How do I verify that CUDA and the NVIDIA drivers are actually working on a new GPU Droplet?

Hi there,

Run this first:

nvidia-smi

That confirms the driver is loaded and the GPU is visible. You will see the driver version, GPU model, and memory in the output.

For CUDA specifically:

nvcc --version

If you are using the AI/ML Ready image, both should work out of the box without installing anything extra.

For Docker with GPU support, the NVIDIA Container Toolkit should already be included in the AI/ML Ready image. Quick test to confirm:

docker run --rm --gpus all nvidia/cuda:12.0-base-ubuntu22.04 nvidia-smi

If that returns the same GPU info as the host, you are good to go. If Docker throws an error about GPUs not being available, run:

nvidia-ctk runtime configure --runtime=docker
sudo systemctl restart docker

That should be all you need before installing your ML stack.

Heya, @a89b0068f48c46e7b7b3e927903355

The fastest way to verify everything is working is to SSH into the Droplet and run:

nvidia-smi

If the GPU and driver are configured correctly, you’ll see details such as the GPU model, driver version, CUDA version, temperature, memory usage, and any running GPU processes.

If you also want to confirm the CUDA toolkit is available, run:

nvcc --version

or

cat /usr/local/cuda/version.json

(if the CUDA toolkit is installed).

For Docker workloads, you’ll typically need the NVIDIA Container Toolkit so containers can access the GPU. You can verify it’s working by running a GPU enabled container, for example:

docker run --rm --gpus all nvidia/cuda:12.9.0-base-ubuntu24.04 nvidia-smi

If that prints the same GPU information, your Docker setup is ready for GPU accelerated workloads.