PC rebooting (semi-)randomly

That's an interesting problem.

I can theorize I guess.

1) The board could be making voltage adjustments under load that it is not making at idle. The adjustments could be to the CPU or RAM. As a result the system becomes more stable under load. This would point to either MB hardware or incorrect default BIOS settings.

-The first possible solution to this is the easiest and cheapest. Update the BIOS.
-The second test would be to use only one stick of RAM. This would lower idle voltage demands...another easy attempt.
Along this same vein you would run CPU-Z and make sure the settings for the RAM in the BIOS match what you find in the SPD tab. Take any RAM settings off auto and enter them manually where you can.
-If the actual voltage regulation of the board is messed up you will need to RMA it. That would be the last thing to try on my list.

2) Bad PSU. Power supplies must make adjustments for different loads and yours might be fine at say 120W, while at 80W it is struggling. Not unusual actually, although usually they just get very inefficient. This could be influenced by the temp of the PSU, which is then dependent on the temp of the room. A cooler room would result in fewer reboots.

-Be Quiet does not make PSUs. Like most companies, they buy from the makers. As I recall that unit is ATNG... heh nope. It's FSP, and not one of their better offerings. That's going to be a prime suspect.

I doubt the 12V rails are really separate, but it looks like there is a dedicated rail for the motherboard. That might explain why the GPU stress is not impacting stability. It's only the 12V1 rail that is causing issues. That's the rail attached to the MB.... but again, I don't know the actual architecture of that PSU. I will call in an expert  

Thanks for the ideas, Proximon. I'll try them out and post the results.

A simple way to check if a "multi-rail" PSU is truly a multi-rail PSU is to disconnect it completely from the wall and the computer. Then use a DMM to measure the resistance between each yellow wire and every other yellow wire in both directions (polarity matters).

If the readings are all low and similar, you really have a single rail PSU.

That makes a lot of sense. I would be fairly confused trying to puzzle it out from the internals I think. Some time soon I'll have to go find some old ones and break them down. Have to finish my basic electricity reading first.

I decided to just buy a new PSU: a Corsair HX850. So far (after two stress tests and two days of normal use), it seems to have fixed the problem. I will confirm in a few days.

^ Well to recommend a cheaper but still good PSU for you to save 20$ is the XFX 850. Same 80+ certified and modular just the same. Plus has great reviews and it's cheaper.

http://www.newegg.com/Product/Product.aspx?Item=N82E168...

Reviews: http://www.hardocp.com/article/2009/12/10/xfx_black_edi...

Thanks for the suggestion, but it's been bought and installed already.

Awww ok then.

Still no problems since the PSU swap. Thanks again to Proximon for helping identify the problem as the PSU.

That's great news. The stranger the PC problem the more likely it's the PSU. It's a maxim.