How do I identify the GPUs installed in a dedicated server?
Check both hardware enumeration and the installed driver’s view. A PCI listing can identify a visible device without proving that its compute driver works. NVIDIA’s management utility can then report the GPUs recognised by its driver; other vendors require their own supported tools.
Before you start
- Authorised shell access and the hardware specification supplied for the server.
Instructions
1. Inspect visible PCI hardware
Run the basic listing without changing device configuration. Compare vendor and device identifiers with the expected GPU. Display controllers, compute accelerators and management graphics can appear differently, so inspect the full listing if a simple name search misses the device.
lspci -nnLists PCI devices with readable names and numeric vendor/device identifiers. It does not install a driver or enable a device.
2. Check the NVIDIA driver view where applicable
Use this only on a server expected to have an NVIDIA GPU and driver. Record the model and UUID for reliable identification; a numeric index can change after a reboot or hardware change.
nvidia-smi -LLists devices recognised by the NVIDIA driver, including applicable GPU or instance identifiers.
3. Reconcile the two views
If the expected hardware is visible on PCI but absent from the driver, check driver compatibility and relevant kernel logs. If inspecting a virtual machine hosted on the dedicated server, confirm device assignment with the hypervisor owner; guest visibility is not the host’s complete inventory.
Check the result
- The expected GPU model and stable identity agree with the server specification and the driver view.
Troubleshooting
- nvidia-smi is not found
- The utility may be absent or outside PATH. Confirm the installed driver and vendor before concluding that the server has no GPU or installing packages.
Official references
Use the documentation for your installed version when a flag, service name or platform feature differs.