i have some questions about how to install and configurate multi-gpu for running CUDA programs, hope someone who knows or has experience can give me some advices.
1 i have a workstation, HP xw 8600, which has two nvidia GPU, one is quadro FX 370 and the other is Tesla C1060, and this machine has been tested for compatibility with the Tesla products by NVIDIA Corp. OS is Microsoft Windows XP Professional 64 Edition. on XP 64, i inserted the C1060, installed the drive\tooklit\SDK, then every project under SDK dir can be correctly compiled and runed now. before that, FX 370 had been there and worked well. now i want to replace it by the newest redhat enterprise linux. since there are two different gpu, i am not sure about what i should to do with the drive, just install the newest drive once or install each respectively?
2 i diyed a new PC yesterday, using a asus motherboard with gpu\network card\audio card integrated, i inserted a C1060 on the PCI-E 2.0 X16 slot. because there is no any video ouput on the display, so BIOS was reseted to disable video output from PCI-E, then fedora 10 was installed, video output on display had disappeared several times when i activated network card, then everything is normal. when i finally installed the driver for C1060, then the driver-installer told me there wasn’t a gpu can couldn’t continue. i guess this is something wrong with BIOS set or driver installation problem, or motherboard’s compatibility with Tesla, does anyone has such a diy experience or know how to figure my problem out?
Note: driver of the motherboard provided is for vista, so i am finding driver for linux.
3 maybe i should not do like this, it may be right to put two C1060 in HP workstation and to buy a geforce seriese gpu for the diyed PC. if i do, can i just install the driver once for two C1060?
any advice is appreciated