Remote Visualization using M2070Q Virtual remote display

Hi All,

I am not sure if it is the right place to ask this question. But, as it is related to Tesla, I am asking this question here.
We have a solution for remote graphics visualization. Till now we have been using Quadro cards on remote server. We used to extend monitors to these cards using dummy vga dongle on windows.

Now, we are trying to replace Quadro cards with Tesla M2070Q cards. Before, I start explaining the problem following is the machine configuration:

OS: Windows server 2008 R2 64-bit
Graphics: Onboard VGA - ATI ES1000 (there is no other discrete card for display)
Number of Tesla M2070Q: 3

When I install latest nvidia driver (270.90) and open nvidia control panel - it gives an error saying no display attached to GPU. I saw the display driver model using nvidia-smi and it was TCC. I changed it to WDDM for all the cards. But, on rebooting and running nvidia-smi -q again, it gives an error saying failed to allocate resources. To reconfirm that driver model has been changed, I saw the registry value (location mentioned by TMurray in some other post) and it mentioned adaptertype = 0. When I open nvidia control panel, it is still giving the same error.

All the Graphics applications are trying to use onboard Graphics. Is there any way this problem can be resolved?

Thanks,
Gaurav

PS: Windows device manager detects all three tesla cards. GPU-Z shows all 3 cards as well but, some values like memory etc. are not coming correct.

Looks like a PCI allocation problem. How many quardro cards were installed before you changed them to the Teslas? Are the Teslas on the same slots as the old quadros?

It is not the same machine. These are pre-configured blades (with Tesla M2070Q) from HP.

Hi,
Is this any answer for this problem? I got the same problem with TESLA M2070Q but using IBM Blade HS22.
Please help…

I just disabled the onboard VGA. This is a constraint in Windows 7 (server 2008 R2). All the GPUs available in system should support same display driver model i.e. XPDM on XP machines and WDM1.1 on Windows 7 machines. If the driver for on-chip graphics doesn’t support WDDM1.1 model, you won’t be able to work with Tesla cards.