X Config file - Not running device driver
We were given a blade-style box running the Tesla M2050 with OS - Centos Final 5, which is supposed to be compatible with Red Hat 5. The updated driver for the Tesla 2050 was installed and re-installed (V290.10) upon reboot, I receive an error with the X configuration file that I am not running NVIDIA device driver. Please edit your X configuration file (just run 'nvidia-xconfig' as root) and restart the X server. I did this multiple times but it does not fix the problem. As a result, (I surmise that) I am unable to finish the install of Visual Profiler even thought the deviceQuery and bandwidthTest both pass and show the Tesla M2050 running.
We were given a blade-style box running the Tesla M2050 with OS - Centos Final 5, which is supposed to be compatible with Red Hat 5. The updated driver for the Tesla 2050 was installed and re-installed (V290.10) upon reboot, I receive an error with the X configuration file that I am not running NVIDIA device driver. Please edit your X configuration file (just run 'nvidia-xconfig' as root) and restart the X server. I did this multiple times but it does not fix the problem. As a result, (I surmise that) I am unable to finish the install of Visual Profiler even thought the deviceQuery and bandwidthTest both pass and show the Tesla M2050 running.

#1
Posted 01/09/2012 09:08 PM   
The visual profiler does not need to be installed separately, it is part of the toolkit.

If you have a server with on-board VGA and plan to use the Tesla for compute only, assuming you have a working X11 environment for the on-board card:
1) Install the nvidia-driver without configuring X11
2) Use the script in the release notes to load the nvidia module and create the /dev/nvidia entries
3) Enable persistent mode: nvidia-smi -pm 1 (this is not required but it will speed up the startup time of CUDA program)
The visual profiler does not need to be installed separately, it is part of the toolkit.



If you have a server with on-board VGA and plan to use the Tesla for compute only, assuming you have a working X11 environment for the on-board card:

1) Install the nvidia-driver without configuring X11

2) Use the script in the release notes to load the nvidia module and create the /dev/nvidia entries

3) Enable persistent mode: nvidia-smi -pm 1 (this is not required but it will speed up the startup time of CUDA program)

#2
Posted 01/09/2012 10:13 PM   
[quote name='mfatica' date='09 January 2012 - 03:13 PM' timestamp='1326147239' post='1353845']
The visual profiler does not need to be installed separately, it is part of the toolkit.

If you have a server with on-board VGA and plan to use the Tesla for compute only, assuming you have a working X11 environment for the on-board card:
1) Install the nvidia-driver without configuring X11
2) Use the script in the release notes to load the nvidia module and create the /dev/nvidia entries
3) Enable persistent mode: nvidia-smi -pm 1 (this is not required but it will speed up the startup time of CUDA program)
[/quote]
[quote name='mfatica' date='09 January 2012 - 03:13 PM' timestamp='1326147239' post='1353845']

The visual profiler does not need to be installed separately, it is part of the toolkit.



If you have a server with on-board VGA and plan to use the Tesla for compute only, assuming you have a working X11 environment for the on-board card:

1) Install the nvidia-driver without configuring X11

2) Use the script in the release notes to load the nvidia module and create the /dev/nvidia entries

3) Enable persistent mode: nvidia-smi -pm 1 (this is not required but it will speed up the startup time of CUDA program)

#3
Posted 01/10/2012 03:55 AM   
[quote name='mfatica' date='09 January 2012 - 03:13 PM' timestamp='1326147239' post='1353845']
The visual profiler does not need to be installed separately, it is part of the toolkit.

If you have a server with on-board VGA and plan to use the Tesla for compute only, assuming you have a working X11 environment for the on-board card:
1) Install the nvidia-driver without configuring X11
2) Use the script in the release notes to load the nvidia module and create the /dev/nvidia entries
3) Enable persistent mode: nvidia-smi -pm 1 (this is not required but it will speed up the startup time of CUDA program)
[/quote]
[quote name='mfatica' date='09 January 2012 - 03:13 PM' timestamp='1326147239' post='1353845']

The visual profiler does not need to be installed separately, it is part of the toolkit.



If you have a server with on-board VGA and plan to use the Tesla for compute only, assuming you have a working X11 environment for the on-board card:

1) Install the nvidia-driver without configuring X11

2) Use the script in the release notes to load the nvidia module and create the /dev/nvidia entries

3) Enable persistent mode: nvidia-smi -pm 1 (this is not required but it will speed up the startup time of CUDA program)

#4
Posted 01/28/2012 04:36 AM   
I still cannot get nvidia-settings to see the Tesla. The nvidia-xconfig still wants to set the vga driver from "mga" to "nvidia".

Installed Ubuntu 10.04 and followed install instructions to the detail, including blacklisting, etc. DevDriver 285.05.32 installed fine as did Cuda and SDK. I set my path dependencies. nvcc -V replied, "NVIDIA (R) Cuda compiler driver, Cuda compilation tools release 4.1, V0.2.1221.

nvidia-settings still gives "You do not appear to be using the NVIDIA X driver. Please edit your x config file (just run 'nvidia-xconfig' as root) and restart the X server. When I run nvidia-xconfig - it changes the onboard vga driver from "mga" (Matrox) to "nvidia" and the display does not work. I need to edit the xconfig.org via terminal and reboot which fixes the screen but not the nvidia-settings issue.

Now nvcc -V gives, "No command found, did you mean: 'nvlc', driver 285.05.32 is running fine, nvida-settings gave me a GUI screen with "You do not appear to be using the NVIDIA X driver...

Was able to compile SDK and ran deviceQuery. Failed. No CUDA-capable device is detected. (not surprised)



Your help is GREATLY appreciated!
I still cannot get nvidia-settings to see the Tesla. The nvidia-xconfig still wants to set the vga driver from "mga" to "nvidia".



Installed Ubuntu 10.04 and followed install instructions to the detail, including blacklisting, etc. DevDriver 285.05.32 installed fine as did Cuda and SDK. I set my path dependencies. nvcc -V replied, "NVIDIA (R) Cuda compiler driver, Cuda compilation tools release 4.1, V0.2.1221.



nvidia-settings still gives "You do not appear to be using the NVIDIA X driver. Please edit your x config file (just run 'nvidia-xconfig' as root) and restart the X server. When I run nvidia-xconfig - it changes the onboard vga driver from "mga" (Matrox) to "nvidia" and the display does not work. I need to edit the xconfig.org via terminal and reboot which fixes the screen but not the nvidia-settings issue.



Now nvcc -V gives, "No command found, did you mean: 'nvlc', driver 285.05.32 is running fine, nvida-settings gave me a GUI screen with "You do not appear to be using the NVIDIA X driver...



Was able to compile SDK and ran deviceQuery. Failed. No CUDA-capable device is detected. (not surprised)







Your help is GREATLY appreciated!

#5
Posted 01/28/2012 05:19 AM   
Resolved!

nvidia-settings program is used to configure X to use the nvidia card to drive graphics. The nvidia card (Tesla family: M2050) is for general purpose calculations only. So the error message is ok. Running from a tty - everything works well. For the graphical profiler, computeprof that comes with the CUDA toolkit works fine as well. Running Cuda Toolkit V.1.
Resolved!



nvidia-settings program is used to configure X to use the nvidia card to drive graphics. The nvidia card (Tesla family: M2050) is for general purpose calculations only. So the error message is ok. Running from a tty - everything works well. For the graphical profiler, computeprof that comes with the CUDA toolkit works fine as well. Running Cuda Toolkit V.1.

#6
Posted 02/01/2012 05:04 PM   
Scroll To Top