NVDIA Tesla V100 PCI-16GB card is not detected by NVIDIA control panel. Enviornment Azure GPU enabled VM-Standard_NC6s_v3 //RDP through Azure Bastion host//Windows 10 Enterprise

Question

[Size vCPU Memory: GiB Temp storage (SSD) GiB GPU GPU memory: GiB Max data disks Max uncached disk throughput: IOPS/MBps Max NICs
Standard_NC6s_v3 6 112 736 1 16 12 20000/200 4]

Windows 10 Enterprise Version 10.0.18362 Build 18362

Installed the NVDIA driver through extension blade from Azure portal. It deployed the standard driver- Control panel does not detect the card.
Updated the driver manually, by installing the driver -uppdated the standard driver with "511.65-data-center-tesla-desktop-win10-win11-64bit-dch-international.exe" .The driver got updated to 511.65 (30.0.15.1165)
How do we make NVIDA control panel detect the graphics card with driver standard grid driver or dch driver provided by NVIDIA website?
If this card is reporting 5120 CUDA cores, why do our application can see only one. ?
Expectation is that applications should see the Cores available in the system automatically.
We tried to see the 3D properties, but could not see anything.
Taskmanager does not detect the GPU in Performance tab.

Answer 1

Hello Mathew !

Thank you for posting on Microsoft Learn.

Here’s a breakdown of the issue and what you can do to resolve or properly configure your Azure VM to recognize and use the NVIDIA Tesla V100 GPU:

Try to open Device Manager and go to Display Adapters, you should see: NVIDIA Tesla V100 if not: The driver may not be installed correctly.

Open Command Prompt as admin and run:

nvidia-smi

You should see:

1 Tesla V100 GPU

• Driver version

Memory usage

CUDA version

If nvidia-smi returns an error or shows no devices, the driver is still not installed properly.

Another workaround, try:

nvcc --version

And then run a sample CUDA app, or use:

nvidia-smi -q -d COMPUTE

Applications won’t see individual cores (5120) directly, they see the GPU device. CUDA cores are not directly exposed like threads.

Since Tesla is a data center GPU, NVIDIA Control Panel is not relevant or required since it is designed for gaming/workstation cards.

Don’t expect display rendering or 3D visualization settings here and you can instead use nvidia-smi , CUDA toolkit utilities and application-level settings.