Can I add a GPU to a VPS for AI or graphics workloads?
A VPS needs an explicitly supported GPU allocation, such as device pass-through or a virtual-GPU profile, to provide suitable accelerator access. Installing a driver inside a normal VPS does not create a physical GPU or assign one from the host.
Instructions
1. Specify the accelerator workload
Record the software framework, required GPU model or features, device memory, driver stack and any licensing requirements. A virtual display adapter for console access is not the same as a compute accelerator.
2. Confirm the provider’s assignment model
Ask whether the GPU is exclusive, partitioned or shared, what memory and scheduling limits apply, and how the assignment affects migration or recovery. Confirm that the selected guest OS and driver version are supported.
3. Verify the assigned device through its official tools
For a service with an assigned NVIDIA GPU and an installed compatible driver, inspect the visible device.
nvidia-smi -LThis NVIDIA-specific command lists detected GPU devices. It is not expected to work on a VPS without the corresponding hardware and driver.
Check the result
- The visible device and memory allocation match the approved service specification.
- The intended framework passes a bounded functional test using the supported driver stack.
Troubleshooting
- The driver is installed but no GPU is detected
- Confirm actual device assignment and virtual hardware support with the provider.
- The workload requires a different GPU feature
- Check the exact model or virtual profile; a generic GPU label is not a compatibility guarantee.
Official references
Use the documentation for your installed version when a flag, service name or platform feature differs.