There seems to be limited information on this, and the marketing material makes things more confusing so I thought I would post in hopes to get a definitive answer.
What value does GRID software provide to a 2016 Hyper-V environment? There isn’t even a deployment guide for which lets me believe it provides limited or no value. Unlike in VMWare, GRID on Hyper-V does not allow loading of local NVIDIA drivers in a VM only RemoteFX, correct? I can use DDA and carve off a GPU and assign it to a VM, but I can do this without GRID, correct? I really am not trying to bash GRID, just it does not seem to provide the value it does in VMWare or Citrix. Can someone please let me know the advantages of using GRID on Hyper-V (bare metal) vs just using native RemoteFX and/or DDA that Windows Server 2016 provides?
The answer is Yes you need a vApps license for each user running on the RDSH.
DDA is GPU Passthrough and you have the native NV drivers within the VM with all advantages. With RemoteFX you have an interception based software driver with some limitations.
I agree that it often makes not much sense to run GRID on Hyper-V as you cannot use vGPU and therefore you don’t get the required density especially for VDI deployment.
Can i use the Nvidia Tesla M10 without GRID software license to be used as vGPU with RemoteFX for my Windows Server 2016 Remote Desktop server on Hyper-V?
DDA is a function of 2016 Hyper-V and not of GRID. So you can use DDA and load native drivers in the VM without GRID and it will show all the performance benefits. I have not had any experience with vApps, but natively on Windows Server RDSH there is no hardware acceleration, but does vApps add this to Windows? Would be helpful if NVIDIA provided real documentation and not just marketing material.
@sschaber: Are you using Hyper-V (bare metal) with RDSH and vApps, or using a different hypervisor with Windows RDSH running as a VM?