I used to have crossfired HD4870X2 on which VRMs tend to overheat to 120°C and then system just shut down, monitor went black and PC off, after few secs it started booting up itself.
I managed to "solve" that by mounting powerful fan (100CFM, noisy as hell) to the side of primary graphic card - it was very noisy, but I could enjoy my games.
Now I faced opportunity to upgrade, so I sold both HD4870X2 and installed one Sapphire HD5970 4GB, hoping to get rid of the noise and gain a bit of perfomace. God I was wrong.
It was all nice etc. but from time to time I experienced the same - sudden shutdown of system out of nothing, rebooting few secs after. Then I found out I had turned my 3x120mm front case fans to minimum, turned them up and it didn't happen since then.
Yesterday I installed Lost Planet 2 looking for some graphical orgies (couldn't get to run DX11 - I just got "... stopped working" when menu was supposed to show up), ran DX9 mode and started mission - system shutdown during intro, I've manually set fans to 100% before starting game, GPU was 38°C, CPU 22°C.
I've tried using Everest and MSI Afterburner for monitoring - GPU temps are not a problem as it seems. I've also used MSI Kombustor and I haven't been able to reach shutdown state neither, Furmark same result.
I've launched Lost Planet 2 then and let Everest do the logging, played until shutdown (cca 1 min after start of mission):
GPU1/2 load: 99%
MB temp: 40°C
CPU temp (4 cores average): 30°C
GPU1 temp: 58°C
GPU2 temp: 60°C
GPU1 VRM temp: 68°C
GPU2 VRM temp: 91°C (?!)
GPU1/2 VRM voltage: 1,18V
now a little strange thing probably:
GPU1 VRM current: 40,45 | 60,06 A
GPU2 VRM current: 52,71 | 38 A
those last two readings in my log before shutdown are separated by 5 seconds, why sudden drop of amperage?
Do you think it's my graphic card overheating? I mean, it's brand new, no OC done, I'd really like to know, how come it could be overheating? I have huge case Lian-Li PC80, 3 fans front, 1 exhaust top, 1 exhaust back, 1 exhaust on PSU, the card itself has 3 fans. And how come I had the exact same problems with dual HD4870X2s, where the powerful fan helped? (I can't try the fan now, because with my PSU replace I lost one of my "experimental" rails I got on the old one and cba to mod with this new one)
The PSU is also brand new, it was replaced along with card and imo should be sufficient, I have only 2 HDD + system SSD one.
Something to check.
I took a look at the picture of your psu.
It identifies what power connectors are on each rail.
You have 4 rails of 20A each.
You don't want to use two pcie connectors from the same rail, because that would not be enough amps. http://www.newegg.com/Product/ImageGallery.aspx?CurImag...
Use 1+3 or 2+4
Just don't use 1+2
This is more common than people think even low end cards have this problem and the only solution is to mod the card for better cooling. Usually that means getting new thermal pads or going to extremes as I and others have done. You can try the fan trick as you have done with your old setup and power consumption is down from your old setup which is good.
hmm, I'm using PCI-E1 and PCI-E2, only 2 8pin PCI-E connectors aviable, leading straight from PSU (cannot be detached, it's on 2 separate sets of wires), detachable PCI-E power cords end with only 6pins. I'll try to get my hands on some other PSU, but I guess it won't be easy, since no1 around posseses such powerful PSU :>
That's why I bought card from Sapphire, they have their own cooling, I placed trust in them and hoped it's gonna be effective, yet it seems it isn't. I really don't want to use fan trick, since it's very noisy :S I've tried the same with 120mm Nexus fan, it extended time before shutdown for like 5s :S http://www.cttl.cz/clanky/Novinky09/kubis17/Duben/Sapph...
They make 6-8 pin adapters, I know, I just bought a couple.
That way you can utilize two different rails.
If the power draw is to high on a rail, it could act just like a underpowered psu. Things start getting inefficient and hot as the load increases. http://www.newegg.com/Product/Product.aspx?Item=N82E168...
edit: You might wonder why they configured the rails they way they did. Well if you had two high powered video cards that each needed 1 6 pin and 1 8 pin for each card. Then each video card would tap in to two different rails.
In your case you have one very high powered single video card , with two gpu's on a pcb.
So the need is for 2 8 pin. Which is not ideal to pull from your psu.
I could be wrong, but its worth troubleshooting it out ?
Underclock your card by 100mhz across both gpu's see if that allows it to run. It might give you some clues.
I have those reductions from Arctic Cooling Accelero Xtreme coolers, so that's no probs, but I have the 6pin cord at home :> I'll try it out tho.
I used to have dual HD4870X2, which was 6+8 pin each, and it still happened.
Nothing is wrong with your psu so I consider that not being a main factor but you can check the volts on the 12v rail. If they are low less than 11.5v then you got a problem. But if more or less 12v even give or take .2v then you have no problems with power. If you are unwilling or unable to physically mod the cooling on the problematic power vrm then RMA the card.
I've grabbed 1 line from home during this weekend and also found out that my PSU has rail 2 and 4 capable of 25A (1 and 3 is 20A, the model is Fortron Everest 1010W 80PLUS), so I connected my graphic card to those 2 strong rails and it works!
Thank you guys for assistance, especially notty22!