GPU Support
Fransys allows provisioning GPU nodes in your clusters to run workloads requiring graphics processing power: machine learning, AI model inference, image or video processing, etc.
Availability
GPU instances are available on certain providers and in certain regions only. When creating a cluster, the wizard asks if you need GPU. If you answer Yes, only regions offering GPU instances are displayed.
On the interactive map, each hosting area indicates whether GPUs are available via the Has GPUs field.
Providers currently offering GPU instances:
| Provider | GPU Regions |
|---|---|
| Scaleway | France (Paris), Poland (Warsaw) |
GPU instance availability evolves regularly. Check the interactive map when creating a cluster to see the up-to-date options.
Configure a GPU cluster
Creating a cluster with GPU follows the same process as a standard cluster:
- Answer Yes to the GPU question in the wizard
- Select a GPU-compatible region on the map
- Configure your node pools with GPU node types (in Manual mode) or let AutoPilot manage it (by setting an appropriate budget)
In Manual mode, GPU types appear in the list when adding a node pool. They are identified by the gpu prefix and display the graphics card specifications in addition to CPU and RAM.
Use cases
- AI Inference — Deploy models (Ollama, vLLM, etc.) on GPU instances for fast response times
- Machine Learning — Train or fine-tune models directly on your cluster
- Media Processing — Video encoding, image generation, rendering
To deploy an AI model on GPU, use the Docker block with an Ollama image on a cluster with GPU nodes.