Anyone knows when we can expect to be able to deploy a VM in Hyper-V 2016 and have a GRID M10 driver recognized inside the VM? (not via DDA but via RemoteFX so I can share GPUs across VMs)
I’m looking to deploy mass "slave" Windows VMs on bare-metal hosts and have graphics rendering inside those VMs offloaded to the GPU. GRID vGPU works best since I’m looking at LOTS of VMs per host.
I’m not sure if I fully understand your request. What is the issue with the current driver for M10? This works for sure also for RemoteFX.
Which driver did you try? I guess you’re just using the wrong driver. Even for RemoteFX you need vPC licenses for all the VMs and therefore the driver is only available via the Nvidia Enterprise Portal.
Just ran into a similar issue.
I have Dell R730 that a personally own. With a Tesla M2090 in it.
So if I slap an AMD graphics card or and Nvidia 1030 in it I can use Remote FX, but if I want to use my Tesla card I have to pay for the drivers or Nvidia won’t let me?
Partly correct. You cannot use Geforce for virtualization. And be aware that RemoteFX is deprecated with 2016 1709 and won’t be supported any more in future releases.
Ahhh Okay sorry. :)
Do you know what the new solution path is for vGPU going forward with Server 2016?
I’m testing in on my R730, but I’m tasked with figuring it out on Server 2016 and Vmware Horizons. I’ve got plenty of docs for VMware. But I’m not quite sure how I’m going to complete the task of using Nvidia products for vGPU performance on VM’s in Server 2016. Unless Microsoft is going to have something new come out some time soon, I guess I’m just pushed straight to Horizons only. :(
For RDSH I would always recommend to use the Tesla M10. You can already use DDA on Server 2016 to "Passthrough the 4 GPUs to 4 RDSH VMs and this should be more than sufficient for a single host.
A follow-on question regarding DDA, the M10 and setting up pass through. The M10 has 8 GPU’s so technically, I can setup 8 VMs each connected to a M10 GPU. Do I have that right?
I have 3 Dell R730’s in a failover cluster, each with M10 cards. They are running several VM servers for our school district. I would like to run 8 Win10 VM’s for student operations to drive remote access for running Adobe Premier and Maya.
Has anyone in this community been successful in a production environment yet?
Thanks in advance for any input you may have on this topic.
the M10 has 4 GPUs. For sure you can DDA the 4GPUs to 4 Win10 VMs but M10 is not meant for Adobe Premiere or Maya.
You should better wait until MSFT supports "vGPU" in a future Hyper-V release.