depends on kernel & driver versions.

If the module fails to build or load, check kernel headers, Secure Boot, and nouveau status (see Troubleshooting).

🛡️ Secure Boot

If Secure Boot is enabled, DKMS may fail to load unsigned modules. Options:

Disable Secure Boot in firmware, or

Enroll a Machine Owner Key (MOK) and sign the module.

The script prints a note when Secure Boot appears enabled.

🚫 Nouveau (Open-Source Driver)

The script blacklists nouveau to prevent conflicts:

Config: /etc/modprobe.d/blacklist-nouveau.conf

Rebuilds initramfs automatically

If you prefer to keep nouveau, remove that file and run sudo update-initramfs -u.

🧰 Troubleshooting

Black screen or login loop: Likely driver conflict or Secure Boot. Boot to recovery/TTY, remove conflicting drivers, verify blacklist, check mokutil --sb-state.

DKMS build fails: Ensure headers match the running kernel:

uname -r apt-cache policy linux-headers-$(uname -r)

nvidia-smi not found: Ensure /usr/local/cuda/bin and driver packages are installed; re-login or source /etc/profile.d/cuda.sh.

🧩 Uninstall

To remove CUDA & drivers:

sudo apt remove --purge 'cuda*' 'nvidia*' sudo rm -f /etc/apt/sources.list.d/nvidia-cuda.list
/usr/share/keyrings/nvidia-cuda-archive-keyring.gpg
/etc/profile.d/cuda.sh
/etc/modprobe.d/blacklist-nouveau.conf sudo update-initramfs -u sudo apt autoremove -y sudo reboot

📦 Installation

1. Clone the repository or download the script: git clone https://github.com/dennishilk/nvidia-driver.git

   cd nvidia-driver

2. Make the script executable: chmod +x install-nvidia-cuda.sh

3. Run with root privileges: sudo ./install-nvidia-cuda.sh

4. Reboot to load the NVIDIA kernel module: sudo reboot

5. Check After reboot: nvidia-smi nvcc --version

   

No Warranty Disclaimer

The software in this repository is provided "as is", without warranty of any kind. I make no guarantees regarding the functionality, correctness, or suitability of this code for any purpose. Use it at your own risk. I am not responsible for any damages, data loss, or issues that may arise from using this software.