I’ve had a few questions about what licensing is needed under the GRID 2.0 and up software licensing for the M60/M10/M6 GPUs for RDSH solutions such as XenApp. I think the confusion arises because it’s possible to use a number of GPU/vGPU different profiles for a server OS VM. The key point is to remember that the licensing is always per user.
- GPU-passthrough requires a per user license
- vApps licenses can be used to deliver server OS hosted desktops
- Multi-monitor and resolution support is different to VDI and limited by the virtualisation vendor e.g. VMware/Citrix.“Virtualized applications are rendered in an off-screen buffer. Therefore, the maximumresolution for the A series of GRID vGPUs is independent of the maximum resolution of the display head (of the physical GPU).”
My personal recommendation would be to favour vGPU above pass-through as the GRID 2.0 and up products (inc. M60/M6/M10) benefit from the recently announced monitoring enhancements in the GRID 4.0 (August 2016) software release.
Imagine you want 20 users to connect to a server OS VM shared under RDSH. So 20 user sessions per VM, here are your licensing options. (pGPU = physical GPU).
|Each server OS VM using||Licensing requirements for 20 XenApp users / sessions connecting to that one server OS VM||Quadro features||Monitoring features|
|M60 pGPU on passthrough||20 vApps
|Yes (only with vWS)||None it’s PCIe technically no host driver|
|M60 vGPU=8Q (vGPU)||20 vWS||Yes||Yes|
|M60 vGPU=8A (vGPU)||20 vApps||No||Yes|
As ever the licensing details can be found in the licensing guides on the resources page (under deployment guides) here: http://www.nvidia.com/object/grid-enterprise-resources.html. In particular these two are useful:
Update: to answer Tobias’ questions on resolution / multi-monitor support
If you trawl through the specs of the vGPU profiles in the licensing guides you will spot (or not) a teeny tiny “1” linking to a footnote. This gives the key information that Multi-monitor and resolution support is different to VDI and limited by the virtualisation vendor e.g. VMware/Citrix.“Virtualized applications are rendered in an off-screen buffer. Therefore, the maximumresolution for the A series of GRID vGPUs is independent of the maximum resolution of the display head (of the physical GPU).”
The resolution / multi-monitor support will be determined by the RDSH solution provider e.g. Citrix XenApp (for Citrix you can find info on XenApp monitor and resolution support here: http://support.citrix.com/article/CTX201696. In fact this non-depency of the M60-8A and similar profiles also applies to the M60-8Q and similar when used under XenApp with vGPU to deliver Quadro features e.g. enhanced CAD acceleration (there isn’t a teeny “1” in the licensing guide though). If it’s RDSH the multi-monitor/resolution support is dictated by the virtualisation stack.
The NVIDIA Knowledge Base also carries a lot of useful information on licensing, here: http://nvidia.custhelp.com/app/answers/list/st/5/kw/grid%20licensing/page/1
And you can ask questions on the support forums here: https://gridforums.nvidia.com/