I am planing to use tesla M10 cards with XenApp (shared desktops)
On VMeare Vsphere (Standard Edition) hypervisor
Since I have only Stnadard Edition I only can make a PCI passthrough of each GPU into XenApp VMs.
But after I installed NVidia driver, it shows that it want Grid Virtual Workstation license
But vApp license should also support GPU passthru
All that nvidia licensing is really confusing
And how I can verify that lisence is working
because even without any license it seams like GPU is working
Do I even need licenses in my scenario?
In license guide it says that GRID vGPU is a licensed feature.
I am not using any any GRID software (no GRID Manager on ESXi host) or any vGPU. I am using physical GPU and just a driver in guest VM.
And what the difference in features or anything else
in using physical PCI passthrough or using M10-8A profile
which basically both dedicate a single GPU to a VM?
one thing for sure Nvidia has really poor documentation
Hi Spiderman,
you need vApps licenses for the XenApp users (CCU). It doesn’t matter if you use Passthrough or vGPU from a GRID perspective. As soon as you use the Tesla M10/M6/M60 you need GRID licenses.
Hope this helps to make it clear. The documentation is not poor, I don’t agree in this but I agree that it is sometimes hard to find. We started a new docs page to make it more transparent and better to find:
Please give it a try. Click on the latest version and you will find the licensing guide as example that decribes what license you need for your szenario
What is the difference between 8A and Passthrough?
8A is a vGPU profile with all current and future advantages of the GRID software stack (monitoring, motion support (in the future) …) and Passthrough is PCI device mapping without software stack.
Regards
Simon
Ok I need Licenses, and need vApp ones
but when I get Tesla M10 into passthru, Nvidia control panel
gives me only 2 options - "Grid Virtual Workstation" and "Tesla (Ulicensed)"
http://imgur.com/wJMwYnn
SO I guess if I put Vapp licenses into Licenses server it want like it
since it would want a Workstation license
Now I have Trial licenses which is Workstation and it is working
But will it be working with VApp licenses?
So any insights how to use vApps licenses in passthru scenario
when NVidia control panel expects to see Workstation license ??
(talking about good documentation)
Can I just purchase vApp licenses and don’t bother to install them
since all is working well without any license
Will it be OK? or this will be violation of licence agreement?
Hi Spiderman,
as I already said the correct licensing is vApps and for sure this works and will be accepted from the NV control panel.
Regards
Simon
It is very confusing. As I understand the information from the topic https://gridforums.nvidia.com/default/topic/1971/xenapp/xenapp-7-14-1-and-nvidia-m10-licensing/ correctly, the Passthrough option results in having support for CUDA and this specific feature requires a Quadro vDWS license (NVIDIA License manager would not accept vApp). So if you have a M10 with vApp licenses, the only option to make it work is to select a M10-8A vGPU profile and that requires you to buy an expensive vSphere license (With XenServer, this would not be the case). Is there any option to use vApp licenses with passthrough (could be to disable CUDA support) in order to avoid the high costs for vSphere?