Computer power cycles every 3-5 seconds

Hey all,

My computer decided to shut down on me in the middle of playing a game a couple of weeks ago. Hardware shut down - no warning, nothing. This is always worrisome, but it was even worse because when I went to turn it back on, it failed to post, stayed on for a couple seconds, then shut itself down. It repeated this cycle many times until I simply turned the power off on the PSU. I turned the power back on and it continued this cycle. Finally I just turned the power on the PSU off and left for the day. When I returned, I was able to boot the machine normally.

About a week after that, I was once again playing EVE and the thing shut down on me and did the whole power cycle thing again. Remembering the fix from before, I turned the PSU off and left for the day. However, when I got back home, the problem persisted.

Currently, I am unable to get the computer to post on a regular basis with anything more than a single stick of ram (though it doesn't matter which stick and even then it's about a 25% chance). This includes removal of GPU, PCI cards, and HDD. Sometimes it WILL POST, but the Hard Drives are not detected, then then machine shuts down after about 30 seconds or so and returns to the power cycle mode.

I cannot boot from my hard drive in my buddy's computer, but it is a much older AMD based system that originally had a 230w psu. I can however access it as a slave disk.

I just replaced the PSU less than a year ago, and the graphics card about 18 months ago (an upgrade from an x1950pro, which blew the 550w PSU I had at the time).

I'm not entirely sure what the problem is - is the CPU temp sensor freaking out and shutting the machine down? Did the power supply go kaplooey (or about to)? Is my southbridge bad and causing all of the above (this is my #1 guess, as HDD's and PCI cards all cause major fail)?

How can I test these things? I have a solid year of EE under me and another couple of practical EE experience, so I might be able to actually find an electrical issue if it's suggested.

Do you have a multimeter or any tools of that type?

Do you have another presumed good PSU you could swap in as a test?

Any known temperature issues?

Are you overclocking?

Multimeter: Check
Another PSU: I can steal one from my roomate's machine (550w) if he lets me.
Temp: No known temp issues
Overclocking: No. Slight voltage overclock for RAM timings (+.2v? +.3v? it's been 3 years since I set it so I honestly don't remember.)

Well, if you are working on an EE degree and have a multi-meter, dive into that PSU and see if its putting out as specified.

Otherwise, swap in your friend's PSU.

I'd go back to BIOS defaults until you get it sorted out, rather than any sort of overclock.

You say no known temp issues. Is that because you do not monitor temps? Or do your monitors say no high temps?

I do not monitor chip temps. I have a case temp sensor that sits around 35 or so when the machine's operating. I do need a new CPU fan though... broke one of the crappy plastic legs putting it back in after checking the CPU

OK, I reseated the CPU fan. It has enough force and grip with 3/4 legs holding it, it seems. I was able to get the machine to POST with 4 sticks of ram, but I did not have the GPU plugged in. I shut it down, put the GPU and HDD in, but once again it did not detect any hard drives whatsoever. After a minute or so of me coming up with ideas, it shut down again and started power-cycling.

I'm thinking this is a heat-related issue as it seems to only happen after a minute or two of running, then has to sit for a while to work again. I believe the PSU is still under warranty, so I don't want to crack it open, but I can check my friend's in my machine. Might be a heavy BGA on the board giving issues somewhere...