I am looking for knowledgable types to help me diagnose this annoying issue (or at least maybe to confirm or correct my thinking about it).

Basically, the rig in my SIG has been acting up in intensive 3d games. Well, in HL2 and F.E.A.R. anyway (and the Age of Empires III demo, which is not so gpu intensive, I'd guess). What happens is after 10-15 minutes in F.E.A.R. and 20-30 minutes in HL2/CS:S (and pretty much exactly 10 mins in the AoE III demo) there is an explosion of coloured pixels on a black screen, then the monitor loses its signal, there is no more sound, and I have to hard reboot.

It is only these games that present a problem, however. I have successfully played hours of Dungeon Siege II, Fable: The Lost Chapters, Psi-Ops, GTA San Andreas, Boiling Point and Fahrenheit with no problems at all, all at maxed settings.

Now, this rig *used* to play HL2 fine six or so months ago. Since then I have added a SATA HDD and a tv card and replaced 2x512 sticks of ram with 2x1024.

I have removed the tv card, unplugged the new drive, given the gfx card a dedicated molex, reinstalled older drivers, turned up the cpu fan speed, restored default clock speeds, cleaned the dust out of the gpu fan. Still the same problem.

Most people seem to think there are 3 possibilities:
1. Bad memory stick(s). I have pretty much ruled this out, THOUGH, as all other games play fine, and memtest and prime95 will run happily for days with no errors. I could possibly make a trip to my dad's and borrow his ram to check it out, but I'm not sure if it's looking likely enough to help to be worth the effort. Agree/disagree?
2. Power issues. I have also ruled this out of my thinking as the psu has powered everything I now have connected in the past with no problems and the gpu is on a molex with no other devices.
3. Overheating GPU. This is the one I am left with. I have tried running with the side of the case off, and this *seems* to buy me a few more minutes on average in F.E.A.R, but this might be just my perception. I have tried alt-tabbing out of FEAR to the gpu temp monitor, and it gets up to 90-92C before crashes. This seems very hot, but I have read, I'm sure, that 6800GTs can go up to 120C without damage?

One other thing: in event viewer when I restart after a crash there are always between 3 and 6 entries for Event: 2, Source: nvatabus, description: device identified. Is that significant? It means nothing to me, and is listed as 'information' rather than 'error', but it's always there.

Any help/ideas I would be extremely grateful for, enough to consider naming my firstborn in your honour...You Don't need to name your firstborn GX1_Man, but I, like you, think it is a heat issue with your Gfx card. Are more case fans an option? Does it go fine for hours with normal tasks - webs surfing, etc. If so, then the higher requirements of the game seem to excite it beyond coolness.  That is a pretty warm temp!

I would rule out #1 and 2 for the reasons you mentioned, except I didn't SEE that you mentioned the name and specs for your PSU.GX1_Man... hmm, well I only said I'd consider it

In racking my brains over this and specifically why it only happens in HL2 now when it didn't before, it suddenly occured to me that, of course, I'd moved house since then, and the pc is now in a different room, and in a new computer desk that doesn't allow so much AIR movement. Aha!

The only doubt I have here, though, is that I get none of the usual graphical corruption you'd expect with overheating gfx cards before it crashes...

More system fans aren't really an option as I have used up all the slots for them already and I don't want to dremel my case.

My PSU is still a possibility, though, as it is cheap generic one from ebuyer. Rated by them at 450W, but I wouldn't trust it too much.

But, yeah, the only time crashes ever happen are with the three games I mentioned: HL2, FEAR and the AoE III demo (which, surely, is not as graphically demanding as, say, Fable, which runs fine?!).

Thanks for the reply. Think I will try moving the pc to another room with better airflow...

Edit: Well, moving the pc and leaving the side of the case off made no difference at all! I did notice that the gpu hsf was very hot to the touch, but it still crashes after 10 mins whether the side of the case is on or off! Surely the case being open should make *some* difference to temperature?

Anyone got any opinion as to whether it'd be worth buying a new hsf, some cleaing solution and AS5 to try to cool this card? Or am I missing some other possibilities?Before you go out and purchase a new HSF try running it with the side panel off and a small desk size fan pointed at the vid-card...
If the results are satisfactory then it confirms a heat issue.

patio.  

p.s. what brand of PSU is it ? ?

Yeah, you know I've been searching the house for a desk fan all day. I could have sworn there was one in the garage or the spare room or... somewhere, but no

The PSU has a label on announcing it is made by ENCER. NEVER heard of them. It's probably a piece of junk, tbh.

I found something interesting just now, though: I got hold of a spare 512mb stick of ram from a friend, and stuck that in in place of my 2 1024 modules. Result: the Age of Empires III demo no longer crashes! It's juddery as sin, but it doesn't black-screen crash anymore. Hmmmm.

I can't really check if FEAR runs now, with only 512mb in the system, but it certainly makes all this look like a RAM issue after all.

A question, though: why would only certain 3d games but not others crash like this if there's some ram/motherboard conflict, but everything else run fine, and memtest and prime95 report no errors? I am bamzumped and gaboozled!Power Supply....

patio.  You think so? As I say, I haven't ruled this out, but I have no idea how a psu problem would manifest. Does it fit the profile that the crashes would always come at pretty much exactly the same time after starting one of the problem games (i.e. 10-12 mins in FEAR; 20-25 mins in HL2)?

Heck, maybe I could do with a new psu anyway. Wouldn't hurt, I suppose. Could always carry it over to my next build...PSU's can have heat issues too. Any PSU that cannot deliver full power up to 50C is a waste of time and money. I suspect that is your problem. I have seen 450W PSU's fail miserably where a 250W worked. Cheap PSU's will deliver full power at about 29C (roughly body temp) and lose approximately 10% for every degree above that. A PSU typically runs at 45-50C so your el cheapo is likely failing when demand increases. With the specs on your system you should get a really good quality PSU. It is the heart of your system. Quote

One other thing: in event viewer when I restart after a crash there are always between 3 and 6 entries for Event: 2, Source: nvatabus, description: device identified.