A few weeks ago Citrix released a tech preview for vGPU, true hardware sharing of vGPU for XenDesktop on XenServer utilising the NVIDIA GRID technologies. At the time I blogged about how this augmented the Citrix product stack to offer complete solutions and an unrivalled product stack from Citrix for a whole range of CAD/CAE and visualisation uses. My colleague, Mayunk also wrote about the diversity of GPU solutions available for XenDesktop exploiting their HDX Pro technologies.
It’s now super exciting, only a few weeks in to the tech preview, to be seeing really positive industry and user reaction to vGPU from end-users both inside but also external to Citrix. The tech preview is proving to be extremely robust and the user feedback is extremely good. We knew we had a good solution but there is always an element of the unknown when you release a technology that your users will expect something different.
Today, I was delighted to get an email pointing out how well the XenServer solution performed in a comparison of the vGPU Tech Preview to vSphere vSGA by a Gartner analyst Gunnar Berger. It’s particularly satisfying to see something you’ve worked on performing so well, but in this case the sheer prettiness and visual attractiveness of Gunnar’s videos on YouTube made it much more tangible for all our teams. Gunnar’s videos are showing some really interesting results using the Unigine Benchmark to assess power of the XenServer/XenDesktop vGPU tech preview. I found it extremely interesting to read Gunnar’s blog about why he found vGPU so compelling, see here.
Update: Gunnar has added a second video to his blog:
- adding a second demo using the redsdk turbine benchmark from Redway3d, a more realistic benchmark for CAD and CAE users of products like Ansys, Solidworks, SolidEdge, Siemens NX and Catia etc.
- elaborating on the different solutions available from VMware and Citrix, vDGA, vSGA, vGPU, GPU-passthrough etc
- comparing the technologies under VMware View rather than XenDesktop to highlight the underlying difference and performance gains a lack of software emulation and API intercept makes at the hypervisor level for vGPU
Update: If you found this post of interest please come to our webinar on 11th December 2013, register here.