Ask Your Question

Install nvidia driver for CUDA without turn it on for XWindows

asked 2015-11-04 19:04:20 -0500

hackingbear gravatar image

Hi, I put an nVidia Quadro K1200 (without attaching to a display) into my desktop which also has a built-in Intel i965 graphic and is running Fedora 22. I tried the following

1) install the nVidia driver using the nVidia installer without overriding X Windows configuration. After installation and eboot, I can verify that the nvidia is loaded with modprobe. But X Windows no longer start (on the Intel graphic.)

2) try to use the rpmfusion package as described in Same problem: X Windows fails to start afterward. This is the case after I remove the nouveau driver, nvidia X Windows modules/shared libs under /usr/lib64/xorg/modules or /usr/lib64/nvidia/xorg.

From the log, the start up failed because it couldn't load GLX due to undefined symbol: _glapi_tls_Dispatch. I checked the Xorg.log that nothing related to nvidia is mentioned anymore, but still something is lingering around.

How do I find/disable the nVidia GLX that's still haunting around in my system? Or any other proper way to install nVidia for CUDA without impacting X Windows? (I can reinstall the OS if need to.)

edit retag flag offensive close merge delete

2 Answers

Sort by ยป oldest newest most voted

answered 2015-11-19 23:48:48 -0500

hackingbear gravatar image

I got it work by

  1. make sure the nouveau driver is disabled in both X Window and kernel module
  2. download the cuda toolkit run file installer (not the nvidia driver by itself and not the rpm package.)
  3. run the cuda installer (with --override option for the newer gcc version)
  4. answer yes to the driver installation
  5. answer NO to the OpenGL installation
  6. answer yes to the cude toolkit installation
  7. answer NO to the sample insallation (as it would require
  8. comment out the GNUC version check error pragma in $CUDA_HOME/include/host_config.h

then X window will still work with the Intel graphic while the nVidia is dedicated to GPU computing.

edit flag offensive delete link more

answered 2015-11-05 10:38:01 -0500

kimbn gravatar image

I have not tried CUDA yet.

edit flag offensive delete link more

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Question Tools


Asked: 2015-11-04 19:04:20 -0500

Seen: 915 times

Last updated: Nov 04 '15