Skip to main content

Install CUDA

Set the os variable below to match your Ubuntu version (e.g. ubuntu2404 for 24.04, ubuntu2604 for 26.04):

# Set the Ubuntu version
os=ubuntu2604

# Download and install the CUDA repository keyring
wget https://developer.download.nvidia.com/compute/cuda/repos/${os}/x86_64/cuda-keyring_1.1-1_all.deb
sudo apt install ./cuda-keyring_1.1-1_all.deb

# Update package lists
sudo apt update

Choosing a Driver: Pinned vs Rolling​

Two ways to install the NVIDIA driver from this point:

  • nvidia-driver-pinning-XXX — pins to a specific major driver version (e.g. 610). Recommended in general, since CUDA Toolkit and TensorRT versions are tested against specific driver versions, and a pinned install won't silently jump to a newer (potentially incompatible) driver on a routine apt upgrade.
  • nvidia-driver-open — always installs whatever the latest open-kernel-module driver is. Simpler, but less reproducible.

Check what version is currently available before deciding:

apt policy nvidia-driver-open

This shows the candidate version apt would install (e.g. 610.xx-xxx). Use that major version number to pin:

sudo apt install nvidia-driver-pinning-610
tip

If you'd rather not pin and just want the latest driver, skip the pinning package and run sudo apt install nvidia-driver-open instead.

Install the CUDA Toolkit​

sudo apt install cuda-toolkit

Configure Environment Variables​

Configure CUDA environment variables system-wide by creating /etc/profile.d/cuda.sh:

sudo tee /etc/profile.d/cuda.sh << 'EOF'
# Runtime paths — required to run CUDA binaries and load shared libraries
export PATH=/usr/local/cuda/bin${PATH:+:$PATH}
export LD_LIBRARY_PATH=/usr/local/cuda/lib64${LD_LIBRARY_PATH:+:$LD_LIBRARY_PATH}

# Build-time paths — required when compiling C++ projects against CUDA (e.g., LibTorch)
export CPATH=/usr/local/cuda/include${CPATH:+:$CPATH}
export C_INCLUDE_PATH=/usr/local/cuda/include${C_INCLUDE_PATH:+:$C_INCLUDE_PATH}
export LIBRARY_PATH=/usr/local/cuda/lib64${LIBRARY_PATH:+:$LIBRARY_PATH}
EOF

Apply the changes by logging out and back in, or run:

source /etc/profile.d/cuda.sh

Validation​

nvidia-smi
nvcc --version