NVIDIA vGPU Grid Profile is missing

Hello, my organisation want to use vGPU.

We use VMWare vSphere 7.03 and vCenter 7.03. I allready have installed the driver over putty and the command nvidia-smi shows me the graphics card. The Problem is, that i can`t choose a profile in the configuration of the VM. I can choose nvidia grid, but thats all.

I have allready activate shared direct in the GPU Settings of the host. Direct IO is activatet too. The persistence-mode is on.

Can anyone help me with this issue?

Which GPU? Which server? Which driver version?
A bit more detail would help to get a better response

Regards Simon

Its an Quadro RTX 8000 in an Intel 3.Gen Server with ESXi 7.03.

The driver number is 14.1 (510.73.06) in the license portal.

Hi,

thanks for the additional information. But it would still be helpful to know which server (OEM, type). We have a HCL and only supported systems are tested. Often there are specific BIOS settings (MMIO, Above 4G and others) necessary, the OEM needs to advise in order to run the GPU properly with vGPU profiles.

In your specific case do you also have the Enterprise Plus license from vSphere in place?
Sometimes there is also an issue if you only have the GPU in one host in a cluster. You need to make sure that the VM is located on the right host, otherwise you also won’t see the profiles.

Best regards
Simon

Hey,

its an Intel Serverbarebone R2208WFTZSR with two Intel XEON Gold 6244.

Can you tell me what kind of bios settings are need to see the profiles?

Yes i have an Enterprise Plus license. There is no cluster at the moment. I have create an Datacenter in VCenter, but there is only one Host.

Do you need more informations?

Greetings
Tobias

Hard to say. Really depends on the OEM BIOS. But you could check this document from an example:

Above 4G decode needs to be enabled as example.

Thanks for the details. I try it tomorrow, when i am in the office.

I think there some options not configured in the BIOS.

Hello,

so i made it work now, but if I restart the vSphere Server its gone and the only way to solved this issue is to reinstall the driver.

The driver couldn`t found by nvidia-smi. The only way to find the driver is first to activate passtrough and then to deactivet it.

Unfortunately I assume this is related to the host you are using. This is not tested/supported and therefore there is not much we can do. There is the HCL for good reason. Please try to get some certified servers after your evaluation period to avoid such issues.

Best regards
Simon

The server from Wortmann TERRA SERVER 7220 G3 is supported by VMWare (HCL). Its only the Intel Barebone whats used.

You can see it here:
https://www.vmware.com/resources/compatibility/search.php?deviceCategory=server&details=1&partner=93&releases=578&keyword=TERRA%20SERVER%207220%20G3&systemTypes=2&cpuSeries=128&page=1&display_interval=10&sortColumn=Partner&sortOrder=Asc

Yesterday i had a call with VMWare support and they toke a look in the logs and make sure that all is configured the right way. The problem can´t solved by them. They say, there are no problems in VMWare and they think thats an Firmware Problem of the card who makes this issue.

Do you have any idea?

Sorry but I’m not talking about the VMWare HCL. As long as the server is not on our HCL it is not tested for GPUs and therefore not supported/working.
We don’t have firmware issues and there is no option for customers to modify the firmware (as not needed) so the VMWare response is not helpful nor correct.
Please check the link I posted previously with our HCL. If your server is not present there we cannot do anything further.

regards
Simon