Screen corruption, crashes, no NVIDIA-GLX

[ecoffey]

Hi, I'm trying to get X up and running on my Debian system to no avail.

The symptoms are as follows:

On startup short black horizontal lines are displayed across the top of the stippled b&w X background, then there is corruption all over the screen as the Gnome splash screen displays the Gnome initialization process.
Random junk sometimes appears alongside new windows and other on-screen changes.
xdpyinfo shows GLX, NV-CONTROL and NV-GLX but no NVIDIA-GLX
glxgears runs and crashes the computer (niether CTRL-ALT-DEL nor CTRL-ALT-BKSP work), or prints "Invalid instruction" and exits, or prints "Sefmentation fault" and exits.
X crashes sometimes when opening an application or simply clicking a link or button in Mozilla, or for no apparent reason at all.
"NvAgp" "0" prevents X from loading at all. Other options have no effect.
No serious looking errors/warnings in X log.

The setup is as follows:

Soltek 75MRN-L motherboard, AMD Athlon XP 2600, onboard GeForce4 MX (Video BIOS 04.1f.00.08.00), 512MB ram (32 for video), nForce2 chipset.
Newly installed Debian system with kernel package 2.4.24-1-k7 and XFree86 4.2.1.1.
Installed nForce drivers from NVIDIA_nforce-1.0-0261.tar.gz tarball - all went smoothly.
NVIDIA-Linux-x86-1.0-5328-pkg1.run produces no warnings
All files and symlinks in place
glxgears linking against tls libraries

I have looked long and hard bat cannot find any reports of symptoms matching these, let alone a solution - does anyone here have any idea what might be wrong?

[whig]

Do any of the XFree provided drivers (vesa, nv) go? Either will give you a starting point - run xf86config

[ecoffey]

nv gives less screen corruption, but crashes soon after launching. VESA, in glorious 8-bit color, works just fine.

[ecoffey]

Has anyone even seen this kind of behavior before? I have attached a log from X for your viewing pleasure.

[Andy Mecham]

As a sanity check - are you sure your system RAM is good? Corruption with the nv driver worries me. You can run memtest86 to check your ram.

--andy

[ecoffey]

Just downloaded the 5336 drivers - no change.
glxgears came up with an interesing new message the first time I ran it:

glxgears: relocation error: glxgears: symbol gettimeofday, version GLIBC_2.0 not defined in file libc.so.6 with link time reference

It crashed the box te next time.

I'll run memtest86 today and see what it reckons about my RAM.

[Andy Mecham]

It sounds like something fundamental is wrong with your install, or something has become corrupted beyond use.

--andy

[ecoffey]

This is my second fresh install on this machine, the previous install also suffered this problem. The machine has been tested with Windows and works perfectly. I ran a single pass of memtest86 and it found no problems. I have removed all PCI cards from the machine, underclocked it, turned off as many features that might increase instability as I could find and there is no change. The error and repeted segfaults from glxgears seem like some kind of library/binary-interface/compiler incompatibility.

I've attached my current XF86Config-4 in case I'm missing something obvious.

[ecoffey]

I'm running out of things to try. I assume other people have got the driver running without screen corruption and crashes, and with the NVIDIA-GLX extension loaded using an onboard GeForce4 MX, nForce2 chipset, Athlon XP CPU and the latest Debian-unstable/testing yes?

Another thought - the screen the GeForce puts up just after power-on says "Engineering Release - Not For Production Use" - should that worry me?

[Andy Mecham]

Another thought - the screen the GeForce puts up just after power-on says "Engineering Release - Not For Production Use" - should that worry me?

Yes - where did you get the card? Did you buy it at retail?

--andy

[ecoffey]

I bought the system retail from a small supplier in Adelaide (I'll send you his contact details by email).

Soltek's website for the SL-75MRN-L motherboard makes no mention of the issue:
http://www.soltek.com.tw/soltek/prod...N%2FSL-75MRN-L

But it seems some others have had similar problems (see response by poster Philippe L):
http://www.pc97.com/forums/index.php...1&overclockid=

[ecoffey]

So what do I do next - This problem renders the computer essentially unusable, so I'd really like to get a replacement board as soon as possible, should I be looking to my supplier, Soltek or nVidia?