Our lab use tesla 1070. At first we try to connect it to ASUS P5N-D with a NVidia NF750I chipset and a Core2 Duo CPU (E8400), It only recognize one GPU. Then we use EVGA 780i SLI (with NVidia chipset 780i), this time it finds the Tesla GPUs (i.e. /dev/nvidia0-4, depending on how many of the Tesla PCI-e interfaces I use), but it always returns an error (“can’t open /dev/nvidia2”) for every program in the CUDA_SDK (e.g. ‘deviceQuery’).
Any help will be appreciate.