My CUDA toolkit installation on Ubuntu platform has failed. I have a GTX960M installed Asus N551VW notebook PC Xubuntu 16.04 is the OS on it. I applied all the steps for a package manager installation, which is the suggested way of installation. No previous installation of CUDA exists by the way, therefore no conflicts should be bothering me. I followed the network package manager installation (deb [network]) method and the toolkit “seemed to” install free of any problem or error. Finally, I updated my PATH variable to include the /usr/local/cuda-9.0/bin directory.
However, when it comes to verification of the installation, everything is not so fine. The installation page states "When the driver is loaded, the driver version can be found by executing the command
$ cat /proc/driver/nvidia/version
"
which returns:
No such file or directory
nvcc -V returns the version of the toolkit as V9.0.176, and I can see that the system uses the GPU driver v384.81 as stated in Software & Updates > Additional Drivers dialog.
I issued sudo make command in sample codes home directory. After that ran the deviceQuery program under the 1_Utilities folder, which returned
CUDA driver version is insufficient for CUDA runtime version Result = FAIL
Further, nvidia-smi command does not work.
Any ideas? Thanks in advance.
BTW: There has been some kind of a flaw during driver installation, which is automatically issued during toolkit installation. There was a step where I defined a pwd for UEFI secure boot disabling. It said I should disable the secure boot option for driver to work, and will be asked this pwd upon reboot. I wasn’t. So, I manually disabled secure boot and set a new User pwd on the BIOS and booted with those options, but that made no difference.