GPU upgrade = powering off

[Slaymate]

A Google search will also show that Nvidia mobo's are known to have an issue with the Northbridge running hot. Are you using the water-cooling feature of the NB? I would apply new paste just to make sure it's not an issue.

One of your video cards is running significantly hotter than the other when both cards are installed. It's usually the bottom card that runs hotter because it's blocked in and doesn't get the ventilation of the other card. The paste could be degrading/drying and out and causing problems that aren't noticeable with just one card.

Your Voltages don't look bad but there also not under any strain at all. Try running a Benchmark like LinX or SuperPi and watch your voltage during the test and see if any of them fluctuate. Open HW Monitor and HWMonitor will both keep track of the high and low values so while the system may not crash, we still may see a problem.

 

[gyaku_zuki]

I think its a bit pointelss, but here you are, temperatures and voltages after running a LinX routine:

There is no real fluctuation, but tbh I didn't expect much - with both GPUs sitting at 0% utilization the PSU can easily handle the CPU running at max. The problem only occurs with both GPUs installed and under stress.

I am not using the water cooling feature, and I will consider reapplying new paste but I'll wait and see if the ordered PSU solves the problem before I get to that - would rather avoid taking the entire thing apart if possible, and for the sake of a day or two waiting on the new PSU. I don't clearly identify a temperature reading in HWMonitor that is the Northbridge, downloading Asus' own PC Probe software only shows a motherboard temperature, currently 26 (my room is quite cold atm and its been sitting cold overnight). I'll check in the BIOS perhaps. But none of the temps in HWMonitor seem particularly high to me - eVGA 8800GTX KO/ACS3 chips, from Googling, usually run hotter idle than I have here, and my CPU temps considering the overclock from 2.66GHz to 3.10GHz is consistent with what I have always experienced even with my old GPUs, and as you can see, only just hits 70 on the cores under 100% and sustained load, which isn't there when the power cuts anyway.

In case you are wondering, I'm using an Antec 900 Gamer Case. I've always been impressed with its cooling ability and ventilation.

I honestly do think, on the balance of evidence, the PSU is the issue. I know what you are saying, the 600W should MAYBE be able to handle it. But factor in the age (I ordered it off Dabs and its no longer in my order history which means older than 36 months) and I think its right on the knife-edge. I mean, the thing was even able to have Supreme commander running for 2 hours, it was only the explosions of multiple nukes (and the associated bloom, smoke, particle effects stressing the GPU I imagine) which cut the power. Synthetically, FurMark achieves similar, by putting full load onto both GPUs simultaneously. Re-reading my maths from a previous post, at full load I would estimate my draw to be 580W. So in a hypothetical situation where my PSU can only supply 570W now, only a full load would push me into the danger zone.

Hopefully when the new PSU arrives this will be proven to be the case. Is there any software that would allow a selective load to be placed onto the GPU?

EDIT: In BIOS (so idle load) the Northbridge and Southbridge temperatures are 46 and 43 respectively which is fine.

 

[gyaku_zuki]

Giving you another screenshot, this time a full one. Discovered I could get an approximate 90% GPU load with high CPU load by using BOINC processing using cuda. Screenshot below. Right click and 'view image' to see the full size version if necessary in your browser.

You can see I have also successfully identified the north and southbridge temperatures in SpeedFan, and voltages are shown in PC Probe and HWMonitor. GPU load and temperature is shown in eVGA Precision, you can see that the temperatures remain stable at about 70/60 over the course of the testing. Northbridge rises to 54 during the test, which is perfectly fine.

 

[gyaku_zuki]

Ugh... as an aside, don't order anything from Purely Gadgets... kept me waiting since last Thursday for an 'in stock' item... suddenly there is a supplier problem and my order is 'backlogged'...

 

[gyaku_zuki]

Final post for this thread - I said I'd update and hopefully it helps others in the same situation. New PSU has indeed solved the problem. Relieved and pleased. Thanks for the bouncing of ideas everyone, I don't mean this in an 'I was right' way but I'm glad I was right, as the alternatives were bad :S lol

Thanks again everyone

 

[patonb]

Actually, no said it wasn't the psu, but why not take the 25mins to do free tests to double check things before you shell out any money.

Nobody here really wants to be responcible for someone spending money they didn't have too.

That said, glad the psu did it.