ubuntu 14.04.2 - what nvidia drivers for CUDA

Question

after installing cuda toolkit and cuda samples via runfile installation (nvidia driver installed separately - NVIDIA binary driver - version 340.76 from nvidia-340 (open source)))

after running ./deviceQuery i have this:

./deviceQuery Starting...

CUDA Device Query (Runtime API) version (CUDART static linking)

cudaGetDeviceCount returned 30
-> unknown error
Result = FAIL

nvidia-smi gives me this:

+------------------------------------------------------+                       
| NVIDIA-SMI 340.76     Driver Version: 340.76         |                       
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 780 Ti  Off  | 0000:01:00.0     N/A |                  N/A |
| 37%   35C    P8    N/A /  N/A |    287MiB /  3071MiB |     N/A      Default |
+-------------------------------+----------------------+----------------------+

So the question is which nvidia driver will satisfy CUDA set?

As far as I know there is no xordg/edgers nvidia drivers for ubuntu 14.04.2 ?

?

Solution

I use Nvidia drivers from nvidia-331 package from official Ubuntu repository on Ubuntu 14.04.2.

Recently something changed in drivers packages and CUDA stopped working, the symptom was error 30 ("Unknown error") in first CUDA-related operation.

After some debugging I investigated that problem was with missing nvidia_uvm kernel module. Check that it is loaded:

$ lsmod | grep nvidia
nvidia_uvm             34855  0 
nvidia              10744943  84 nvidia_uvm
drm                   303102  2 nvidia

If it's not loaded, check that you have installed appropriate package, for 331 version it is nvidia-331-uvm:

sudo apt-get install nvidia-331-uvm

But this is not enough: nvidia_uvm can't be loaded in straightforward way (e.g. by running modprobe nvidia_uvm), you need to either run you program as root (once, then module will be loaded and root permissions will no longer be required), or (recommended) install nvidia-modprobe package and reboot:

sudo apt-get install nvidia-modprobe

Comments

[Mitch]

Comments are not for extended discussion; this conversation has been moved to chat.