Linux & NVIDIA Drivers

Before you begin, make sure you have:

• A running Linux instance with an NVIDIA GPU on MegaTech
• SSH access to your instance (see How to SSH into your instance)
• Root or sudo privileges on your instance
• Basic familiarity with Linux command line

Select your Linux distribution for specific installation instructions:

After installing the drivers and rebooting, verify that the drivers are working correctly:

Check driver installation

nvidia-smi

This command should display information about your GPU(s), including the driver version, GPU model, and utilization. If you see this output, the drivers are installed correctly.

Check CUDA installation (if installed)

nvcc --version

This command displays the CUDA compiler version if CUDA is installed.

If you plan to use your GPU for machine learning or CUDA-accelerated applications, you'll need to install the CUDA Toolkit:

Ubuntu

# Add CUDA repository
wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-keyring_1.1-1_all.deb
sudo dpkg -i cuda-keyring_1.1-1_all.deb
sudo apt-get update

# Install CUDA
sudo apt-get install -y cuda-toolkit-12-0

# Add CUDA to your PATH (add to ~/.bashrc)
echo 'export PATH=/usr/local/cuda/bin:$PATH' >> ~/.bashrc
echo 'export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH' >> ~/.bashrc
source ~/.bashrc

CentOS/RHEL

# Install CUDA
sudo dnf install -y cuda

# Add CUDA to your PATH (add to ~/.bashrc)
echo 'export PATH=/usr/local/cuda/bin:$PATH' >> ~/.bashrc
echo 'export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH' >> ~/.bashrc
source ~/.bashrc

Arch Linux

# Install CUDA
sudo pacman -S cuda

# Add CUDA to your PATH (add to ~/.bashrc)
echo 'export PATH=/opt/cuda/bin:$PATH' >> ~/.bashrc
echo 'export LD_LIBRARY_PATH=/opt/cuda/lib64:$LD_LIBRARY_PATH' >> ~/.bashrc
source ~/.bashrc

nvidia-smi command not found

This usually indicates that the NVIDIA drivers are not installed correctly. Try the following:

• Make sure you've rebooted after installing the drivers
• Check if the NVIDIA kernel modules are loaded: lsmod | grep nvidia
• If no modules are shown, try loading them manually: sudo modprobe nvidia
• If you get errors, check your kernel logs: dmesg | grep -i nvidia

Failed to initialize NVML

This error may appear when running nvidia-smi. It usually indicates a driver-kernel mismatch. Try these steps:

• Reboot your system
• If that doesn't work, reinstall the drivers
• Make sure you're not using Wayland as your display server (some NVIDIA drivers have issues with Wayland)

Black screen after reboot

If your screen goes black after installing drivers and rebooting, it's likely a graphics issue. SSH into your instance and:

• Remove the drivers: sudo apt purge nvidia* (Ubuntu) or sudo dnf remove nvidia* (CentOS)
• Try installing a different driver version
• Check if your GPU is compatible with the driver version you're installing