A distilled version of this has been added to the Wiki, so you may want to just go there.
http://knoppmythwiki.org/index.php?page=UpdateNvidiaDrivers



Since the kernel source wasn't included with R5E50 by default, it has been a bit difficult to install any nvidia drivers other than 8776 and 7184. Thanks to the post at http://mysettopbox.tv/phpBB2/viewtopic.php?t=13281 we now know how to get the source. However, that's not the complete picture on how to get the latest and greatest drivers installed.

First, of course, follow the directions in the above post so that you have the /usr/src/linux-source-2.6.18-chw-13 directory. Now perform the following:


cd /usr/src/linux-headers-2.6.18-chw-13
cp -r . ../linux-source-2.6.18-chw-13


This merges the kernel-headers with the kernel-source, since the source by itself doesn't have the configuration files that were used.

Next, on my system at least, I had to at least partially make the kernel again, because certain kernel module scripts weren't in a usable state until you did that. So I did the following:


cd /usr/src/linux-source-2.6.18-chw-13
make
[wait about 20 seconds and then hit ctrl-c]


Now we're ready to install the new driver. The easiest way to do this is the rename the startx script and reboot so that X is not even trying to boot up.

mv /usr/bin/startx /usr/bin/startx.save
reboot


For some reason the NVidia install script isn't yanking out the original nvidia kernel module, so I manually pulled it out with this command:

rmmod nvidia
lsmod | grep nvidia

The second command will let you verify that you no longer have the module loaded.

Now we're actually ready to run the binary package from NVidia. It doesn't quite detect everything correctly, so issue the command as such:

sh NVIDIA-Linux-x86-1.0-9746-pkg1.run \
--kernel-source-path=/usr/src/linux-source-2.6.18-chw-13


You will most likely get a message during the installation process that it had to make assumptions about the paths for X.org. The assumptions that it made are fine.

Once the installer is finished you should be able to safely start X back up, but don't forget to rename your startx executable.

mv /usr/bin/startx.save /usr/bin/startx


As always, check the /var/log/Xorg.0.log file for hints as to what the problem may be. That's how I figured out that the original kernel module was still loaded!

Scott Milliken[/url]

Thanks for this, i'm trying it right now. Is there an easy way to overwrite all the files when doing the cp -r? there's alot of files to hit y, enter, y, enter for.

Edit --
Found my answer, if you do \cp -r you don't have to hit y constantly

after running these instructions, X fails to start and I have the following messages in dmesg:

NVRM: API mismatch: the client has the version 1.0-9746, but
NVRM: this kernel module has the version 1.0-8776. Please
NVRM: make sure that this kernel module and all NVIDIA driver
NVRM: components have the same version.

I reinstalled fresh and followed the instructions again and got the same problem. Any ideas?

Wow, I managed to really break my system. The second reboot after trying this, myth backand and frontend both seg fault when trying to start. Any idea what could cause that?

If you type:


It will remove the need for you to verify overwrite (for the current session only). Once you log out and back in the alias will be back in place.

Scott

Before you try to install 9746, verify that the old nvidia module was removed. I got this, too, and had to manually remove the old one first with:



Scott

I made sure I did rmmod nvidia before hand. Should I do rmmod nvram as well (I think I tried that already but I don't remember anymore). Is there anything else I need to remove before running the install script?
Thanks for the help so far.

Eek! No, nvram is the Non-Volatile RAM module and has nothing to do with nvidia.

You have to get the system to the point where X has not even attempted to start (by renaming startx as a suggested way), the nvidia module has been removed (via rmmod) and the NVIDIA installer script can find the kernel source (via the --kernel-source-path parameter).

Scott

Ok, tried this about 4 times now with the same results as I had before. Any other ideas as to what could be the problem?

Here's the last chunk of my Xorg.0.log:

Ok, I think I've made some progress...

There was an install-nvidia-debian being called from /etc/init.d/ that would automatically overwrite my installation of the latest nvidia drivers.

Bad news is that now I'm just getting a crazy yellow screen. But progress is progress.

Edit:
OK, before installing the drivers, I ran

update-rc.d -f install-nvidia-debian remove
update-rc.d -f nvidia-glx remove
update-rc.d -f nvidia-kernel remove

After doing that, I installed the driver and now it looks like it works!

Thanks for the help

I am trying to do this and get the following:

sh: NVIDIA-Linux-x86-1.0-9746-pkg1.run: No such file or directory

Any thoughts?

you need to download the drivers from nVidia's website first.

Run wget http://us.download.nvidia.com/XFree86/L ... 6-pkg1.run

Thanks for writing that up. I'm a bit new at all this and it solved a huge problem for me (no support for my card)

_________________
Skill level: 10^-3

so im following all this but i get errors that the NVIDIA script cant figure out the source version still i followed all directions on the post.

Did you first download and extract the kernel source? Can you post the nvidia installer log file?