CUDA Card compatibility two NVIDA cards doing separate duties.

Accelerated Computing CUDA CUDA Programming and Performance
1

One of my colleagues at work installed a NVIDIA 275 video card and is using it only for GPU parallel processing. At least I thought he was doing just that. He had removed the original graphics card which was also an NVIDIA card, but was not CUDA compatible. That was okay since this card would never be used with CUDA only the doing traditional graphics duty on this quad processor machine.

As of now this NVIDIA 275 CUDA compatible card is doing GPU and computer graphics. I only want it to do GPU parallel processing - nothing more.

The question is this: how does one use two NVIDIA graphic cards one of which is CUDA compatible and one which is not. I want the NVIDIA CUDA compatible card to do GPU parallel processing only, and the older non CUDA compatible card to do the old graphics duties, only.

Newport_j

2

Whether it is possible or not really depends on how old the second card is. If it is older than a GeForce 6 series, then the short answer is that you can’t, because only a single driver can be loaded, and that is the oldest card which the current 180.x and 190.x series will support (and which also support your GTX 275).

If the card is a Geforce 6 or 7 series (and based on your earlier postings I don’t think it is), then you only need to make some configuration changes in the xorg.conf file to specify the display card by PCI address. I have a box with a pair of GTX 275s installed:

avid@cuda:~$ lspci | grep nVidia

01:00.0 VGA compatible controller: nVidia Corporation Device 05e6 (rev a1)

04:00.0 VGA compatible controller: nVidia Corporation Device 05e6 (rev a1)

and use a Device and Screen configuration like this, so that a display runs on PCI:1:0:0, and not on PCI:4:0:0 :

Section "Device"

	Identifier	 "Device0"

	Driver		 "nvidia"

	VendorName	 "NVIDIA Corporation"

	BoardName	  "GeForce GTX 275"

	BusID		  "PCI:1:0:0"

	Option		 "Coolbits" "1"

EndSection

Section "Screen"

	Identifier	 "Screen0"

	Device		 "Device0"

	Monitor		"Monitor0"

	DefaultDepth	24

	SubSection	 "Display"

		Depth	   24

		Modes	  "1440x900"

	EndSubSection

EndSection

Nothing else is required for Xorg. With only a single CUDA capable card, you don’t even need to worry about setting compute exclusivity or anything like that. It should just work.