System unstable

[makken]

I just put together a new rig not too long ago and seem to be having stability problems in games. Here's what I've been observing:

Dawn of War 1: randomly crashes with the warning "Display driver atikmdag stopped responding and has successfully recovered."
Tomb Raider Underworld: randomly crashes to desktop, seems to throw an access violation error message. artifacts with shadow effects seem to happen sometimes, though not frequently.
Dawn of War II: randomly crashes to desktop, very frequent, also seems to throw an access violation error message
TF2: rock solid
Prince of Persia, sands of time: rock solid
Quake 4: crashes whenever I try to enable AA in game, no issues when I use AA from the catalyst center.

Any idea what may be causing this?

I started with cats 9.3, upgraded to cats 9.4 a few days ago, with no effect on stability.
Temps seem to be a bit high, but acceptable - CPU at low to mid 50s under load, GPU at mid to high 40s depending on ambient temps
Everything's running at stock speeds - 2.5Ghz on CPU, 815 gpu core, 925 memory
I have no stability issues when stressing my cpu alone with linpack

Specs are
E5200
Gigabyte UD3L
Asus 4870
4GB ram
BFG GS550 550W PSU

 

[Sylvanas]

1. Check you installed your drivers correctly and removed old ones from the way I describe at the end of this thread
2. Makesure games are patched
3. Try a different 6pin connector attached to the graphics card.

 

[Beanie46]

Honestly, I'd worry about your power supply. Why? My reasoning is given below....a little long read but will make sense if you take time to follow through it all.

First, that GS550 is a way old designed power supply and is group regulated. How do I know that? Simple. That power supply uses a voltage selector switch on the rear of the case to choose proper voltage. This is indicates an older design, before PFC was introduced and also shows the rails are regulated as a group instead of independently, as newer designs with PFC are. This can and occasionally does introduce poorer voltage regulation in the power supply's output, esp. when pushed to its limits.

Next, that power supply has two +12V rails, each rated for 18A max., but combined can only put out 396W, or 33A, meaning that if one +12V rail is pulling like 17A, that leaves only 16A on the other. Pull 18A from one and you only have 14A left for the other rail. Not exactly the stoutest of 550W power supplies, esp. when contrasted to more modern designs. Take a Corsair 450VX.....that 450W unit has the same +12V output rating as your 550W unit. The Corsair 550VX has a +12V rating of 41A, or 492W, a LOT more than your "comparable" 550W BFG power supply. Heck, even an Antec Earthwatts 500W unit has more output on its +12V rails that yours does.

Given your ps's low output on the +12V rails, I wouldn't be surprised if you're not trying to pull too much power out of your PCI-e connectors, given that you have two 6-pin connectors on your video card and no matter how the connectors are arranged, if pulling anywhere near their theoretical max. output (6-pin PCI-e connectors are rated to give 75W each), you're pulling over 10A through them for the card itself.

To add insult to injury, that GS550 had its output rated at 25C.....a completely unrealistic temperature, room temp. Once placed inside a case with the computer running, the power supply will be subjected to temps approaching 40C. And it's a fact that as power supplies get hot, the output they produce derates, or drops in other words. So, a power supply rated for 550W at 25C may drop its true output enough when subjected to temps approaching 40C.....enough to make your +12V rails now only have like 12A-13A per rail....maybe less. (A drop to 13A per rail is only down 3.5A per rail at max output, not a lot, like ~20%, not unheard of at all.)

So, you've got an old design, weak +12V output power supply trying to run a known power hungry video card. Bet that's where your problem is coming from.

 

[lavaheadache]

Originally posted by: Beanie46
Honestly, I'd worry about your power supply. Why? My reasoning is given below....a little long read but will make sense if you take time to follow through it all.

First, that GS550 is a way old designed power supply and is group regulated. How do I know that? Simple. That power supply uses a voltage selector switch on the rear of the case to choose proper voltage. This is indicates an older design, before PFC was introduced and also shows the rails are regulated as a group instead of independently, as newer designs with PFC are. This can and occasionally does introduce poorer voltage regulation in the power supply's output, esp. when pushed to its limits.

Next, that power supply has two +12V rails, each rated for 18A max., but combined can only put out 396W, or 33A, meaning that if one +12V rail is pulling like 17A, that leaves only 16A on the other. Pull 18A from one and you only have 14A left for the other rail. Not exactly the stoutest of 550W power supplies, esp. when contrasted to more modern designs. Take a Corsair 450VX.....that 450W unit has the same +12V output rating as your 550W unit. The Corsair 550VX has a +12V rating of 41A, or 492W, a LOT more than your "comparable" 550W BFG power supply. Heck, even an Antec Earthwatts 500W unit has more output on its +12V rails that yours does.

Given your ps's low output on the +12V rails, I wouldn't be surprised if you're not trying to pull too much power out of your PCI-e connectors, given that you have two 6-pin connectors on your video card and no matter how the connectors are arranged, if pulling anywhere near their theoretical max. output (6-pin PCI-e connectors are rated to give 75W each), you're pulling over 10A through them for the card itself.

To add insult to injury, that GS550 had its output rated at 25C.....a completely unrealistic temperature, room temp. Once placed inside a case with the computer running, the power supply will be subjected to temps approaching 40C. And it's a fact that as power supplies get hot, the output they produce derates, or drops in other words. So, a power supply rated for 550W at 25C may drop its true output enough when subjected to temps approaching 40C.....enough to make your +12V rails now only have like 12A-13A per rail....maybe less. (A drop to 13A per rail is only down 3.5A per rail at max output, not a lot, like ~20%, not unheard of at all.)

So, you've got an old design, weak +12V output power supply trying to run a known power hungry video card. Bet that's where your problem is coming from.

Most helpful post of the year? nice write up!!

 

[makken]

hmm... I originally suspected my PSU (considering i got it for $30 bucks on sale), but if that was the issue, wouldn't I see stability problem in all games instead of just select titles? I haven't had any issue with TF2, Prince of persia, or Quake 4 as long as I don't enable AA in game (which turned out to be an issue with Quake4 and ATi's adaptive AA). Dawn of War I and Dawn of War II, which IIRC isn't that GPU demanding crashes the most, and Tomb Raider, which is the most GPU demanding title I'm playing right now rarely crashes.

Nevertheless, thanks for the info, and I will definitely look into it.

 

[SSChevy2001]

With any factory OC card problem you might want to try downclocking to default 4870 clocks 750/3600 for testing. Also IMO your better off without CCC, as ATT is a much better alternative, and it doesn't require .net framework.

 

[vj8usa]

Oddly enough, I've gotten the same issue as you with DoW 1 (technically Soulstorm, but that's pretty much the same thing) quite a lot. Every once in a while (I'd ballpark it at every 30-60 minutes), I'll get a black screen with the display driver stopped responding message. If I alt tab back into the game, it continues to work, so I didn't pay much heed to this issue. I have no such issues with more demanding, modern games. My old X1900XT also had no issues with DoW1 though, so I'm thinking this might be a common issue for 48xx series cards.

As for the PSU... when I first got my 4850, I tried to use it with my 450W Antec PSU (32A, I think, on 2 12V rails), but had all sorts of issues. TF2 ran flawlessly, but Crysis was miserable (single digit framerates on the normal island GPU benchmark, significantly slower than my X1900). Once I upgraded my PSU, performance was completely normal. I didn't have system stability issues though. Maybe try running Crysis (or another extremely GPU hungry game) and see if performance is what you'd expect it to be?

 

[Tempered81]

Update operating system, service packs, system drivers, vga drivers, cats, game patches, directx, utils, etc... This might help.

edit: oh and a cheap-o PSU very well may be your cause also.

 

[makken]

think I might have found the problem - was reinstalling cat 9.4 after using the driversweeper as mentioned above, after I did that, rebooted and everything, I got a windows update message saying new updates were installed, it looks like MS thinks that the latest version of ATi's drivers is outdated

curious, I did a quick update history check, and sure enough windows update downloaded a display driver update 3 times, once after i first installed 9.3s, again right after I upgraded to 9.4, and once again just then. I disabled recommended updates and sweeped / reinstalled the driver and everything seems to be working fine now. I just finished running DoW I and DoW II for 2 hours without any crashes.

Oddly enough, I've gotten the same issue as you with DoW 1 (technically Soulstorm, but that's pretty much the same thing) quite a lot. Every once in a while (I'd ballpark it at every 30-60 minutes), I'll get a black screen with the display driver stopped responding message. If I alt tab back into the game, it continues to work, so I didn't pay much heed to this issue. I have no such issues with more demanding, modern games. My old X1900XT also had no issues with DoW1 though, so I'm thinking this might be a common issue for 48xx series cards.

yup, exactly what I was getting, except that I couldn't recover via alt-tabbing. Tabbing back into the game just gave me a blank screen.