[sclug] Nvidia and Ubuntu.

Pieter Claassen pieter at claassen.co.uk
Sun May 21 12:09:41 UTC 2006

On Sunday 21 May 2006 12:58, ed wrote:

> There is every possibility that I was doing things wrong, but last time
> I tried, no matter how much mucking around I could not get dual head to
> work.

I have the below packages installed which gives me fine dual head and glx 
support (RMS and other philosophical concerns aside).

A few things I picked up:
- You will notice that I also had a third card in my PC (radeon) which slowed 
things down so badly in xinerama that I had to drop using it. Just the two 
heads on the single GeForce FX 5200 work fine. Without cinerama three 
xservers on three heads worked fine.
- I had to hack the xorg.conf file by hand because debconf didn't handle dual 
heads (might have changed by now).
- The nvidia-settings packages had some nvidia tool that can do the dual head 
config but I found that it didn't work for me (can't remember why but you 
should try it. It did uninstall nvidial-glx which might be a dapper 
dependency thing. It actually recommends it so it shouldn't uninstall it).
- Do not forget "sudo nvidia-glx-config enable" if things don't want to work 
for no reason at all. Don't know what it does but there was a hoary/breezy 
ugpade bug that left glx disabled unless you ran this after the upgrade)
- I  don't know why I have DRI switched off?
- I strongly recommend against compiling your own kernel module because unless 
you use the apt build system (which means you have to either create a new 
custom package yourself or use the glx source package which would have anyhow 
given you what is compiled in the restricted modules package), you will not 
get a .deb and everytime your kernel is upgraded, you have to rebuild the 
package (in the dark with only the CLI flashlight so to speak) which is cool 
for techies but not your grandpa.
- I did once get the nvidia binary going but that required the flashlight 
modules recompile when your kernel upgraded.
- This config worked on breezy and now on dapper.
- The actual nvidia kernel module is in the linux-restricted-modules.
- The "screen" settings were important for a single card with dual heads 
because the PCI bus identifier was the same for both cards. Multiple cards 
seem to not care about it. This is an internal driver issue I think with hand 
waving logic to its configuration i.e. the logic of the action of the driver 
is not inherently apparent unless you have an IQ of 120 (the average I think 
is 110).


ii  nvidia-glx                             1.0.8756+                  
NVIDIA binary XFree86 4.x/X.Org driver
ii  nvidia-kernel-common                   20051028+1                            
NVIDIA binary kernel module common files
rc  nvidia-settings                        1.0-3ubuntu7                          
Tool of configuring the NVIDIA graphics driv
ii  linux-restricted-modules-2.6.15-22-386                           
Non-free Linux 2.6.15 modules on 386

Relevant sections in xorg.conf are:

Section "ServerLayout"
        Identifier      "Small Layout"
        Screen          0 "1" LeftOf "2"
        Screen          1 "2" RightOf "1"
#       Screen          1 "3" RightOf "2"
        #Screen                 2 "2"
        #Screen         "2" 0 0
        InputDevice     "Generic Keyboard"
        InputDevice     "Configured Mouse"

Section "ServerFlags"
        Option          "Xinerama"

#Section "DRI"
#       Mode    0666

Section "Device"
        Identifier      "NVIDIA Corporation NV34 [GeForce FX 5200]"
        Driver          "nvidia"
        BusID           "PCI:1:0:0"
        Screen          0
Section "Device"
        Identifier      "rad"
        Driver          "radeon"
        BusID           "PCI:0:10:0"
#       Screen          2

Section "Device"
        Identifier      "DVI"
        Driver          "nvidia"
        BusID           "PCI:1:0:0"
        Screen          1

More information about the Sclug mailing list