Video card fails, but still works with nouveau??

[doctordruidphd]

I have an interesting situation here. My system has two nvidia 9800GT cards in sli configuration, and works great with the proprietary nvidia driver. Until last night. First, I noticed some odd artifacts -- every time I tried to move a window, I would get snow all over the screen that cleared up after a second or two. Reboot, and system hangs dead. Hmmm. Booted into single mode, deleted xorg.confg, rebooted, goes into VESA mode just fine. Can't boot windows at all. A little more playing around, and I conclude that one of the cards is whacked. Can't physically remove it, as the neat little NVIDIA system board has these locking devices on the slot connectors I can't figure out. So I try booting the maverick system that has the nouveau driver installed, and up it comes, albeit with nouveau's minimal compositing. I'd much rather hack at it myself, but after all I did pay for the warranty, so until the tech gets here, the system is still alive.

So either the nouveau driver simply hasn't got to the point that it implements whatever feature of the card is actually failing, or it's fault tolerant beyond belief.

[dibl]

Re: Video card fails, but still works with nouveau??

Can't boot windows at all. A little more playing around, and I conclude that one of the cards is whacked. Can't physically remove it, as the neat little NVIDIA system board has these locking devices on the slot connectors I can't figure out. So I try booting the maverick system that has the nouveau driver installed, and up it comes, albeit with nouveau's minimal compositing.

Ouch! Good luck with the warranty .....

I find it odd that Windows won't boot. I wonder if the apparent video problem could be a side-effect of something else, like a problem in main memory? Just a wild guess. You're probably right about one of the Nvidia cards failing, in some 3D mode that nouveau doesn't use.

[doctordruidphd]

Re: Video card fails, but still works with nouveau??

I wonder if the apparent video problem could be a side-effect of something else, like a problem in main memory?

Main memory passes all its tests. My suspicion is that it's the memory in the card, which I really can't test -- there don't seem to be any diagnostics available. The system dies at the point that the fans slow down, both in windows and linux. Windows is using a proprietary nvidia driver as well; in safe mode, it boots, but ugly beyond description.

After about half an hour of playing around, Dell is sending new cards and a tech. Cheaper just to do it that way, rather than really dig into it. So until next week, I get nouveau, for which I am truly greatful. It has VESA beat by a mile.

[HighHo]

Re: Video card fails, but still works with nouveau??

Fairly common fault with some of the older Nvidia cards, the solder on the graphics chip breaks/cracks. Can be fixed with a reflow (although normally only a temp fix). Glad to hear that dell are replacing them for you.

[doctordruidphd]

Re: Video card fails, but still works with nouveau??

Actually, these were fairly new 9800GT's.
Replaced. working properly.
Seriously considering extending the warranty on the system.That little glitch more than paid for it.

[HighHo]

Re: Video card fails, but still works with nouveau??

I really recommend extending the warranty, almost however new they seem the solder is the same. I can assure you that this issue is far more common than nvidia admits, from the 6x series to the 9x series should be a permanent replacement program. I fix computers to component level everyday, and even if you dont take my word a simple google search will confirm the issues with nvidia solder used. I currently fix newish nvidia northbridges with the same issues on motherboards.

I personally believe nvidia have the best linux support with priority drivers, however there manufacturing process with 6x to 9x should be on a permanent warranty after the issues. This should of been done with permanent fix, but mostly is done with a replacement fix for another couple of years.