Hi,
I am in the process of building a VDI environment which requires GPU acceleration. I have a NVIDIA Tesla T4 card installed in a server running vSphere 6.7. vSphere has the NVIDIA GRID host driver installed which allows the Tesla card to be exposed as as Shared PCI Device. I tried adding a Shared PCI Device to a VM. But, vSphere has the option to add Shared PCI Devices as a greyed out option. I'm able to add the Shared PCI Device using vCenter which is tied to the vSphere instance. Is there a way to add a Shared PCI Device directly from vSphere? Or, is vCenter required?
Thanks,
Matt Villa
National TeleConsultants
Pretty sure you mean the Telsa card as a shared pci card. Did you follow all of nvidia's instructions, I haven't done a 6.7 one, but the 6.5 one I have as a POC I had to install vib before it works on the esxi hosts.
If I'm not mistaken, Horizon View is going to be required for this (which requires vCenter). Reference.
Actually I'm using Teradici instead of Horizon for VDI.
Not sure what you mean by that. Teradici makes hardware clients which usually work with Horizon to provide end-user access.
Pretty sure you mean the Telsa card as a shared pci card. Did you follow all of nvidia's instructions, I haven't done a 6.7 one, but the 6.5 one I have as a POC I had to install vib before it works on the esxi hosts.
I also don't know what teridicia software your using, but that doesn't keep you from adding the pci card, but esxi needs the vib and I think in some cases you need to set specifc options. Once that is done you can add the card and they get the different gpu profiles.
And I think you need vcenter
Yes I followed the instructions provided by NVIDIA. Their documentation says to add the Shared PCI Device by using vCenter. That option works from vCenter but not from vSphere. Trying to understand why. This is a POC environment which has only one vSphere. Ideally we could be able to make this change from vSphere.