Axel Thimm wrote:
> But you still need to restart X to allow the installed nvidia driver
> to take over. You would even need to rmmod nvidia and modprobe it
> again otherwise you'd be running the old nvidia kernel module with the
> new nvidia X11 driver.


I am relatively inexperienced in GNU/Linux (recent full adopter), the upgrade/installation 
process I have been following is this:


1) Remove previous nvidia rpms if present.
2) Install the latest nvidia rpms
3) Reboot.
4) Run nvidia-settings-1.0-xxxx  (or just plain nvidia-settings).
5) Run nvidia-xconfig-1.0-xxxx
6) Reboot.


xxxx= version number




> So the full steps are:
> 
> o install the driver (e.g. via ATrpms' rpms or manually via the
>   installer, whatever suits you best)
> o exit X (go runlevel 3)
> o rmmod nvidia
> o modprobe nvidia
> o restart X (telinit 5)
> 
> It's easier to ask for a reboot, though, and it is not guaranteed that
> rmmod'ing an old nvidia kernel module will really restore the chip to
> what the next modprobe expects. nvidia recommends to power-cycle your
> system when upgrading the kernel module, but it usually works w/o.


I suppose this is the process to avoid reboots?