I’m trying to set up a test system to use a basic PCI video card for system video output and a Tesla card exclusively for debugging/profiling CUDA code. Is there a way to force Windows XP to use different drivers simultaneously for the two cards? The PCI card is a GeForce 8400GS and the Tesla is C2050.
So far, I managed only to use the PCI card for display with its driver installed and Tesla happily ignored. After the developer driver has been installed the system refused to boot up.
If someone has done this sort of setup, I would very much appreciate any suggestions.
Thanks in advance
You would have better luck with Win 7.
Actually, I tried Win7 first, and it wouldn’t even boot the system with the PCI card installed and Tesla in (no developer driver).
I ended up reading posts about WDDM vs. the older XPDM display model and how Win7 and Vista wouldn’t allow two cards with different drivers to even be used. Hence, the question about WinXP.
If someone has had specific experience setting this up, please let me know.
Have you tried changing the C2050 to only TCC mode?
no, I haven’t. From the Nvidia docs looks like it should be set…
Trying to figure out the sequence to get this working… If I start off with the basic PCI card and its driver installed, do I then follow up with installing the Tesla, CUDA toolkit and nvidia-smi setting change? Plus the switch to PCI/PEG in BIOS to get the video out of the PCI card. Then does it all influence the CUDA driver installation? Can the developer driver for the Tesla be installed to not interfere with the PCI card?