Thursday 15 March 2007

"Computers don't make mistakes -- the things they do they do on purpose."

Well, I suppose I should provide a little background before I launch into my tale. In September of 2006 it was time for me to do some upgrading to my aging computer system. At that time I was running an AMD XP 2000, Geforce 3 ti 200, 512 megs RAM, etc. I was anticipating the release of Neverwinter Nights 2 as I had been fairly involved in the first version of the game creating custom content, doing scripting, etc. I like to mod games -- it's my hobby. My old system wasn't going to cut it for running the new version of the game -- particularly the video card as the game's graphics engine had been completely replaced and required at least a shader model 2.0 capable video card. I was faced with simply upgrading the video card or doing a more full system upgrade. With PCI express interfaces becoming the standard over the older AGP interface it didn't seem particularly wise to buy a new video card based on an old interface since I would then be locked in to staying on an older platform. So, if I was going to go with a PCI express video card I also needed to upgrade my motherboard, RAM and CPU as well.

After a bit of research and poking about on the net I found an online retailer and put together a few possible combinations of hardware and priced them out. I ended up going for a MSI motherboard based on the 550 Nvidia chipset, an X2 3800 AMD dual core CPU, 1 gig of Geil DDR2 low latentcy RAM and a 7900GS Palit video card. The total came in at just over the budget I had set for myself -- £400. The system wouldn't be bleeding edge, but I never go for bleeding edge anyways as it's just not value for money. I probably would have been better off to wait a bit, in hindsight, as the core 2 duo's were out shortly thereafter and showed a fair bit of performance edge over the AMD offerings. No biggie really, as you have to pull the trigger at some point and actually make a purchase or you would never buy anything waiting for "what might be next". The system was still plenty beefy for what I do with it. The dual core CPU was a nice step up as I tend to have several hefty applications open at the same time while working on mods.

Anyway, I assembled the system without any serious hitches and all looked good. Neverwinter Nights 2 was released on October 31, 2006 (boo, scary, hehe) and the sytem handled it pretty well. The game is a bit of hog in terms of video card performance and received a patch or two to tighten it up, so performance improved a bit. All was well until the end of February or beginning of March this year. At this point I would occassionally notice that in a 3d scene it would appear that textures or polygons on certain areas of a model would be stretched out as if someone had pulled a vertex or two way out of position. I first noticed this on the cloak of my game character and originally thought a skinmesh weight was way off for the animation -- like a vertex or two had some whacky weight assignment. However, minimizing the game and then maximizing it would correct the problem, so it seemed it was actually a rendering error and not something inherent to the game, model or animation. I shrugged it off as I was running the game in windowed mode with a few other apps running as well.

I then began to see this behaviour in other apps/games as well as textures flickering, pixels turning white or other incorrect colours. Now sensing something was going seriously wrong somewhere I first checked temperature levels for the video card -- all looked good as it idled at the desktop between 39c-42c and under full load it was hitting about 55c-57c -- all well within normal ranges. I took the card out and checked for dust in the PCI-e slot as well as for anything on the card's "golden fingers" interface. All looked fine so I put the card back in and booted back up. Things seemed okay for a bit so I thought it was just "one of those things" that computers are so good at doing -- anyone who hasn't experienced the odd unexplained computer quirk is a lucky person. Although, hehe, Dale Gribble from King of the Hill had a great line in one episode where the permanently paranoid Dale intoned, "Computers don't make mistakes -- the things they do they do on purpose."

Unfortunately, it wasn't "just one of those things" as I was now starting to see pixel corruption on the desktop where a few pixels would be the incorrect colour (usually red, green or blue) or even flicker between various colour states -- not good. At this point I reached for Google and start looking for any accounts of similar types of display corruption (1 2 3 4). I came upon a few threads on the Nvidia boards (1 2 3) where the exact same corruption was being discussed as well as screenshots posted. The screenshots were identical to what I was seeing and there were even accounts given of the same pixel corruption on the desktop I was seeing. The problem didn't seem hugely widespread, but those that did have it all shared almost identical symptoms. Most of the discussions were of the 7900GT line of video cards, but those discussions had mostly began in May of 2006 before the GS line was released. The GS line is basically a cut down GT core (and may even share the same printed circuit board [PCB] layout as the GT), so it seemed reasonable that the fault could have easily migrated through to the GS line of cards. Most everyone who had the problem needed to RMA (return merchandise authorization) their video card and have it replaced. In short, if you had/have this problem your video card is dying and needs to be replaced...

No comments: