We recently upgraded our ESXi hosts and VMware Horizon 7.12-managed virtual desktops from vGPU software version 8.0 to version 10.2. According to the release notes for the software, I thought that this would allow the M10-1Q profile to allow up to 4 display heads.
However, when in Horizon, the "max number of monitors" is locked to 2. I’ve tried creating a new pool with the snapshot, but as soon as Horizon detects that the graphics drivers is the M10-1Q profile, it still locks the max number of displays to 2.
Any assistance understanding and/or fixing this would be appreciated.
On further testing, I note that I can in fact do at least 3 monitors, so I guess there’s no actual issue with the amount of pixels the profile will allow displays.
My issue, then, is the for non-persistent pools in VMware Horizon, in the pool settings the max number of monitors shill shows up as 2, and this is incorrect. My understanding is that the there is no maximum number of monitors, just a maximum number of pixels, so I guess there’s no "correct" value to put here, but perhaps it could be indicated somewhere in the documentation that Horizon will continue to display this value.
Back when vGPU 10.0 was released, NVIDIA switched from a "Fixed Resolution" to "Pixel based Resolution". They did this due to the large amount of non-standard panels that are becoming increasingly popular, such as the Super Ultra-Wide variants. I myself use a Dell U4919DW which has a resolution of 5120 x 1440. At first glance, the resolution looks quite high, however once you calculate the total amount of pixels, it’s not very high and is actually less than a single 4K panel. The problem is that because it’s a non-standard resolution, the previous versions of vGPU drivers wouldn’t support it, and when trying to run Virtual Machines in full screen, I’d have a full screen vertical display (1440), but with the black "bookend" on both ends of the screen as the vGPU driver had a maximum horizontal resolution of 4096 (and my monitor is 5120) so wouldn’t accept the resolution, even though that same version of vGPU driver would support multiple 4K monitors (up to 4) which as said, with even just 1 of those being a higher total pixel count than my U4919DW. The logical option for NVIDIA was to move to a far more flexible configuration, which is where we are today.
The reason for asking how many monitors you have and the resolution of them, is that with the 1Q profiles you can support up to 17694720 pixels. A few examples of common resolution options are available here: https://docs.nvidia.com/grid/10.0/grid-vgpu-user-guide/index.html#vgpu-types-tesla-m10 so you can calculate what you’ll be able to run.