Minimal vGPU profile required for use with three screens on Tesla M6

NVIDIA Virtual GPU Forums Virtual GPU Talks with the Developers
#1

Hi,

For my client we are seriously considering the use of nVidia Grid technology. Our users (about 600 concurrent(!)) can be considered knowledge workers. They use a 2D GIS applicaton and sometimes streaming video CCTV, the rest of their apps is not graphically intensive. That is why we would like to use the profile with just 512 MB of memory, so that we can support 16 concurrent users per M6 card. The unusual thing about this client is the use of multiple screens. The thin clients used will (based on the current plans) be equipped with 3 screens, one 34 inch with a resolution of 3440x1440 and two with a resolution of 2560x1440 (so somewhat lower than the maximum supported resolution of UHD). Based on the information on this page: http://www.nvidia.co.uk/object/buy-nvidia-grid-uk.html, there seems to be a limit, with implies that with the use of more than two screen, you need to use the profile with 2GB of memory or more. This would imply that we can only support 4 concurrent knowledge workers per M6 card. This makes the business case for the use of NVidia Grid much less interesting. Could you please let me know if there is a possibility to use the 512 MB (or 1024 MB) profile in combination with the three mentioned screens? Are the limits of two and four screens hard technical limits, or also dependent on the resolution used?

I would very much appreciate a serious answer, given the size of this project. Thank you very much in advance!

With kind regards,

Roel van Vuuren

#2

This is correct and is a fixed number of displays regardless of their resolution.

Some further information can be found in the documentation here

http://us.download.nvidia.com/Windows/Quadro_Certified/GRID/354.56/ESXi-6.0/352.70-354.56-nvidia-grid-vgpu-user-guide.pdf

For the solution you’re describing you’ll require a 2GB vGPU Profile licensed for Virtual Workstation Extended as you require 3 displays, one of which exceeds 256x1600.

#3

Dear Jason,

Thank you for the clear answer, unfortunately exactly the answer we were fearing, but very understandable nonetheless. Since this seems to be a software limitation only, I wonder if a formal request for a custom profile might have a chance. Based on the fact there is no suitable profile (only one with performance overkill for our usecase), isn’t unique, and has quite a large potential usergroup, I hope you can take this into consideration. If not with this card then perhaps with a next release.

My usecase is emergency room personnel (911 dispatch for Police, Firedepartment and Ambulance), which require the concurrent use of multiple applications (GIS, dispatch system, Office, CCTV, etc.). They require a system with multiple screens (more than 2, currenty 6 19 inch, going back to 3 big screens) with some graphical performance, but not comparable with 3D designers. The requirements for the virtual desktops are not that high either, about 2 VCores and 8 GB of memory will do. That is why the use of Grid cards with 2 GB of memory would be quite devastating for the consolidation ratio (VM’s per server) as well. For my client in a small country like the Netherlands alone we require at least 600 concurrent virtual desktops. I don’t imagine the sale of 40 M6 cards will lead to a stock rise for NVidia, but this specific use case and its importance for national security might be of interest.

Could you please let me know if, given this usecase, there is a chance of success if we submit a formal request for a custom profile which will fit our (and probably others) needs. If you think that Nvidia Grid isn’t designed or suited for this usecase, also please let me know.

Thank you very much again!

With kind regards,

Roel van Vuuren
IT Infrastructure Architect
MDC (part of Dutch Police)