flat panel hard lockups

[elanthis]

I've never been able to get my flat panel (SyncMaster 172t) to work in digital mode. I posted a problem report on this previously and never had a fix, so I'll try again.

Some notes: this problem happened both on my old mobo (asus, via k266 based iirc) and the new one (fix an19c). This happened both on Debian/Sid with a custom 2.4.{18-20} kernel, and now on RedHat 9 with it's stock kernel. Happened both on older 3xxx series, and the 4xxx series.

The first post I made, I was told to check my ram (512 DDR, I forget manufacturer, some "cheap" brand, admittedly) w/ memtest86 - I do so, and no problem were found, and I posted those results (and never had a further response; ah well).

The problem only happens in digital mode, not using the CRT connection (this monitor supports both, and can switch between them on the fly). When using the DFP setting, the kernel locks hard - pings do not respond from other machines on the network once X tries to start. The monitor also starts acting rather funky (like it too locks up) until I unplug it from power and back in again.

If I eneable GLX, the logs show it locking up there. If I disable GLX, it locks up when RANDR is loaded. Will of course attach those.

The digital mode does work in the console just fine.

Specs and info:
Video Card: GeForce 3 Ti500
Driver Version: 1.0.4349
Distro: RedHat 9 + rawhide (including XF86)
Video Bios: 03.20.00.16.10
Processor: Athlon XP 1800+
Motherboard: FIC AN19C (kt400 chipset)

Those are just the current specs; had this same problem with older mobo and os, as stated.

[elanthis]

here's the log

[Andy Mecham]

First of all - make sure you're not hitting the KT400 hang. Update to 1.0-4363, and use NVAGP (not AGPGART).

I'll look into this.

--andy

[elanthis]

Updated to latest revision, and set NvAgp 1. Same problem. Attaching updated log.

One line sticks out to me tho:
(II) NVIDIA(0): Verifying AGP usage for Screen 0...
(WW) NVIDIA(0): Failed to verify AGP usage

Anyways, got to the GLX bit and locked up, like before - monitor got all screwy, altho this time it started working again after the computer rebooted (I usually need to power cycle monitor). The console and all that was running over CRT tho ( I have both the CRT and the DVI cables plugged in to both the display and the geforce card, to make testing easier - as a note, the problem is the same whether I do this or have only the dvi cable plugged in).

[Andy Mecham]

Why are you using the ConnectedMonitor option? If you have just the DVI connected as your sole boot device (ie: only monitor connected), what's the output?

--andy

[elanthis]

With only the DVI cable plugged in, the machine comes up fine, all the bootup/console stuff works fine, etc. When I start X, everything comes up, GNOME destkop all that, only I don't see anything. If I quit X (ctrl-alt-backspace) I go back to the console where everything is fine.

The monitor tries some mode changes it appears, then goes into the "no valid input" mode where the power led turns orange.

The log shows that it can't find a display, and is assuming CRT. DVI definitely works, tho, else the console wouldn't show...

[Andy Mecham]

(II) NVIDIA(0): SyncMaster 172t: Using hsync range of 30.00-81.00 kHz
(II) NVIDIA(0): SyncMaster 172t: Using vrefresh range of 56.00-75.00 Hz

From a quick search, it looks like your hsync range is inappropriate for digital. Can you try setting your hsync to 30-63?

--andy

[elanthis]

OK, tried hsync at 30-63, same thing - hard lockup.

And the log...

[Andy Mecham]

OK, i'll track one of these panels down and take a look.

--andy

[elanthis]

righty, thank you.

these things are hard to find; the only place I've seen that carries them in stock is CDW - dunno if you have super nVidia powers and can get samples/test-units tho. ;-)

is it also possibly a hardware problem with the graphics card? i've had several geforce cards fails on me, i'm not too confident in their workmanship. ;-)

the only other card I have to test around here is an old ati radeon 7000, I think it has dvi off the top of my head, but it certainly wouldn't be using the same drivers to test against...

(the card failures all were related to power issues I think - I've since changed the power supply, motherboard, and added a ups system, after having to get two replacements and an upgrade; this card has at least kept working for over 4 months, so that was probably the issue)

[elanthis]

Just an update; same hardware, still doesn't work. I have just yesterday upgraded to the latest nVidia drivers. The machine no longer hard-locks (I can ctrl-alt-del to reboot the dumb thing, anyways), altho I still need to power cycle the monitor afterwards for it to start working again.

DVI works fine until X starts.

Just a refresher:
RedHat9 + various Rawhide and other updates
Kernel 2.4.20 (athlon optimized)
XFree86 4.3
GeForce3 Ti 500
Samsung SyncMaster 172T tft display
Athlon XP 1800+
512MB DDR RAM
Via k400 based mobo (I think)

I'm getting pretty desperate to get this working; I want me my DVI. ;-)

[capybara]

I'm getting pretty desperate to get this working; I want me my DVI. ;-)

Wow, finally I found someone who has nearly my exact problem. See my post for more details. I'm using a GeForce3 Ti 200, so maybe the similarity between our cards explains why we see it and nobody else does.

With the 4496 drivers I get "No signal" and sometimes a brief "Out of range" error when starting X over DVI. My monitor doesn't need power cycling, and actually ctrl-alt-bksp brings me back to the text console with no trouble. I have played with all possible permutations of the options, and still see nothing on the screen.

Any ideas? I'll be watching this thread now...