GPU upgrade = powering off

[gyaku_zuki]

Hi everyone, new here to ask for some advice.

My current specs...
Striker II Extreme motherboard
Intel C2D e6750 2.66GHz running @ 3.10GHz
2 x 2GB DDR3 Corsair RAM
2 x nVidia 8600GT XXX XFX graphics in SLi
2 x Seagate Barracuda 7200.10 500GB HDD
OCZ GameXstream 600W PSU

For power purposes I also have:
Thermaltake Flower CPU fan
DVDRW optical drive
Floppy drive
Linksys Wireless G PCI card
Total of 5 x 80mm fans with LED and 1 x 200mm fan without LED

Everything is around 2 years old, apart from the motherboard and RAM, which are about 1.

Recently, I got 2 used eVga 8800GTX KO ACS3 nVidia cards from eBay. I know, sounds dodgy but bear with me. So my aim was to swap out the 8600s for the 8800s.

Did it and fine, not a problem - system was grand. But, last night I was playing some Supreme Commander, maxed settings, and after about 2 hours the PC just powered off. Not a restart or a blue screen, just off. Now, I had SpeedFan logging (I often build PCs and temperatures, especially on these big 8800GTXs, is something I watch) and the temperatures seem fine, the processor is around 35 idle 42 under load, and the GPUs are 60/50 idle, 70/60 under load (primary card / secondary card).

I can replicate the power down with FurMark, where the power down occurs in about 2 seconds. Using a measuring power plug, I read about 320W idle. Unfortunately the refresh rate on the power plug isn't great and all I can see when I run FurMark is a spike to about 480W before the PC shuts down.

I'm going to try with only one card in later but for now I wanted your opinions - power issue right? A shut down (rather than a blue screen or reboot) would only be caused by either heat or power, am I correct? So, if confidence is high that heat is not an issue, then that only leaves power.

The OCZ 600W PSU has been solid but it is 2+ years old... capacitor aging maybe putting me just under at peak usage?

Just curious because if the GPUs are damaged then I need to return them, but otherwise I'll just buy a bigger PSU.

Really appreciate the help, thanks.

 

[juampa87]

hey!

In My opínion you are just asking too much of a 600 Watts PSU, with two years of ussage for two monster cards...

in my guess yes, it's a Psu problem, i just got an OCZ Fatality 750 Watts.. before this PSU i nvere even thought about using SLI

maybe im being a little ignorant, but in my opinion with such a build you have, you're overwhelming the PSU

 

[Slaymate]

It sounds like a power supply but it could be a problem with the power circuitry on the mobo or one of the cards. Download Open Hardware Monitor, it shows all kinds of Voltage and Temperature readings

Open Hardware Monitor 0.2.1 bta download from Guru3D.com

 

[MindoverMaster]

IIRC, 2 x 8800ghtx's can easily be used by a 600W.

I suspect there is overheating. Get and run HWMonitor, give us a screenshot.

 

[gyaku_zuki]

OK so I tried each card individually and there isn't a problem. They run fine on FurMark. Temperatures really aren't going at all high, even when only one is in, at full 100% load they aren't going high into 70s. As for when both are running, well I can't get you a screenshot as it dies in 2 seconds of FurMark. I got the eVGA Precision software, which measures and shows GPU load, and allows fan speed overrides. Even turning up the fans to 100% did nothing, and the power dies as soon as load approaches 100%.

When running one card only, my power meter reads ~210W at idle (0-2% GPU utilisation). Under 100% GPU utilization FurMark, this rises to ~350W. When running both cards I have ~300W idle. Following the math - 100% on one card is increase of 140W meaning 100% on two is increase of 280W, meaning a peak utilisation of 580W, right on the cusp. Factor in the age... you get the impression I am getting I think.

I've ordered a CoolerMaster M850 Silent Pro as it is SLi certified for dual GTX 470 let alone two 8800GTX. Until that arrives, I will keep things relaxed - I mean you might be right MindoverMaster but even before buying I reckoned I would be tight on the power - and like I say, the PSU is 2-3 years old and PSUs do age. Hopefully it will solve the problem, and I will let you know - but I am keen to hear any more advice or insight, even disagreement with my assessment!

 

[Slaymate]

Your only telling us what you want to tell us. Post a screenshot of Open Hardware Monitor and let us see the rest of the info.

 

[gyaku_zuki]

OK I will when I get home later, I'm at work atm so can't right now. I had been running it and that's where I took those temperatures from in my post above, but how does it make a difference if I tell you or post a screenshot? I can't screenshot the temps at the crash. And I've told you all the idle temps.
At idle, my processor cores according to HWMonitor (from memory) are Core 0: 41 and Core 1: 40. Board reads the processor temperature as something like 35. GPU 0: 61 and GPU 1: 54. I have three thermocouples, one near my RAM and one at each of the GPUs, but they aren't thermally bonded to anything so aren't much use. As for NB / SB temperatures I can't recall but they were <50. The point I'm making is, I'm not inexperienced with builds and hardware. I know an excessive temperature when I see one. And I don't see any excessive temperatures. Which would suggest bad card or power. Since each card on their own passes 30 minutes of FurMark without a peep, there isn't a problem with the cards. So when all else is eliminated, that leaves only power.

I'll post later with the screenshot. Thanks for the continued advice.

 

[patonb]

The screen of hardcware monitoir will show your volts as well as temps on all cores, gpus and all that jazzzzzzz.

I screen is easier than writing 15 values.

 

[Slaymate]

Open Hardware Monitor shows us all kinds of information about your whole system

Open Hardware Monitor 0.2.1 bta download from Guru3D.com

 

[gyaku_zuki]

I don't usually used Open Hardware Monitor as it doesn't read a load of my voltages for whatever reason, normally I use HWMonitor, so I include both to keep people happy

Before anybody says "Look at that 80C on CPU and Mainboard!!!" I will point out that a Google search will show that, in fact, this is due to the fact that the Striker II Extreme doesn't use these sensors (Tcase) and HWMonitor doesn't handle this well.