First, I have built Fedora7 environment on my research desktop that has Geforce8800GT for displayer. Then, I have installed NVIDIA-Linux-x86-171.05 driver after connecting S870 to the desktop using the two special cables. Next, I have followed the release note to set up the CUDA Development Environment. Finally, after succeeding ‘make’ process in the projects directory, I could run some example binary files. Then, I could see the following error message when I ran matrixMul file.
[root@CUDA release]# ./matrixMul
NVIDIA: could not open the device file /dev/nvidia2 (Input/output error).
There is no device supporting CUDA.
If someone knows this problem, please let me know the solution.
Check with “lspci |grep 3D” if all the Tesla cards are recognized by the system.
If this is a 32bit system, you may need to pass some parameters during boot (see the readme).
It works on CUDA 1.0 with Nvidia drivers 100.14.11 (Quadro NVS 280) but don’t work on CUDA 1.1 with the same drivers! I think we must install the 169.09 (or later) NVidia driver…