GL: colored garbage + freeze (7664, 8174, 2.6.13, 2.6.14, X11R6.9rc3)

[mfranz]

Since a few weeks I get occasional colored garbage when I run OpenGL applications (Blender, FlightGear, KDE's krita). There are colored 1 pixel horizontal stripes and squares of different sizes with random or repeating pattern. The problem didn't occur after any particular software upgrade. It happens with drivers 7664 and 8174 (no others tested). It happens with kernel 2.6.13 (SuSE 10) and vanilla 2.6.14. It happens with composite and renderaccel turned off. When such garbage occurs, it only takes some seconds until the driver freezes (only the HW cursor moves). Alt+SysRq+{S,U,B} works. When I exit immediately from the respective OpenGL application as soon as the garbage occurs, I can avoid the freeze. 2D does still work flawlessly, even with renderaccel turned on.

Now, since the problem started (IIRC) between software upgrades (X11, nvidia, kernel) and, thus, there is no apparent connection to any of them, I assume that the nvidia card is about to die. Also, it seemed to become worse over time. Could the onboard RAM be damaged? (I didn't do anything bad to the card. Also I checked that it's not loose in the slot.)

Is there a way I can verify this assumption? Can nvidia developers make any sense of the Xids or something else in the attached bug report log? If it's an unfixable hardware problem, I will have to buy another card (and hope that it lasts a little longer), but I don't want to waste money if it isn't.

Attached: bug report log; cropped screenshot (normally, the effect is more impressive and disgusting, but this is hard to catch in a screenshot if the driver constantly freezes).

[netllama]

Its difficult to make out much in your screenshot, as its really small. Are the green dots the corruption?

At any rate, this problem might be a hardware issue, but its hard to tell for certain. A few things to check:
0) Verify that you don't have an overheating problem. Make sure that there is adequate ventilation inside the case, and that all surfaces are relatively clean and free of dust.
1) Make sure that you have the most recent BIOS for your motherboard applied
2) Run a memory checker, such as memtest86 for at least two full passes

The fact that this problem is worsening over time does suggest that its a hardware issue. Before you run out and purchase new hardware, you might want to try performing a fresh OS installation to verify that the problem persists.

If you can provide me with a reliable means of reproducing the corruption, I can attempt to reproduce it here.

Thanks,
Lonni

[mfranz]

Thanks for your reply.

Originally Posted by netllama

Its difficult to make out much in your screenshot, as its really small. Are the green dots the corruption?

Yes. Attached is a better image. It is supposed to depict realistic landscape. Every colored spot in it that you wouldn't expect in Real Life is caused by the described problem (except the red one on top of the airport beacon pylon), including the black blobs on clouds.

Originally Posted by netllama

0) Verify that you don't have an overheating problem. Make sure that there is adequate ventilation inside the case, and that all surfaces are relatively clean and free of dust.

I should have mentioned what I had checked before: instable or too low voltage (I checked with a brand new power supply), loose card (removed and plugged in again; firmly in place), overheating (no sign of it -- the passive cooler was only warm after 2D use), dust (the card was extremely and surprisingly clean on the downside (mounting side), and only very slightly dusty on the upside; I blew that away with a bike pump to avoid moisture)

Originally Posted by netllama

1) Make sure that you have the most recent BIOS for your motherboard applied

The motherboard is only a few months old, and there were no updates on the vendor's site (MSI 865PE Neo2-V). It worked very well with the graphics card until the problems started to occur.

Originally Posted by netllama

2) Run a memory checker, such as memtest86 for at least two full passes

I'll do that over night. I had planned it, of course, but should have done so before posting. Sorry for that.

Originally Posted by netllama

If you can provide me with a reliable means of reproducing the corruption, I can attempt to reproduce it here.

I'm afraid I can't. All the programs that are causing this now (FlightGear, Blender, krita) worked very well before. They'd probably work for you, too.
I'm not doing anything special, and certainly nothing that I didn't do before it started.

I'll check RAM with memtest86 and will report tomorrow. Thanks for now.

[netllama]

Based on the new screenshot, this looks like it might be a hardware problem. I'd be inclined to agree that this won't reproduce for me.

If you could test the same videocard in a different system, that would be a good test as well.

Thanks,
Lonni

[mfranz]

Originally Posted by netllama

If you could test the same videocard in a different system, that would be a good test as well.

Not easily. Only skeletons lying around here, and friends don't trust me. ;-)

I did the memtest86, though. Over night, more than 8 hours. Not a single RAM error. So, essentially there were the motherboard and the graphics card remaining as possible causes. I would have expected more problems with 2D graphics and other PCI services if the motherboard were the cause, so I purchased a new (nvidia) card. It works very well so far, the artifacts are gone.

Thanks for your replies and expertise. This made the decision a lot easier. And maybe the thread even helps others. :-)