NVIDIA: could not open the device file /dev/nvidia0 (input/output error) Error obtained when running

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi all,

I am having trouble running the CUDA SDK samples. when running nbody i get the following result:

[codebox]adam@Deimos2U:~/NVIDIA_CUDA_SDK/bin/linux/release$ ./nbody

Run “nbody -benchmark -n=” to measure perfomance.

NVIDIA: could not open the device file /dev/nvidia0 (Input/output error).

cudaSafeCall() Runtime API error in file <nbody.cpp>, line 553 : initialization error.[/codebox]

This happens when I run any SDK code. I have a dell studio xps 13 laptop, with the hybrid sli vid cards (9200G+9400GS). Is CUDA trying to use one of the two vid cards that I do not currently have enabled?

How do I get around this?

I also just ran the sample cuda code found at http://llpanorama.wordpress.com/2008/05/21…t-cuda-program/

and get the following output:

[codebox]adam@Deimos2U:~/Documents/CUDA$ nvcc cuda_test.cu -run

NVIDIA: could not open the device file /dev/nvidia0 (Input/output error).

0 0.000000

1 1.000000

2 2.000000

3 3.000000

4 4.000000

5 5.000000

6 6.000000

7 7.000000

8 8.000000

9 9.000000

[/codebox]

Does any one have any idea whats going on here? I tried doing a search in this forum but came up with nothing.

THanks all,

Adam

2

Check the permissions on the /dev/video device, and give yourself group membership to whatever group owns it. If you don’t have it, then you need to make sure that the driver module is actually loaded and you have some ruleset active to create the devices and set permissions at boot time. What distro are you running?

3

Did you start X with the nvidia driver?

4

Yes, I have the driver working, get the NVIDIA logo before logon and have compiz working as well (my usual indicators - fyi i have compiz off while doing cuda work though).

5

I’d like to see an nvidia-bug-report.log generated while this problem is present.

6

I checked the permissions, both are set to root root. I should mention that on my desktop (8800GTS), I have the exact same settings/distro/everything (and also have /dev/nvidia0 as root root), and all CUDA code runs great.

The driver module is actually loaded on the laptop, and do I realy need to have a ruleset active to set my permissions at boot, given that I do not have this on the desktop and it works fine?

I have been suspecting that it is the Hybrid SLI that is confusing the computer. /dev/nvidia0 may be the one I do not have activated in my xorg.conf files.

I am running Ubuntu 8.10 x86_64, with the 181.22 drivers.

Thanks again,

adam

7

I am suffering from the same issue on the same Laptop Dell Studio XPS 13
@laxsu19 : did u get any solution for your issue?

8

Nothing past what is discussed here: [topic=“96451”]Topic 96451[/topic]

But that basically set it up so CUDA would work - but only on the integrated GPU - never got access to the discrete card on our laptop. Its pretty frustrating - It was one of the reasons why I got this laptop - ability to do useful things on CUDA and get good overall performance otherwise while still having decent battery life.

Any plans to incorporate this, NVIDIA?

9

Hi I am running into the exact same problem one year later on Ubuntu 10.04 LTS. Is it officially impossible to use CUDA on a Dell Studio 13?

Many thanks,

Frédéric

10

Hi I am running into the exact same problem one year later on Ubuntu 10.04 LTS. Is it officially impossible to use CUDA on a Dell Studio 13?

Many thanks,

Frédéric