Official Linux driver 1.0-4191 thread

[volt]

NVIDIA has posted new Linux drivers:

IA32
Version: 1.0-41.91 (http://www.nvidia.com/view.asp?IO=linux_display_ia32_1.0-4191)
Release Date: Dec 11 2002
Release Highlights
* OpenGL 1.4 with CineFX architecture support
* Support for AGP 8x and nForce2 IGP
* Support for index overlays on Quadro4 to support legacy applications
* Support for separate X screens on a nView enabled GPUs
* GLX 1.3 support

IA64
Version: 1.0-40.50 (http://www.nvidia.com/view.asp?IO=linux_display_ia64_1.0-4050)
Release Date: Dec 11 2002
Release Highlights
* OpenGL 1.4 with CineFX architecture support
* Improved SPEC Viewperf performance by up to 20%
* GLX 1.3 support

Athlon 64
Version: 1.0-4180 (http://www.nvidia.com/view.asp?IO=linux_display_amd64_1.0-4180)
Release Date: Dec 11 2002
Release Highlights
* First publicly available Athlon 64 graphics driver
* OpenGL 1.4 with CineFX architecture support
* GLX 1.3 support

[mtrr]

before anyone complains without reading the documentation:

the name NVdriver is obsolete,
new name is nvidia.o (modprobe
nvidia should work)


cheers

[tiger]

I just installed the new driver but when starting X the screen turns black and nothing happens. CTRL-ALT-Backspace has no effect. CTRL-ALT-Del causes the machine to start the shutdown procedure but shortly after the machine freezes. When I disable AGP support, I can start X but of course with degraded performance.

The XFree log looks normal, the last entry is something about initializing AGP8x

My hardware-configuration: MSI KT4 Ultra (KT400 chipset with AGP8x), Athlon XP 2400, MSI GeForce 4 ti 4200 AGP8x 128MB, kernel 2.4.20-ac1 (for KT400 support)

Is it possible to disable AGP8x initialisation to try AGP4x? I guess my problem is the KT400 chipset and/or AGP8x...

btw. using kernel agpgart fails with "agpgart: unable to determine aperture size" so I am using the internal AGP support of the driver.

Tiger

[misleb]

Holy cow. I just got this driver and the kernel module (nvidia.o) is 1.7MB! Thats insane. Its up 50% from the previous release. Please tell me that its "mostly useful, human readable, debug/error strings and sanity checking" :-)

Seriously, what is gong on here? 1.7MB seems a little excessive for a kernel module. I realize that to make video as fast as possible, a lot of stuff needs to go into kernel space, but what about system stability? Isn't one of the reasons that Windows is/was so notoriously unstable is because they put the graphics at the kernel level?

Can anyone comment on the possible tradeoffs of disadvantages of having so much graphics code in the kernel? Is the size of the kernel module a result of the "universial" nature of the driver? I'm not so much concerned with the amount of RAM used as the bloat itself.

-matthew

[3777]

You people better help me RIGHT NOW!!! I installed the supposed Nvidia driver Rpm for 8.0 and along with GLX, now I cannot get into Xserve. It took me 3 days to install the last drivers and get my system configured and everything working, and now everything is blown thanks to this b$. I am getting a XIO fatal IO error 104 on Xserve. Fatal server error no screens found, I can edit my Xconfig but changing driver from nvidia to nvidia.o has not helped. WHAT DO I DO???:mad:

[3777]

this is very frustrating, I am trying to manually re-install the old drivers to get Xserve again. Anyone reading this DO NOT install the rpm by Nvidia, thay are junk and will wreck your system. Try to install the new drivers via Tar.GZ, that what I'm going to do if what I'm trying to do now works

[bwkaz]

OK, the error that first brought me here ("dlopen: /usr/lib/libGLcore.so.1: undefined symbol: __divdi3" message in /var/log/XFree86.0.log) has apparently recurred. :mad:

Version 3123 fixed the problem, which was good, but making 4191 reintroduce it? Come on, guys! ;)

Well, I for one will be going back to 3123 until this gets fixed... :rolleyes:

Oh, by the way, the problem as I understand it is with gcc 3.2, so anyone who used that to compile glibc would be well-advised to either just not bother with 4191, or try it anyway but be warned -- you may not get "glx" to load due to this error.

Is anyone from nVidia reading this? (Andy?) Can you get someone to fix that bug again? -- I'm assuming it's a simple change, since the last time I asked linux-bugs, it was a known problem.

3777 -- post your X log (/var/log/XFree86.0.log, AS AN ATTACHMENT PLEASE ;) ) and your hardware.

[mtrr]

calm down boy, don't change your XF86Config(-4)
or whatever redhat calls it.

look in /etc/modules.conf and change every appearance of 'NVdriver'
with 'nvidia' and do a depmod (-ae) after.


cheers

[3777]

On a laptop right now, system with Red Hat is Athalon XP2100+, 1 gig DDR, and TI4600 card, running Red Hat 8.0 ver. 18.8 & XP Pro. I have everything back, but those Nvidia rpm's SHOULD install a proper Xconfig file as part of the process. I dont' understand how Nvidia can just put out rpm's that for some people .......could lead them to have to re-install .....or completely give up on Linux. I will try to make the new drivers from Tar.gz, if that works great, if not I will go back to 3123's, but please people build the drivers, do not just install the rpm, they just don't work.

[3777]

Originally posted by mtrr
calm down boy, don't change your XF86Config(-4)
or whatever redhat calls it.

look in /etc/modules.conf and change every appearance of 'NVdriver'
with 'nvidia' and do a depmod (-ae) after.


cheers

Ok so this a new one. I edit the modules.conf file as well? (See Nvidia you should have the rpm's do this !!!!) I wll still create the new driver via Tar.GZ, and rename the modlues.conf and see if it works...

[michael]

This sucks !!! I dont know if it`s a SUSE thing but Ti4200 is NOT working with SUSE 8.1 SMP Athlon. Holy shi*

The last message in *.log is :

(**) Mouse[1]: Buttons: 5
(II) XINPUT: Adding extended input device "Mouse[1]" (type: MOUSE)
select returned 1
select returned 1 .... (a few times)
...
select returned 0 ....
....
this goes on forever. The System is not crashed it can be reached via telnet and rebooted without any problem.

When using "nv" instead of "nvdriver" and NOT loading "glx" at least X can be started beside the fact that some minor applications like xterm or emacs do produce some fly footprints on the screen. I'll give it a last try with mandrake and if this is not working you'll find a Ti4200 at a reasonable price at ebay

Regards

Michael

[volt]

Why the hell do you use RPM's. That's the worst thing you could have done. :rolleyes: Compiling from source aint hard nor time consuming. I haven't compiled new drivers but NEVER had problems with older versions (Both Redhat and Slackware). Have you read README ? You could have done something wrong, so please stop blaming Nvidia unless you are 100% positive. Once you install the drivers MAKE SURE your XConfig file is correctly edited (read the notes in README). The new drivers may overwrite your existing file so double check your entries.

[3777]

Originally posted by volt
Why the hell do you use RPM's. That's the worst thing you could have done. :rolleyes: Compiling from source aint hard nor time consuming. I haven't compiled new drivers but NEVER had problems with older versions (Both Redhat and Slackware). Have you read README ? You could have done something wrong, so please stop blaming Nvidia unless you are 100% positive.

If the RPM's aren't any good...... then why does Nvidia offer them in the first place:o

[volt]

Because it's point 'n click but it doesn't mean it will always work.

[Andy Mecham]

The RPM will automatically edit /etc/modules.conf for you. If they didn't, then you should make sure that you properly installed the RPM according to the instructions in the README.

--andy

[mattmcg3]

Originally posted by mtrr
calm down boy, don't change your XF86Config(-4)
or whatever redhat calls it.

look in /etc/modules.conf and change every appearance of 'NVdriver'
with 'nvidia' and do a depmod (-ae) after.


cheers


Hi mtrr, I'm wondering what "do a depmod (-ae) after" means. Can you explain? Thanks.

Matthew

[3777]

Ok the rpm's are junk, but if you downlaod the tar.gz files and install manually......... here is what you get if you have a high performance system....when playing UT2003 on Asbestos:

Under 3123: XP Pro fps: 45-80 Linux 10-40fps

Under 4191: XP Pro fps 45-80 Linux never below 50


WOW...... and into the 100's most of the time........ so the rpm's $uck, but GREAT JOB on the drivers themselves, now I can really play UT2003 whether I'm on XP Pro or Red Hat 8.0 desktop

:D

[michael]

I have compiled from source. Modified XF86Config and so on.

No error messages at all in the log files just a black screen. Does anybody have a working configuration with Suse8.1. I gonny blast his head off (possibly :-) in UT2003

Michael

[adnans]

Excellent work on the new drivers NVIDIA!!!

The only problem I've seen so far is massive 2D slowdown. When starting the X server 2D is really unbearably slow (RenderAccell is ON btw). However, things are back to normal after running anything that uses libGL (glxinfo is sufficient). Can anyone explain this?

-adnans

[John]

Seems to be working on an ASUS P4S533, P4 2.5Ghz with PNY Geforce4 Ti 4600.

Been running it all day with Openscenegraph and Redhat 8.0 with 2.4.20 kernel. It seems faster and hasn't frozen yet. I also use full screen antialiasing with OpenGL/OSG by typing
export __GL_FSAA_MODE=2 before running the scene graph viewer (a 3d viewer).

Installed by removing old GLX and NVdriver without X loaded. (replace xxxx with your current version)
rpm --erase NVIDIA_GLX-1.0-xxxx
rpm --erase NVIDIA_kernel-1.0-xxxx

Then rebuilt for my kernel the nvidia driver.
rpmbuild --rebuild NVIDIA_kernel-1.0-4191.src.rpm

Then installed it and GLX
cd /usr/src/redhat/RPMS/i386
rpm -Uvh NVIDIA_kernel-1.0-4191.rpm
rpm -Uvh NVIDIA_GLX-1.0-4191.rpm

Rebooted into redhat without X and made sure that the new NVIDIA kernel module will load by typing (this is the module thats changed names from NVdriver to nvidia:
insmod nvidia

Then started X:
exec startx

I like this method because I can recompile the kernel src for what ever kernel I'm running at the time. Note I use the Option "NvApg" "1" for X so that the Nvidia AGP is used although I have used agpgart for some chipsets.
YMMV

[dlpierce]

Does "Support for separate X screens on nView enabled GPUs" mean I could connect two monitors, two keyboards and two mice to my computer and configure it to have two workstations using a single video card? It would be sweet to be able to get school work done next semester while my g/f plays neverwinter nights.
~Dan

[thecombatwombat]

I'm also seeing some serious 2d slowdown like adnans, (on debian sid with xfree 4.2.1) haven't tried running glxinfo, did run glxgears, saw a nice speed boost (~150 fps) but still uninstalled the drivers just after for 3123.

I'll give it another shot and try running glxinfo, that's awfully annoying though, maybe it'll work if I add it into my ~/.xinitrc, still seems like a pretty serious problem to me, I may just not screw with this release.

[gigadude]

I tried upgrading my Mandrake 8.2 Dell Inspiron 8000 laptop with the RPMs and ran into two problems:

1) I've got two different monitor sections in my /etc/X11/XF86Config-4, and for some reason running rpm -Uvh switched from the one that works to the old Dell monitor setting which had bad horizontal/vertical refresh values. The old Monitor line was left intact but commented-out, switching back to it got things working once I resolved #2

2) The nvidia.o module was not getting installed at boot time. I'm not sure which thing fixed this problem, but I did both a depmod -ae and performed an edit on /etc/modules (NOT /etc/modules.conf) to change the NVdriver to nvidia. /etc/modules now looks like:

------------8<----------
# /etc/modules: kernel modules to load at boot time.
#
# This file should contain the names of kernel modules that are
# to be loaded at boot time, one per line. Comments begin with
# a `#', and everything on the line after them are ignored.

nvidia
------------8<----------

I can't find any documentation on /etc/modules, either on google or in the manpages.

Rebooting after those two changes did the trick.

[Soul-Crusher]

Great job! Installation by tarball took only a matter of seconds and UT2003 performance seems to have jumped way up! WOOHOO!

[etr]

Originally posted by michael
When using "nv" instead of "nvdriver" and NOT loading "glx" at least X can be started beside the fact that some minor applications like xterm or emacs do produce some fly footprints on the screen. I'll give it a last try with mandrake and if this is not working you'll find a Ti4200 at a reasonable price at ebay

It sounds like you may be mis-configuring a couple of files.

First, you should use a driver of "nvidia" in your XF86Config/-4 file. The "nv" driver hear applies to the stock XFree86 driver, rather than nVidia's. As far as I know, there is no XFree86 driver named "nvdriver".

It's a different story in the /etc/modules.conf file. Here, you used to need to provide an alias for the "NVdriver" (note the case) module. However, with the 4191 drivers, you need to provide an alias for the "nvidia" module, instead.

Edited because I hit the wrong button and got the wrong quote