VMware Horizon Community
whibr
Enthusiast
Enthusiast
Jump to solution

higher PCoIP supported 3D resolutions?

I was wondering if the maximum resolution or Max number of monitors supported has now increased with Horizon 6.1 and vSphere 6.0 vGPU technology?  It still seems to only show 2 monitors up to 1920 x 1200 resolution in the pool properties, when choosing PCoIP and 3D renderer.  We have some users with larger monitors and 2560 x 1440 resolution.

Reply
0 Kudos
1 Solution

Accepted Solutions
nwincey
Enthusiast
Enthusiast
Jump to solution


Yes, it is a little confusing as to which pieces work together and which do not.  I will explain how I understand them, but hopefully others with more knowledge on the subject will interject.

It appears you are using a ver 11 VM, since you can assign all the way up to 2048 VRAM.  So to start off you have the VM configured for vSGA and vGPU.  While this setup will work technically it wastes resources and only one virtual GPU approach should be used.  So if you want vSGA then delete the "shared PCI device" which is the vGPU portion.  When you check the "Enable 3D support" box that will enable vSGA mode.  Then you can select the 3D Renderer to be whatever you want.  If you want to allow your VM to work across ESXi hosts that have a real GPU (that's supported for vSGA) and ones that do not, then leaving the setting set to "Automatic" is best.  However we typically switch it to "Hardware" only because all our ESXi hosts have a real GPU in them (nVidia GRID K2).  The vSGA VRAM setting is the "3D Memory" setting.  If you had a ver 9 or 10 VM it would only allow 512MB max.  The "Total Video Memory" field is only for 2D virtual video card setups and used when just the standard 2D VMware vga adapter is utilized.  When using vSGA or vGPU I typically just set it to some small value between 5- 50MB.  You can reference another post I made about performance issues we have experienced when using the new VM ver 11 and vSGA =>https://communities.vmware.com/thread/506977  I haven't gotten much response on that thread so I'm not sure if others are seeing it in their apps or if its just us because something else is messed up.  We have been working with this virtual gpu technology for several years so we don't think its an error on our end, but I'm sure something will break loose eventually whether its from us figuring out our error or VMware finding a bug.  From that we only use VM ver 9 or 10 until we can figure out what's going on.

Ok so the vGPU section.  If you use the vGPU section then you don't need to check the "Enable 3D support" or any of the 3D video settings (since those are strictly for vSGA).  Just select the appropriate vGPU profile you want to use and the profiles will set the VRAM, max monitors and monitor res supported.  You can look at the attachment Tony referenced to see those settings.  All this is automatically set when you use vGPU based on the profile you select.  Now those are not hard limits...  for instance we use the k240Q profile with four monitors @1920x1080.  We didn't want the 2GB VRAM, but wanted the four monitor head support.  This was accomplished by setting VM config parameters in the VM settings area.  However you have to make sure you stay within the vGPU profile's VRAM buffer size for these alternative resolutions.  Also because we use a tera2 zero client our max monitor count with 2560x1600 is only two, so we can not utilize four monitors @2560x1600.  Those setups would require a standard desktop (as your client device) with the appropriate video card to display four monitors at that res and then utilize the software view client to get those vGPU VM setups working.

Regards,

Nathan

View solution in original post

Reply
0 Kudos
5 Replies
nwincey
Enthusiast
Enthusiast
Jump to solution

Whibr,

We are currently using tera2 zero clients to connect to our environment.  In those environments we run four 1920x1080 monitors and have also run two 2560x1600 monitors.  In the view connection server pool properties area where you set the "2 monitors" value there is also a choice to "manage with viclient" (something like that); select that and make sure in the viclient under VM settings you have four monitors configured and the appropriate vram (max 512MB) set for the VM.  With these settings as long as your client device (zero client or software view client) supports those desired monitor counts and resolutions you should be good to go.  This works with both vSGA, vGPU, and dGPU VM setups (we have used all three with our zero clients).  We are currently waiting for 4K capability, but that involves more pieces to all be working correctly before that resolution is supported.

Regards,

Nathan

Reply
0 Kudos
TonyHuynh201110
VMware Employee
VMware Employee
Jump to solution

If using NVIDIA GRID vGPU, you can get the following - see attachment

Reply
0 Kudos
whibr
Enthusiast
Enthusiast
Jump to solution

Interesting.  I don't actually have a larger monitor to test at the moment, I have just one 1920x1080 available.  So, in the vSphere 6 Web client, I am looking at the video card section of the VM settings (see pic).  When I change the number of monitors to 4, I can't enter any higher than 128MB for Total Video memory.  For 3D Memory, it will allow me to enter up to 2048MB.  Do you know how these different memory settings work?  This particular VM has the NVIDIA vGPU k260q profile assigned to it.

video-memory.png

Reply
0 Kudos
nwincey
Enthusiast
Enthusiast
Jump to solution


Yes, it is a little confusing as to which pieces work together and which do not.  I will explain how I understand them, but hopefully others with more knowledge on the subject will interject.

It appears you are using a ver 11 VM, since you can assign all the way up to 2048 VRAM.  So to start off you have the VM configured for vSGA and vGPU.  While this setup will work technically it wastes resources and only one virtual GPU approach should be used.  So if you want vSGA then delete the "shared PCI device" which is the vGPU portion.  When you check the "Enable 3D support" box that will enable vSGA mode.  Then you can select the 3D Renderer to be whatever you want.  If you want to allow your VM to work across ESXi hosts that have a real GPU (that's supported for vSGA) and ones that do not, then leaving the setting set to "Automatic" is best.  However we typically switch it to "Hardware" only because all our ESXi hosts have a real GPU in them (nVidia GRID K2).  The vSGA VRAM setting is the "3D Memory" setting.  If you had a ver 9 or 10 VM it would only allow 512MB max.  The "Total Video Memory" field is only for 2D virtual video card setups and used when just the standard 2D VMware vga adapter is utilized.  When using vSGA or vGPU I typically just set it to some small value between 5- 50MB.  You can reference another post I made about performance issues we have experienced when using the new VM ver 11 and vSGA =>https://communities.vmware.com/thread/506977  I haven't gotten much response on that thread so I'm not sure if others are seeing it in their apps or if its just us because something else is messed up.  We have been working with this virtual gpu technology for several years so we don't think its an error on our end, but I'm sure something will break loose eventually whether its from us figuring out our error or VMware finding a bug.  From that we only use VM ver 9 or 10 until we can figure out what's going on.

Ok so the vGPU section.  If you use the vGPU section then you don't need to check the "Enable 3D support" or any of the 3D video settings (since those are strictly for vSGA).  Just select the appropriate vGPU profile you want to use and the profiles will set the VRAM, max monitors and monitor res supported.  You can look at the attachment Tony referenced to see those settings.  All this is automatically set when you use vGPU based on the profile you select.  Now those are not hard limits...  for instance we use the k240Q profile with four monitors @1920x1080.  We didn't want the 2GB VRAM, but wanted the four monitor head support.  This was accomplished by setting VM config parameters in the VM settings area.  However you have to make sure you stay within the vGPU profile's VRAM buffer size for these alternative resolutions.  Also because we use a tera2 zero client our max monitor count with 2560x1600 is only two, so we can not utilize four monitors @2560x1600.  Those setups would require a standard desktop (as your client device) with the appropriate video card to display four monitors at that res and then utilize the software view client to get those vGPU VM setups working.

Regards,

Nathan

Reply
0 Kudos
whibr
Enthusiast
Enthusiast
Jump to solution

Nathan,

Thank you for the detailed and helpful explanation of how this works.  We have since been testing vGPU Horizon desktops @ 2560 x 1440, and other resolutions, with excellent results...and using the beta NVIDIA 348.20 driver as well.  So far, all is looking very good for supporting high-end resolutions for our engineers and designers.

Reply
0 Kudos