Reinstalling Drivers

If your GPU encounters CUDA errors, Docker issues, or is not detected properly, reinstalling the NVIDIA drivers can resolve most problems. Follow these steps to safely remove, reinstall, and verify your drivers.

---

Disable Hosting & Docker Services

Before making any changes, stop all services to prevent conflicts during installation.

Run the following commands:

bashCopyEditsystemctl disable nebula-hosting.service
systemctl disable docker.service
systemctl disable docker.socket
reboot

This stops all GPU-dependent processes, ensuring a clean reinstallation.

---

Remove Old NVIDIA Drivers

A corrupted or outdated driver installation can cause stability issues. Completely remove all existing NVIDIA drivers before reinstalling.

Run:

bashCopyEditsudo apt-get remove --purge '^nvidia-.*' -y
sudo apt autoremove -y
reboot

This removes all NVIDIA-related packages and clears system dependencies.

---

Install the Correct NVIDIA Driver

After rebooting, install a fresh NVIDIA driver version that matches your system.

Option 1: Install the Latest Recommended Driver

Run:

bashCopyEditsudo apt install -y nvidia-driver-535
reboot

This installs the latest stable driver from Ubuntu's official repository.

Option 2: Install a Specific NVIDIA Driver Version

If you need a specific version, use:

bashCopyEditnvidia-driver-update 535.129.03 --force

Or download a driver manually from NVIDIA’s official site and install it with:

bashCopyEditnvidia-driver-update https://us.download.nvidia.com/XFree86/Linux-x86_64/550.67/NVIDIA-Linux-x86_64-550.67.run

Ensure that the driver version matches your CUDA toolkit version to avoid compatibility issues.

---

4) Verify Driver Installation

After installation, confirm that the system detects the GPU properly.

Run:

bashCopyEditnvidia-smi

You should see a table with GPU details, including driver version, power usage, and processes.

💡 If nvidia-smi returns ‘No devices found’, run:

bashCopyEditsudo modprobe nvidia

---

5) Re-enable Services & Restart System

Once the driver installation is complete, reactivate the hosting environment:

bashCopyEditsystemctl enable nebula-hosting.service
systemctl enable docker.service
systemctl enable docker.socket
reboot

Your GPU is now ready for hosting on Nebula AI.

---

6) Troubleshooting Common Issues

Issue	Cause	Fix
Black screen after installing NVIDIA drivers	Xorg (GUI) conflict	Run sudo dpkg-reconfigure gdm3 and reboot
nvidia-smi shows ‘No devices found’	Driver modules failed to load	Run sudo modprobe nvidia
Docker fails to detect GPU after driver update	NVIDIA runtime issue	Run sudo apt install --reinstall nvidia-container-runtime
CUDA errors when running workloads	Incompatible driver or CUDA mismatch	Install the correct CUDA version (apt install cuda-toolkit-12-2 -y)

---

Final Steps

Reinstalling NVIDIA drivers should resolve most GPU detection and performance issues. If problems persist, check logs:

bashCopyEditjournalctl -u nebula-hosting --no-pager | tail -n 50

or visit the Nebula AI Discord for live troubleshooting.