So I have contacted plenty of folks about this, and nobody seems to have really any idea about what might be causing it. So... here's to hoping someone around here has an answer:
I have this video card. It's an ATI 5670 HD manufactured by HIS. I replaced it after my 9800GT fried itself out, of which this new problem never occurred with. With the 5670 in, while I am playing certain games (I believe it only happens with certain games, anyway) my computer will just spontaneously restart. There is no error, no difficulty in reloading, it just goes black and restarts. I did RAM swaps, heat checks, fan checks, a processor check, and the new 5670 is supposed to run on something like 30-40% less power than my previous 9800GT did. I have no idea what the issue is. The only time I've been able to avoid the issue entirely is when I've swapped out the card completely with an old AGP card, the 2600 PRO. Just to make sure it wasn't a graphical thing, I played again on the 5670 on the same low settings as I would with the 2600 and the restart still occurred. So, I sent it in for RMA, and after 4 weeks I finally receive the new product and the same issue still happens. The Guild 2 and Starcraft 2 are the most notable games that come to mind where this bizarre reboot seems to happen, and Street Fighter 4, Lord of the Rings Online, and Civilization V all seem to run the game just fine without restarting my computer. There is no drop in performance before or after my computer restarts, either. My motherboard is an ASUS P5K Pro.
Does anyone have ANY clue as to what the hell is happening here? There are probably details I'm missing here, but if they're necessary, I can easily provide them.
Whats your GPU/CPU temps
Yeah, see, I thought about all this. And the weird part is that it only appears to happen in SOME games. I've checked with the HIS manufacturer on it and they're just.. not being very helpful at all. I played Modern Warfare 2 for an hour or so tonight and nothing happened. Usually I get about 3-4 Starcraft 2 maps before it all restarts, and only about 15-20 minutes of the Guild 2 before it shuts off.
I will try to get my temps tomorrow, and yeah it's a home rig.
Processor: Intel Core 2 Duo 2.54Ghz, 4GB of RAM, windows XP 32-bit, ATI 5670 HD.
Power supply: 450W
450W - What manufacturer? What model? How old is it? Computer on 24/7/365 or only while in use?
I've had this sort of behaviour crop up a couple times now between the 4xxx and 5xxx series. Though that's out of quite a few of the cards.
Not sure if I'll be home tonight or not, so may have to wait on this stuff. But I do turn my computer off overnight.
He does not need a Megawatt PSU
650watts should be plenty for his PC and for future upgrades, and it can be gotten for much cheaper should he need to buy one for testing.
if his PSU was low quality or really old, I could see it, as PSU's lose capacity over time
I suggest bringing home a more powerful PSU, swapping it out, and seeing if the issue occurs. My guess is that it won't.
Usually video cards crapping out are either thermal problems (or chip damage accumulated from thermal problems), and power. The system spontaneously restarting isn't necessarily indicative of a software problem (unless you've set your system to reboot on a bluescreen, and I don't think that's the default).
The card's new (although one should be cautious about rejecting possible sources of problems just because equipment is new - lemons are out there). OP hasn't posted card temps, but for the moment, let's assume they're fine. If it was a thermal problem, I'd tend to think it'd happen pretty reliably as soon as the card went into 3D mode at all - not just when the card's really taxed. Chip temperatures spike as soon as you turn the thing on, it only varys very little with load. Being right on the very edge of thermal failure only at the high end of GPU loading isn't impossible, but it would be somewhat exceptional.
450W should be OK for that card, it is supposed to pull less juice than the one the OP replaced it with. However, 450W power supplies tend to be at the low-end (and the no-name ones at the low end tend to be really bad). The power supply could have drifted out of specification, or although the new card is drawing less power it might be drawing it over over a different voltage bus than the previous card did (one that might already be heavily loaded), and it's overloading.
I concur, 650W should have more than enough headroom, even with spec slop, and since they're much more common, a good name-brand one ought to be reasonable. I think replacing the power supply would be a reasonable next step in troubleshooting.
I mean, just look at this review which is using an overclocked i7 965 which is (I think) pretty much on par with the power draw and it's only using 226 watts under load running furmark.
It's not so much how much total wattage (measured at the mains outlet) you're drawing, it's how much your pulling off of any voltage rail at any one time. PC power supplies aren't designed to supply all (or even a large fraction) of their total rated wattage to any one rail at a time. If you exceed the rail's amperage, the power supply is going to take a dump on you.
And since most manufacturers tend to bury (or not provide) those sort of detailed specs, and the plugs aren't necessaily labeled as to which rail they're on - and most people aren't set up to put digital ammeters that have a max reading hold in-line with their equipment... it can be hard to troubleshoot.
I hate telling people to buy new components too , especially if it's not a slam-dunk it's the right answer, but I think the OP Buy/beg/borrow a better-specced PS is the best route for the moment.
I always recommended completely removing the previous drivers then doing a fresh-install of the latest drivers for the new card you're installing. I would also recommend checking the crash logs to see what clues you might find.
Although I would certainly also double-check to make sure it's not an issue of excess heat or insufficient power supply. To troubleshoot those, you might try putting the computer through its paces with the chassis open to allow more airflow, or swapping out the power supply as everyone has helpfully described.
Drivercleaner or Driversweeper, I don't remember which of the two is the newer one, but use that.
A 9800GT consumes more power than a 5670, I highly doubt the PSU is at fault here.
Edit: Heck, I've even had trouble when I went from 10.7 Catalyst to 10.6 Catalyst. I did a revert because of some trouble, but some features remained (GPU rendering of Flash 10), so when my card would try to access them: Bleep! Blue Screen of Death.
Do a driver sweep and report back!
a restart is most likely a PSU issue
I just want to say that my 9800GT NEVER had this issue. Ever. Not once. I only got the 5670 to replace the 9800GT when it had an issue, and this 5670 is supposed to run on 35-40% LESS power than the 9800GT does. That's why I bought it.
I'm pretty technologically inept, but shouldn't that mean it should... work just fine, as far as power is concerned?
Does the new card offload anything onto the CPU compared to the other card- say, physics calculations or anything the old card did for itself?
It might be hiding its power usage by doing that, in which case the CPU would be drawing more power.
For giggles, you could pull out the video card and get some compressed air to blow out your PCI-E slot. A stray dust bunny may not be the culprit but it couldn't hurt.
There should be a sticker on the side of your PSU that denotes its power output. How many amps is it supplying your 12V rail?
Re: Blue Screen vs. Restart; OP's WinXP install is likely set up to restart when it blue screens, as that is the default setup. So he could be getting blue screens but the computer could be rolling directly into a restart so he's missing the message (I've had this happen to me).
I have never understood why this is the default. It is god damned obnoxious.
It is also the default in 7.