Run your AI models in Docker using any graphics card.
D2VGPU gives Docker containers access to your GPU on Windows. If you have an AMD, Intel, or NVIDIA graphics card plugged into your PC, your containerized AI workloads now run at full GPU speed instead of stalling on the CPU.
Why D2VGPU exists & how it bridges the gap.
Scroll through the pipeline to see how D2VGPU restores hardware access to your Windows container environment.
The Universal Hardware Disconnect
Docker is designed to package software so it runs identically on every computer. But on Windows, containerized software has had zero access to the host graphics card unless that card was NVIDIA, simply because only NVIDIA built that proprietary bridge into Docker Desktop.
If you have an AMD or Intel GPU plugged into your PC, your containers are locked out and forced to fall back onto a sluggish CPU. D2VGPU is that missing universal connection: with one part on the Windows host and one part inside Docker.
Implement the YAML File into Docker
Integrating D2VGPU into your project requires zero code changes to your software. You simply define the D2VGPU runtime inside your existing compose.yaml file and tag each container service that needs hardware access.
Whether deploying Ollama, PyTorch models, or custom Vulkan rendering jobs, marking the service guarantees immediate pass through routing without rebuilding images or changing dependencies.
Install Software
Install the lightweight Windows software. Operating cleanly in user space, it actively maintains a high speed shared memory connection between Docker containers and your physical graphics card.
Near-Native Speeds: D2VGPU achieves most of the card's native speed directly out of the box on AMD, Intel, and NVIDIA cards. Faster kernel level versions are in production and available on a per case direct request basis.
Good to Go
Start your containers with docker compose up or standard docker run. The container immediately detects your host GPU and initializes compute or graphics contexts seamlessly.
Your AI models run sealed inside Docker without security exposure to the host system, executing at full GPU speeds regardless of card manufacturer.
What D2VGPU means for your workflow
Zero Code Alteration
On Windows, Docker Desktop gives a container the graphics card only when it's NVIDIA. D2VGPU gives your container whatever card is in the machine. You install the Windows host service, mark your services in your compose file, and nothing changes in your application code.
Isolated Hardware Compute
You want to run local AI inside Docker so it remains sandboxed and sealed off from the rest of your operating system. Without D2VGPU on non-NVIDIA hardware, your models fall back to running sluggishly on the CPU. D2VGPU unlocks GPU acceleration instantly.
Vendor Agnostic Standard
Developers build and deploy in containers. Until now, no vendor agnostic solution existed for desktop GPUs on Windows. D2VGPU establishes a uniform layer that bridges all manufacturers across workstation and virtual machine environments.