Ubuntu 18.04 LTS install. No problem installing new OS with the default nouveau drivers. Desktop working fine. Installing all my normal configuration and utilities. Reboot system to add the graphics-drivers ppa. No problem and select the latest Nvidia 396.24 driver which I am currently using in my other Linux systems. Installs fine. Add more utilities and further set up normal configurations. Nvidia-settings added to the Dock. Everything looks good. Reboot system and nvidia drivers working fine. Now to get the 3 GTX 1070 cards set up for compute. Need to set cool-bits to get fan control and to set all cards up for compute. This is where everything goes to he!!. Run
nvidia-xconfig --enable-all-gpus
nvidia-xconfig --cool-bits=4
xorg.conf reconfigured and new xorg.conf.backup and xorg.conf.original-nvidia-configuration files created. Look at xorg.conf and see that cool-bits-4 has been added to all screens like usual. Everything looks normal and exactly what the xorg.conf looks like in my other Ubuntu 16.04 compute systems. Reboot the system for the new xorg.conf to take effect and to have fan control available in nvidia-settings app.
System boots and I see the normal boot display because I always remove "quiet splash" since I want to see any errors and watch the normal loading of the system. The boot gets to the end where the display manager blanks the screen and I get the normal purple screen. But instead of getting the mouse cursor, the screen stays blank. I finally move the mouse and the cursor is a big "cross" and that is it. I either have to drop to tty with a alt-ctrl-F3 or push the computer reset button. I can't restart the display manager with a gdm restart or systemctrl restart gdm.service does nothing. I can't get to a login screen.
The only way I can get the system back is to boot grub recovery mode, drop to root, mount the drive and cp xorg.conf.backup to xorg.conf and the system boots again to the desktop. I don't know why either of the normal nvidia-xconfig commands corrupt the xorg.conf or whatever and then prevents loading the desktop or display manager. I have tried both commands together and separately. Either will make a new xorg.conf that won't boot the desktop. The file looks absolutely correct and I even compared it to my other 3 gpu Ubuntu 16.04 system. The syntax and section structure is exactly the same.
I have never had any issues with Ubuntu 16.04 adding the use all gpus or fan control before. I normally use cool-bits=28 on the 16.04 machines but I tried several settings of the cool-bits enable parameters to try and see if the more elaborate settings was causing issues. cool-bits=4 is the minimum I need to get fan control setting available in nvidia-settings. I normally just set the gpu fan for 100% fan speed that way since the gpus are run full stop 24/7 with compute loads.
Can anyone offer a reason why nvidia-xconfig corrupts xorg.conf in Ubuntu 18.04 LTS while is no problem with Ubuntu 16.04? Help please!!
Bookmarks