Gartner Blog Network


Virtual Desktop: GPU Testing

by Gunnar Berger  |  June 21, 2013  |  1 Comment

My lab in growing at an enormous pace. The culprit of this is my current mission to push the limits on what you can do with a GPU inside of a virtual desktop. One thing I’ve been bullish on is that a virtual desktop done right can give a user experience that is as good as or better than a physical desktop. All the stuff happening in the GPU space really shows what I mean by this. My hope is that these videos wow people when they see what is possible on a virtual desktop. Something else that should be impressive is that you won’t even notice the protocol in these videos, as the protocols do a fantastic job of pushing this high end content.

Without further ado I give you my first series of tests:

Direct X 9 – Pre-Beta from 2012

This first video I share as a point of reference. I started playing with GPUs on VMs well before they were public knowledge, and well before Gartner hired me. This is a video I uploaded about a year ago where I did a DirectX 9 test. The test goes horribly, and the only reason I make it public at this point is because I find it humorous to see how much has changed in a year. (Please only watch 5 seconds of this video, its very boring)

Direct X 9 Test: Experimental (Tech Preview)

As you can see the videos are night and day differences. They also aren’t apples to apples. I was using different GPUs different drivers, VIBS, servers, you name it and most likely it changed from the first test to the second. However, the goal was the same, you have a VM playing a DirectX game. Not bad considering this is less than a year of changes. (Again no need to watch the whole video, just click somewhere in the middle of it.)

Solidworks Test / Starcraft 2 Test:

The test I have been working on this week is running the NVIDIA GRID K2 on VMware View 5.2. I’m doing a vDGA test (or direct passthrough of the GPU to the VM — this is experimental but public). If you are looking for instructions on how to setup this up you can find them here.

In the original Solidworks test I had a few bugs that I ended up fixing in the second Solidworks and Starcraft 2 test. Again its all experimental however, its very interesting to see how GPU has the potential to change the VDI landscape. I’m doing high end stuff here, but I see a potential to just give this to everybody and make the transition from physical desktop to virtual desktops invisible to the users. I plan on doing lots of fun research around this and in future tests you’ll see me testing XenDesktop on top of XenServer (XenServer does not do API intercepts like View and vSphere do so a side by side test should be interesting). I also will move away from the games and continue to push other areas in graphics. I open it up to the community to suggest which tests you’d like to see next.

Test Suggestions:

  • More powerful model in Solidworks (like a car! BMW/Audi/Telsa)
  • Healthcare 3D imaging (This would be the ultimate test but getting a demo 3D image isn’t easy)
  • I’ll add more suggests as they come in

Category: gpu  shvd  vdi  

Tags: gpu  k2  nvidia  shvd  solidworks  vdi  vmware-view  

Gunnar Berger
Research Director
1 year at Gartner
14 years IT industry

Gunnar Berger is a research director for Gartner's IT Professionals service. He covers desktop, application and server virtualization ...Read Full Bio


Thoughts on Virtual Desktop: GPU Testing


  1. Barry says:

    We’ve done some testing of this technology from a PoC perspective. The “stack” we used was a Dell R720 server, K2 Grid graphics card, ESX 5.1, VDGA (Pass-thru), Win7 VMs, in a XenDesktop 5.3 environment. it took quite a bit of tweaking but we have it working for testing purposes only. We are seeing outstanding virtual desktop performance. And it seems like it will be cheaper to deliver than the high end engineering workstations we currently have.

    The workload is 3D visualization and computationally intensive apps in Oil and Gas. Think of subsurface visualization and interpretation apps.



Leave a Reply

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

Comments or opinions expressed on this blog are those of the individual contributors only, and do not necessarily represent the views of Gartner, Inc. or its management. Readers may copy and redistribute blog postings on other blogs, or otherwise for private, non-commercial or journalistic purposes, with attribution to Gartner. This content may not be used for any other purposes in any other formats or media. The content on this blog is provided on an "as-is" basis. Gartner shall not be liable for any damages whatsoever arising out of the content or use of this blog.