Hello, @Williams Roger ! Summarizing what we've discussed:
Why am I not seeing a GPU in the device manager after installing drivers on my NC24ads A100 v4 VM?
In this instance, while the driver install process is successful, the enable process is not. This is an issue on our end that affects the CUDA drivers on NC A100 SKU VMs.
Edit (3/13/25): The fix is live and you should now be able to deploy the NVIDIA GPU driver extension without issue:
As a temporary workaround, you can try GRID drivers instead of CUDA drivers. For installing GRID Drivers on NC SKU VMs, you need to use Azure CLI and an ARM template:
az vm extension set \
--resource-group <resource-group> \
--vm-name <vm-name> \
--name NvidiaGpuDriverWindows \
--publisher Microsoft.HpcCompute \
--settings "{'installGRIDNC':True}"
ARM Template:
{
"name": "NvidiaGpuDriverWindows",
"type": "extensions",
"apiVersion": "2015-06-15",
"location": "<location>",
"dependsOn": [
"[concat('Microsoft.Compute/virtualMachines/', <myVM>)]"
],
"properties": {
"publisher": "Microsoft.HpcCompute",
"type": "NvidiaGpuDriverWindows",
"typeHandlerVersion": "1.9",
"autoUpgradeMinorVersion": true,
"settings": {
"installGRIDNC":True
}
}
}
You can use nvidia-smi
to confirm installation at this point.
It is expected that after installing the NVIDIA GPU Driver Extension for Windows on your Standard NC24ads A100 v4 VM, you should be able to verify driver installation in the device manager:
Additional Reading: