NVIDIA The Hard way Q's

I’ve read and re-read NVIDIA the hard way but before I do this I was hoping to get a few items cleared up for the text.

First: I have 11.2 installed and the latest reference is to 11.1. Can I assume comments about 11.1 pertain to 11.2?

Second: There’s a line about using Sax2 that states: “With 10.0 you won’t need to specify -m 0=nvidia any more, sax automatically detects and uses it if you activate 3D acceleration.”

I had to use the sax2 -r -m 0=nv to get my display to use the nv driver. Won’t I need to do the same for 11.2?

And how do I determine if 3D accel is activated? Is that a selection in a dialog that occurs during the sh NVIDIA-yada-yada.run -q execution or what?

Thanks – Mike

openSUSE Graphic Card Practical Theory Guide for Users - openSUSE Forums

11.1 or 11.2 the hard way install is basically the same.
Once you reboot after installing the driver you will see if all is well. If it’s the proprietary driver you may need to do sax2 -r -m 0=nvidia

Read the guide above.

Hi Mike,

>> Can I assume comments about 11.1 pertain to 11.2?

Yes.

>> I had to use the sax2 -r -m 0=nv to get my display to use the nv driver. Won’t I need to do the same for 11.2?

I’ve never had to use sax2 - the driver installer writes a new Xorg.conf for you. But even if I did use sax2, it would have been in automatic mode (“sax2 -a”). I know I’ve never explicitly which driver to use.

>> And how do I determine if 3D accel is activated?

I always use glxgears. I get ~2000fps on my 8000-series GeForce and ~10,000fps on my 9000-series GeForce. Without 3D, I’d expect to get something in the hundreds.

Also, an “lsmod | grep nv” should only return the nvidia driver and not the nv one.

But with all of that said, why do you want to manually install the Nvidia driver? The YaST method is fast and it works.

-Jeremy

I have a geforce2 ti AGP card and the display driver hasn’t hooked up correctly when installing anything I’ve tried… debian, mepis, ubuntu, mint and lastly openSUSE.
I became determined to figure out how to reliably fix it while on openSuse and posted here with the problem(display stuck at 800x600). With this forums’ posters help I’ve been able to get driver ‘nv’ installed and working. The guy that helped me suggested doing it “The hard way” (which isn’t really hard) whos directions are in one of the driver guide posts on this site. If a certain method works I’ll use it even if a little more typing is involved. Sure beats having to reinstall the OS.
Hey, I’m real new to linux, been at it now almost 3 weeks. Been a MS guy since the AT’s came out and now I want OUT!
I am all ears.

Isn’t that an old card?
I’m not sure of what support there is…

This card should work with the nVidia proprietary 96.43.16 driver as discussed in this thread: display stuck at 800x600? - openSUSE Forums.

You can find that driver here: NVIDIA DRIVERS 96.43.16

To install “the hardway” you need to do some things:

  1. first install gcc, make, linux-kernel-headers, and also kernel-source and kernel-syms where kernel-source and kernel-syms are the EXACT same version as your kernel. You can tell your kernel version by typing:
uname -a

You MUST install those applications first and what I noted about the EXACT same version number is important.
.

  1. download from the nVidia site the NVIDIA-Linux-x86-96.43.16-pkg1.run driver to your hard drive, … say to /home/your-user-name (where your-user-name is your user name).
    .
  2. reboot your PC, and at the initial boot/splash menu (where you can choose failsafe or a normal boot) press 3 with normal boot selected such that a “3” (no quotes) appears in the options line of that menu. Then press for a boot to a full screen text mode
    .
  3. login as a regular user. Then type “su” (no quotes and enter root password) to get root permissions.
    .
  4. backup/move your /etc/X11/xorg.conf file with:
mv /etc/X11/xorg.conf /etc/X11/xorg.conf.nv.backup

.

  1. build and install the proprietary nVidia driver with:
sh NVIDIA-Linux-x86-96.43.16-pkg1.run -q

.

  1. then build a custom xorg.conf file to work with that driver with:
nvidia-xconfig

.

  1. then restart with:
shutdown -r now

and test your graphics (don’t type “3” this time when rebooting).
IF the above does not work, you can boot again to run level 3 (as noted above by typing 3 at the splash/boot menu) and this time restore the original xorg.conf file with:

mv /etc/X11/xorg.conf.nv.backup /etc/X11/xorg.conf

and restart with

shutdown -r now

and you are back where you started with nothing lost (but some time) for a failed effort.

But the above should work.

oldcpu - you’re awesome!
I’ve printed all the links you’ve referenced in this and the other thread. I’ll be back after the dust settles.

thanks again.

One thing that still puzzles me is the comment about "“With 10.0 you won’t need to specify -m 0=nvidia any more, sax automatically detects and uses it if you activate 3D acceleration.”

How do I check to see if 3D accel is “activated”? Is that in a dialog or do I need to do some cat/ |grep thing?

I installed gcc, gnu make & kernel-source
then I rebooted into the terminal and su
I backed up my xorg.conf
I cd’d to /usr/src/linux
I executed – make cloneconfig
then – make prepare
cd’d to /home/[me]/Download
executed sh NVIDIA-Linux-x86-96.443.16-pkg1.run -q

I got an error unable to build kernel and see the
/usr/log/nvidia_installer.log file

As it was trying to compile I also noted a message that my GPU wasn’t supported by this libraary and to see nvidia.com

I rebooted and found my display back at 800x600 so I rebooted again nad copied my xorg.conf.nv.backup over xorg.conf and now my display is back where it should be.

I also tried the same thing with the NVIDIA-Linux-x86-71.86.13 driver wit the same results.

Maybe I need a different GPU

Post result of this

rpm --query --all '*kernel*'

no linux-kernel-headers? no kernel-syms ?

I’m not convinced this is useful. And if not useful then I am always worried it will cause an unwanted problem. That is why I did NOT recommend that step.

Well, you did not follow my recommendations, so I am not sure what to say.

I suppose you could (and should) provide the output, requested by caf4926

rpm --query --all '*kernel*'

here’s the output:
linux-kernel-headers-2.6.31-3.4.noarch
kernel-default-2.6.31.5-0.1.1.i586
kernel-default-devel-2.6.31.12-0.2.1.i586
kernel-source-2.6.31.12-0.2.1.noarch

I read how the kernels had to match exactly but I only saw one to install from (my) yast and that’s what I used. These aren’t a match are they?

FYI - I followed the directions explicitly the first time, that is without the 2 makes and I got couldn’t find headers message. I figured that step hadlready pointed out enough times that I should know to do it. The second attempt after 2 ‘make’(s) I no longer got the missing header msg but di get the "failed to build’ msg.

**** - I ran yast from su and searched and found many more versions there. Previously I ran yast (install software)from the desktop where I only had one match. Guess I should try this again with different kernel-source, eh?

I went to su and ran yast and searched for kernel-source and found 3 listed that matched 2.6.31.5 and tried to install one of them but got a message that indicated the wrong one. So I close that instance of yast and ran the one from the desktop. That one revealed 4 in the list. Only one of the was called kernel-source and none of them were rev 2.6.31.5

Why don’t I have a 2.6.31.5 in the desktop yast?

Looks like I didn’t install the correct source. But now I’m not sure where to get the correct one. And what appears to be the correct one won’t install.

The kernel-source version MUST match the kernel-default version. What is in place now means this build is likely to fail, and it did.

Either downgrade kernel-source to kernel-source-2.6.31.5-0.1.1.noarch OR upgrade your kernel to kernel-default-2.6.31.12-0.2.1.noarch.

As a matter of practise, I ALWAYS ensure my kernel-source version matches my kernel version.

You need to be careful what you are doing here.

If you go to YaST, and select an rpm, there typically is a tab that one selects that indicates the version (where the tab is called “versions”). One selects in the version area/tab as to what version of an application is to be installed.

from a terminal-> su yast the greatest resolution I have is 2.6.31.5 noarch. I have 4 hits in the window that match this. Are they all the same? The versions appear to be but the descriptions are slightly different.

su terminal and then

zypper up

should bring everything in order

Assuming your repo’s are OK

zypper lr -d

zypper up is running…
Still learning my way around. I found the repositories in the desktop yast

after zypper commands have completed, do I need to make cloneconfig & prepare or can I straight to running the NVIDIA-Linux-x86…pkg1.run ?
I realize I’ll need to reboot into su prior to this…

I see no need for cloneconfig. I’ve never once used that. But I do recommend you reboot after any kernel update.