What I'm talking about is designating one GPU for vSGA and pass-through the second GPU to a VM for high performance 3D work. In theory this should work, but I'm curious to know if anyone has actually done this or could contribute to this line of thinking. Right now I'm not chomping at the bit to buy a 3K+ card to test this theory out.
Regards,
Nathan
wponder,
Now that the K1/K2 cards are available is the plan going forward to include these new APIs to allow these two cards to work with vDGA and View? This feature is highly desirable from our standpoint as well as others I would imagine! If this is enabled we would certainly use this feature as the GRID cards are multi-gpu which offer a higher GPU density than the older Quadro cards.
mikeleahy1234,
Have you tried getting View to use the K1 (vDGA) again after a reboot of the VM like wponder states? We have not risked buying the expensive K1/K2 cards yet, because of the uncertainty on whether they would work or not with View 5.2 like wponder states?
Regards,
Nathan
yep - got it working
The document doesnt state that the montereyenable command has to be run on the desktop once you connect over pcoip - i think i was running that command while connected to the console
Also - i have to disable the vmware svga driver in device manager and then and only then does dxdiag.exe show K1 as being the card in use
So when i reboot teh vm then it stops working again and i have to re-enable svga and do the procedure again
@wponder - anything you can think of here would be great ?
also when i add in the pciHole.start line to the vmx file then the vm wont power on
get an unrecoverable errors /bora/chipset
These sound like symptoms caused by the K1 / K2 not being fully integrated with vDGA. With the quadro cards you do not have to do any of those things.
WP
mikeleahy1234,
Good to hear it works "somewhat" to play with at least. Thats a real bummer about it not being able to survive a reboot, hopefully support for these GRID cards will come soon.
Have you attempted to split up some of the K1's GPUs; that is use one for passthrough (vDGA) and another for vSGA?
I know Warren is the SME on this area but I just wanted to say I have had success splitting the K2, using one PCIe slot as direct and the other as shared. The shared is less than stellar on the performance, the direct was pretty good for 3d models. I even shot a video of me doing direct on youtube.
NVIDIA K2 on VMware View running Solidworks - YouTube
Right now I'm playing around with the Q4000 trying to get direct to work and having more issues than I did getting K2 to work in the direct mode
gunnarb,
Thank you very much for the info!
Did you have similar behavior like mikeleahy1234 did where the view config (using the K1 as primary display) did not survive a reboot of the VM? Since you have a Q4000 have you compared vSGA performance between the two (K2 and Q4000)? Im very interested to know if there is any GPU performance increase with using a K2 over the Q4000 (we are currently using one of these)? Any more info on this is greatly appreciated.
Regards,
Nathan
Blog coming sometime this week: http://www.youtube.com/watch?v=HTzdVGtx1hI
This is vDGA of the K2 doing DirectX 9. Works pretty good.
Great video! Definitely interested to see the same test done with vSGA on the K2 card. Then we could do the same test with our Q4000 to get a decent comparison between the two. Thanks again!
hi
thanks for your inputs. I have seen the same behaviour with svga. I am using the heaven benchmark and in svga the fps would not stay above 25 whereas with vdga then it does (albeit at a low resolution)
Are you seeing the issue that i am , having to disable the svga in device manager to force the nvidia to take over ? This is using k1 by the way
gunnarb,
Any updates with doing this same DirectX9 test with vSGA on the K2?
Gunnarb,
We now have a K2 (to go along with our Q4000) and I'm having the exact opposite problem you had. The Q4000 works fine in a dual monitor passthrough mode (using the monterey command) with view, etc... but when I do the same with the K2 card (single monitor) as soon as I log in using the view client (even a zero client) the screen blanks in/out about every second and the mouse cursor resets to the center of the monitor after each screen blank (basically rendering the desktop unusable). I've tried both nVidia driver versions 320 and 311.5 with no success. Do you have any ideas or advice on what you did with the K2 to have it work successfully (like what you show in your youtube video)? Thanks for any tips you can provide.
Regards,
Nathan
Nathan, there are some bugs being worked on for vDGA and K1/K2 and needs to be fixed before we can declare full support, the Quadros is pretty stable though.
// Linjo
I have K1 and K2 working perfectly now. You may need to upgrade the firmware and / or bios on the Grid cards.
mikeleahy1234,
Where did you get the firmware and or vbios for the GRID boards? Also, do you have more than one monitor assigned to the K1/K2 boards? Right now all I can get is an emulated vga connection. My old way of loading an EDID file with nVidia's driver under the "System Topology setting area" does not work since no other connection options are avaliable (DVI, etc...).
Regards,
Nathan
You will have to ask Nvidia for the bios / firmware of course.
Only using one monitor at the moment and all works perfectly. You need to go to screen resolution - detect and two more monitors will appear . Then select K1 / K2 etc and the card will become active
This part is missing from the vmwrae documentation
Ok, I'll contact nVidia.
I attempted the detect more monitors, but nothing pops up. The only option was "Connect to VGA anyway" and that was for the first monitor... I wonder if that is what the new firmware/vbios will take care of.
By the way, what firware/vbios version are you running? I can at least check that to see if I even need an updated version.
Thanks,
Nathan
This is what shows up on my system. The second monitor is the K2 card; in this screenshot I havent disabled the first monitor which is the VMware monitor. I was remoting into this vm with HP RGS because if I used view it would blank the monitor, etc.. like I was explaining above.
you need to use PCOIP for vdga
It sounds like you need bios / firmware upgrade
Yes, I plan on using PCoIP for vdga once I can fix the monitor blanking issues, etc..
If you look on my screenshot on the K2 properties window. What does your "Bios Information" field read?