Ubuntu 24.04 GPU Image for NVIDIA® GPUs

You get Ubuntu 24.04 LTS for Azure N-series GPU VMs with NVIDIA's signed 580 server driver, the CUDA 12.6 toolkit and the NVIDIA Container Toolkit, ready at first boot. The driver is Canonical's signed kernel module, so it loads with Secure Boot as well as without it. The current image was built on 18 October 2026 with the Ubuntu updates published to that date.

Who it's for

• Teams that run GPU workloads in containers and want the driver and container runtime support already in place
• Developers who build and test CUDA code on Azure GPU VMs

What's included

• NVIDIA driver 580.178.04 (signed server build) and CUDA toolkit 12.6.3
• NVIDIA Container Toolkit 1.20.1, for running GPU containers with your container runtime
• Ubuntu 24.04 LTS, Generation 2 VM, 64 GB OS disk by default

Not included

No deep-learning framework is preinstalled: install PyTorch, TensorFlow, cuDNN or NCCL yourself, or pull a framework container. Docker is not preinstalled either.

After deployment

• Choose an N-series GPU size and an SSH public key, and allow inbound SSH (port 22) from your own IP address only.
• Need room for model weights? Pick a larger OS disk at deployment or attach a data disk.
• Run nvidia-smi to confirm the GPU. On a size without a GPU, the image records that in /etc/app/gpu-status instead of failing.

Why this image

• Checked on a deployed test VM with an NVIDIA T4 GPU before publishing: driver loaded, CUDA sample run, and the Azure Marketplace certification rubric passed
• Email support at support@dcassociatesgroup.com and a quick-start page at www.dcassociatesgroup.com/support/

How it runs on Azure

The VM runs in your own Azure subscription and virtual network. Billing is pay-as-you-go on your Azure invoice: an hourly software charge per vCPU plus standard Azure compute and storage.

NVIDIA and CUDA are trademarks of NVIDIA Corporation. This image is built and supported by Derek Coleman & Associates Inc and is not affiliated with or endorsed by NVIDIA.