CUDA not work for graphic card without monitor

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi Experts
I have been working on this for two days, and now I think I really need some hints from you guys.

I got a Z68 mobo with i7-2600k and a GT430 graphic card. The way I wanted my new desktop to be is use onboard port for video output and the graphic dedicated for CUDA. But after I install the nvidia driver, the X just would not start if I use the onboard intel chipset.
Later I know it is because the nvidia installer changed the xorg.conf file, so I change it back like this

Section “Device”
Identifier “Device2”
Driver “intel”
BusID “PCI:00:02:0”
VendorName “Intel Corporation”
EndSection

also change the screen section to make it reference the onboard video card:

Section “Screen”
Identifier “Screen0”
Device “Device2”
Monitor “Monitor0”
DefaultDepth 24
SubSection “Display”
Depth 24
Modes “1360x768”
EndSubSection
EndSection

Then the onbarod viedo goes back to work again, but the discrete card just stop working, it gives nonsense when running test programs. Then I tried to add the nvidia config back to xorg like this

Section “Device”
Identifier “Devicer0”
Driver “nvidia”
BusID “PCI:01:00:0”
VendorName “Nvidia Corporation”
EndSection

But not referenced by any screen in screen section, still the discrete card does not work! What i am guessing is that when the card is not referenced by any screen, the X just would not activate it and load the module needed for it to work.

So, my question is what is the module needed to make the card work(it seems merely load the module “nvidia” not work), or how to activate a card without a monitor connected???

I know there is desktop in my lab where it use one nvidia card for output and another for cuda, and it works well. But right now I want my intel onboard video to do video output and nvidia card to do CUDA task, anybody have the same experience before? Any suggestion would really be appreciated!

2

Hi Experts
I have been working on this for two days, and now I think I really need some hints from you guys.

I got a Z68 mobo with i7-2600k and a GT430 graphic card. The way I wanted my new desktop to be is use onboard port for video output and the graphic dedicated for CUDA. But after I install the nvidia driver, the X just would not start if I use the onboard intel chipset.
Later I know it is because the nvidia installer changed the xorg.conf file, so I change it back like this

Section “Device”
Identifier “Device2”
Driver “intel”
BusID “PCI:00:02:0”
VendorName “Intel Corporation”
EndSection

also change the screen section to make it reference the onboard video card:

Section “Screen”
Identifier “Screen0”
Device “Device2”
Monitor “Monitor0”
DefaultDepth 24
SubSection “Display”
Depth 24
Modes “1360x768”
EndSubSection
EndSection

Then the onbarod viedo goes back to work again, but the discrete card just stop working, it gives nonsense when running test programs. Then I tried to add the nvidia config back to xorg like this

Section “Device”
Identifier “Devicer0”
Driver “nvidia”
BusID “PCI:01:00:0”
VendorName “Nvidia Corporation”
EndSection

But not referenced by any screen in screen section, still the discrete card does not work! What i am guessing is that when the card is not referenced by any screen, the X just would not activate it and load the module needed for it to work.

So, my question is what is the module needed to make the card work(it seems merely load the module “nvidia” not work), or how to activate a card without a monitor connected???

I know there is desktop in my lab where it use one nvidia card for output and another for cuda, and it works well. But right now I want my intel onboard video to do video output and nvidia card to do CUDA task, anybody have the same experience before? Any suggestion would really be appreciated!

3

You probably need to run a script to setup the /dev/nvidia*
That script can be found in the Linux release notes.

If that doesn’t work, then perhaps you’ll have to try Windows7.