VMWare WPF vMotion (v)GPU Acceleration for vSphere 7 on Terminal Servers

Hello dear VMWare Community,

One of this annoying Developers speaking. We have the Problem that one of our 3rd Party Applications intensivly uses WPF (Windows Presentation Foundation – a 2006 Microsoft .NET Technology). WPF uses DirectX and the GPU to Draw the User interface. But normally the weakest onboard GPU from the last 10 years should be enough.

Now on our old Hyper-V Virtual Maschines and old Intel Processors (Xenon E5-2609 v2) the Application is extremly slow. I highly suspect the reason is that this Server CPU has no sufficent DirectX support in its (non existant?) onboard GPU or the onboard GPU is not virutalized. Since on modern physical Workstations the software is fast (Factor 3x). The Xenon is slow since WPF needs to be emulated on the CPU.

The application is not even graphically demanding (No CAD or 3D renderings) just badly programmed.
The developer of the software does not want to or cannot give us hardware requirements for Terminal usage.

Now we already ordered our new Server Hardware (AMD Eypc 16-core of newest generation – don’t know which excactly) and we don’t know how sufficent the onboard GPU of the Epyc will be for terminal server usage.

Our Head of IT-Hardware is very reculant about the topic since he is afraid vMotion of his 3 server cluster would not work anymore. I found this article that vSphere 6.7 supports hotswap with NVidea vGPU. But according to our head of IT vSphere 7 does not support this anymore. Is that true? I did not find anything about it so far.


Our Head of Hardware would insists on GPUs in the ballpark of Tesla V100S (3×9000 Euro) to be on the save side, which seems to me like overkill to me. We would start with about 10-20 Users on the Terminal Server.

Has maybe someone experience with WPF applications? What kind of GPU power they need for Terminal usage? We want to start with 10-20 Users on the Terminal Servers but maybe go up to 100. I know these are very vague numbers. But examples like hey I have this not very demanding application for 15 Users would be helpfull.

Thank you very much for assistance.

View Reddit by MichaelPeter91View Source

Related Articles


  1. There is no “real” gpu built into EPYC, not like on the ryzen APU line for desktop. As far as vMotion goes, you can leverage vGPU to maintain that ability, you just need to have the correct GPU’s installed in your servers and the licensing to go with it.





    Those Tesla GPU’s the correct ones, they are the lowest end GPU’s that can fulfill the requirement.




    Please bear in mind that the Tesla GPU’s cannot all do graphic (at least not well) as some are computational accelerators.

    It’s all about how many monitors the vGPU is capable of driving. At 100 Users, you will probably need more than 1


    For further confusing information, I recommend reading this document over the weekend.


Leave a Reply

Your email address will not be published. Required fields are marked *