No GUI - Ubuntu 16.04 (4.4.0-119-generic) , Quadro K5200, Tesla K40c, Driver Version: 387.26

Accelerated Computing CUDA CUDA Setup and Installation
1

Hi
I have a Ubuntu 16.04 machine with 2 Nvidia cards a Quadro K5200 and a Tesla K40c . I have setup CUDA 9.1 and it installed from ( http://developer.download.nvidia.com/compute/cuda/9.1/secure/Prod/local_installers/cuda_9.1.85_387.26_linux.run )
The CUDA programs are working ( deviceQuery, nvidia-smi ). But the GUI display is not coming up . I switched from lightdm to gdm3 and this is the error I see ( /var/lib/gdm3/.local/share/xorg/Xorg.0.log )

Screen(s) found, but none have a usable configuration

This is the various outputs of commands
lspci |grep -i vga
04:00.0 VGA compatible controller: NVIDIA Corporation GK110GL [Quadro K5200] (rev a1)

lsmod |grep -i nv
nvidia_drm 49152 0
nvidia_modeset 901120 1 nvidia_drm
nvidia 13918208 1 nvidia_modeset
drm_kms_helper 155648 1 nvidia_drm
drm 364544 3 drm_kms_helper,nvidia_drm

The nouveau drivers have been disabled
cat /etc/default/grub |grep -i nou
lsmod |grep -i nou

Running nvidia-xconfig --preserve-busid --enable-all-gpus --allow-empty-initial-configuration

[b]Using X configuration file: “/etc/X11/xorg.conf”.

WARNING: Unable to use the nvidia-cfg library to query NVIDIA hardware.

ERROR: Unable to determine number of GPUs in system; cannot honor ‘–enable-all-gpus’ option[/b].

any help appreciated in getting the GUI up and running

2

Probably xorg.conf congfiguration is not matching the fact that you want your GUI to be running on the K5200, and the --enable-all-gpus switch is not helping (why don’t you try that command without that switch? That is what the error message is indicating)

This may help you get things sorted:

[url]USING CUDA AND X | NVIDIA

3

Thanks @txbob . But that did not help . I added a K5200 as follows
Section “Device”
Identifier “Device0”
Driver “nvidia”
VendorName “NVIDIA Corporation”
BoardName “Quadro K5200”
EndSection

but now I get an error
[ 907.819] (EE) NVIDIA(GPU-0): Failed to initialize the NVIDIA GPU at PCI:4:0:0. Please
[ 907.819] (EE) NVIDIA(GPU-0): check your system’s kernel log for additional error
[ 907.819] (EE) NVIDIA(GPU-0): messages and refer to Chapter 8: Common Problems in the
[ 907.819] (EE) NVIDIA(GPU-0): README for additional information.
[ 907.819] (EE) NVIDIA(GPU-0): Failed to initialize the NVIDIA graphics device!
[ 907.819] (EE) NVIDIA(0): Failing initialization of X screen 0

Doing lspci | grep nvidia seems to indicate that the device is correct

03:00.0 3D controller: NVIDIA Corporation GK110BGL [Tesla K40c] (rev a1)
04:00.0 VGA compatible controller: NVIDIA Corporation GK110GL [Quadro K5200] (rev a1)
04:00.1 Audio device: NVIDIA Corporation GK110 HDMI Audio (rev a1)

4

what is the result of running

sudo dmesg |grep NVRM

?

5

[ 191.640837] NVRM: RmInitAdapter failed! (0x30:0xffff:662)
[ 191.640870] NVRM: rm_init_adapter failed for device bearing minor number 1
[ 192.070408] NVRM: failed to copy vbios to system memory.

Repeated infinitely

6

I find it hard to believe that deviceQuery is working correctly with those messages in your system message log.

In this state, deviceQuery properly reports both GPUs ?

what kernel version?

uname -a

7

Sorry I should have caught this before . deviceQuery reports only 1 GPU . The lspci reported the Quadro K5200 . This is the output of the uname -a

4.4.0-119-generic #143-Ubuntu SMP Mon Apr 2 16:08:24 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux

The strange thing is that I remove the nvidia drivers and the monitor and GUI works . (with lightdm ). so I never suspected a problem in the hardware

8

which GPU is reported? The K40c or the K5200?

sounds like you haven’t properly removed the nouveau driver

http://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#runfile-nouveau

switch to runlevel 3 (or stop the gdm/GUI) before following those steps
then reboot

you can verify that nouveau is properly removed with a command like this:

lsmod |grep -i nouv

which should return empty

9

The K40c is what is reported by the deviceQuery .

I have one observation ( not sure if that is the cause of the problem )

The K5200 has 2 display output and I had connected the monitor to one of them . I decided to switch the monitor ports and suddenly I am not able to do Ctrl+Alt+F1 etc. and the font sizes even on the console mode have gone to 640x480

I am not sure if that has got any bearing on this at all ?

10

This is all sounding like a problem with nouveau, i.e. you haven’t properly removed the nouveau driver.

11

Thanks for the help so far . But I think I have followed everything and the nouveau drivers are not present . and I have no idea .

I will may be leave at this point and contine to use the CUDA stuff ( since that was the main option ) .

Will debug this later .

But thanks again for the help . Much appreciated

12

I have the same issue.

$ sudo nvidia-xconfig --enable-all-gpus
[sudo] password for alex: 

Using X configuration file: "/etc/X11/xorg.conf".

ERROR: Unable to determine number of GPUs in system; cannot honor '--enable-all-gpus' option.

Backed up file '/etc/X11/xorg.conf' as '/etc/X11/xorg.conf.backup'
New X configuration file written to '/etc/X11/xorg.conf'
$ lsmod |grep -i nouv
$ uname -a
Linux alex-XPS-15-9570 5.0.0-36-generic #39~18.04.1-Ubuntu SMP Tue Nov 12 11:09:50 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux
$ kf5-config --version                                                                                                                                                                  
Qt: 5.9.5                                                                                                                                                                                                                             
KDE Frameworks: 5.44.0                                                                                                                                                                                                                
kf5-config: 1.0