Problem with an ATI Radeon 9600SE video card
 

The system is Debian Squeeze stable.
The problem is:
When the card is the most solicited it give a black screen for 3 seconds and then resume after.
I think it does a hardware reset.
A not too demanding game like NeverBall will do it.
Here is the result of dmesg | grep -Ei Radeon

[ 9.384828] [drm] radeon kernel modesetting enabled.
[ 9.384894] radeon 0000:01:00.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16
[ 9.387770] [drm] radeon: Initializing kernel modesetting.
[ 9.388283] radeon 0000:01:00.0: putting AGP V3 device into 8x mode
[ 9.388300] [drm] radeon: VRAM 128M
[ 9.388302] [drm] radeon: VRAM from 0x00000000 to 0x07FFFFFF
[ 9.388304] [drm] radeon: GTT 128M
[ 9.388307] [drm] radeon: GTT from 0xE0000000 to 0xE7FFFFFF
[ 9.388327] [drm] radeon: irq initialized.
[ 9.388545] [drm] radeon: 128M of VRAM memory ready
[ 9.388547] [drm] radeon: 128M of GTT memory ready.
[ 9.388791] [drm] radeon: 1 quad pipes, 1 Z pipes initialized.
[ 9.388801] [drm] radeon: cp idle (0x10000C03)
[ 9.389322] platform radeon_cp.0: firmware: requesting radeon/R300_cp.bin
[ 10.091617] [drm] radeon: ring at 0x00000000E0000000
[ 10.095880] [drm] radeon: ib pool ready.
[ 10.097023] [drm] Radeon Display Connectors
[ 10.614766] fb0: radeondrmfb frame buffer device
[ 10.630627] [drm] Initialized radeon 2.0.0 20080528 for 0000:01:00.0 on minor 0

You can see I installed a driver which comes from Debian non-free (See the sixth
line before the end).
Before I did that this command above was giving a problem report.

Yesterday I read that some ATI video cards don't support modeset so thinking it was my problem
I legitly configured my boot loader (Grubb) with GRUB_CMDLINE_LINUX="nomodeset"
like shown below. (... Yes I did the grub update and a reboot)


# If you change this file, run 'update-grub' afterwards to update
# /boot/grub/grub.cfg.

GRUB_DEFAULT=0
GRUB_TIMEOUT=5
GRUB_DISTRIBUTOR=`lsb_release -i -s 2> /dev/null || echo Debian`
GRUB_CMDLINE_LINUX_DEFAULT="quiet"
GRUB_CMDLINE_LINUX="nomodeset"

This seemed to have cleared the problem since the card behaved properly from 3:0 0PMto 6:00 PM
It never worked that well for so long.

This morning I gave it a test with NeverBall and it did resets all over.
I am doing some Python prgramming and I am using a terminal screen, when I scroll it sometimes the card will reset again.

So what could I try next to solve this
Thanks,

Is it possible the card is overheating? Does the card have a fan and, if so, have you checked to make sure it's working 100%?

Adam

I tend to not use the automated grub update these days and directly edit the cfg file, though I'm not suggesting that you should do this for one minute. However, you may be able to establish whether or not grub.cfg is being altered in between which could affect your gfx driver. However, I doubt it's a grub problem and I'll explain why it may be and why it probably isn't.

If you have done a full installation using Debian as the base, particularly using the update manager, or apt to upgrade the kernel, this will automatically run update-grub at the end. If there's anything awry with /etc/default/grub then any fixes may be undone during these updates. The easiest (and non-destructive test) is to press "e" on the keyboard at grub with your chosen boot option selected, which will show you (and allow you to edit for this boot only) your boot line. Check that nomodeset is still there.

Alternatively, you can examine grub.cfg directly (cat /boot/grub/grub.cfg) to check the boot entry.

The reason I don't think it's grub is that I would expect your gfx driver to fail before loading the desktop if there was a fault there. You can easily eliminate the driver as a problem with:
And reboot

This should boot using the vesa driver. You can go about your business without glx and see if you have the same problems. If you do, fglrx is not to blame. You can restore your gfx driver with:
And reboot

If the same problem appears after the first mv, then the problem isn't your gfx driver, and you can look elsewhere. Check the log files for any reports especially after a gfx crash.

Ok ,
About the grub tweaking I didn't see any changes with the line GRUB_CMDLINE_LINUX="nomodeset"
so I did change it for GRUB_CMDLINE_LINUX="radeon.modeset=0" and then I wasn't able to change the mode in the monitor setup.
Conclusion it seems the card supports modeset and I dropped this grub tweak.

I did many things here like going in aptitude and removing all the drivers concerning ATI and Nvidia.
I downloaded the proprietary driver from ATI and their doc said to remove and Nvidia driver.
When I tried to build the Debian package following their instructions it failed. These drivers are from over 6 months ago and
my kernel is too recent: if I interpreted the messages correctly.
So I went back and forth re-installing the ATI drivers in Debian.
There is no fan on that board but a large heat sink.

According to what you said Adam:
Now I am starting to think that it is a thermal problem or an hardware problem.
I got that card from Ebay they were supposed to ship me an ATI Radeon 9550 XL with 256 M but instead they shipped me that one.
So after the claim I got re-reimbursed and they told me to keep it: so it is a free-bee.

My next step: I will definitely remove the heat-sink and the old compound and put it back together
with the white compound I bought for sitting power transistors almost 30 years ago.

I looked in the log file Xorg.0.log and there is no error.
See the dmesg below (nothing to worry about)

/var/log# dmesg | grep -Ei "failed"
[ 1.024252] PM: Resume from disk failed.
[ 3.905775] PM: Resume from disk failed.
[ 19.427644] svc: failed to register lockdv1 RPC service (errno 97).
:/var/log# dmesg | grep -Ei "error"
[ 3.905772] PM: Error -22 checking image file

Since the system seems happy about this card I will reseat the heat-sink and
let you know.

Thanks to all for you input

Well well,
I don't have it easy ... must be my Karma.
I did take out the card removed the heat sink clean everything up.
Reseated it with a good silicon paste, power it up and played with it for a few hours -- the problem seemed to be gone.
I then let the power management of the system turn off the screen at night.
Guess what the morning after POW! balck screens all over the place.
So I disabled power management the problem remained ,then logoff logon the problem is still there.
I did a cold boot --Tower unplugged for 2 minutes -- the problem came back.
Then I shut down the tower and turn the screen off for 3 minutes.
The problem is gone :) :)
To test I played with the system for a few hours -- no more problems.
I set power management to shut down the screen and wait for a screen off to happen. POW The problem is back.
I then set power manager to leave the screen alone and I turned of the screen for 2 minutes -- the problem is still there.
I shut down everything and reboot-- Problem is gone.

I would like to know what cause this (Some kind of floating data in the screen memory and / or videp card ?)
I would like some hints about a possible solution.
By teh way the screen is a Samsung SyncMaster SA350 (brand new)

Weird or pfff.. what

Let me know,
Cheers,

Check your monitor. I had a similar problem, except mine wouldn't play ball when cold, but was fine after warm up. Turns out the monitor was dying. New monitor, problem gone. (maybe try and borrow one for testing purposes first - that's what I did).