Pulling the power plug fries CPU.

I have an interesting problem here that has me scratching my head. A Gateway tower came into my shop recently. When powering it on I get three long beeps and no POST. We narrowed it down to a bad CPU. When putting in a test CPU the system POSTs. So, I ordered an identical replacement CPU for the system. We got the CPU, installed it, and the system worked fine. During a test of the hard drive my software was giving me some problems so I just pulled the plug. As soon as I plugged it back in, the system would not POST again and would only beep three times. I inserted the test CPU again and the system again powered on. Therefore, I ran the system and diagnostics longer on the test CPU. One last time, I had to decide for sure, I pulled the power plug on purpose. Sure enough, it fried my test CPU. No POST, only three long beeps. That's three CPUs that are definitely fried due to a power interruption on the power supply. My question is this, is it the power supply or the motherboard doing it? I can't afford to guess and be wrong. I don't know what to decide. The power supply does not provide the regulated voltage to the CPU. The power supply only provides regulated 12v, 5v, and 3.3v. The motherboard decides and provides the regulated voltage to the CPU. If the power supply is the problem why hasn't it damaged anything else on the motherboard? If the motherboard is the problem, what about unplugging the power supply is causing it to do this? Strange problem that has me in a difficult position because I don't know if to say power supply, or motherboard. If I'm wrong, I've fried another CPU, wasted more time, and might end up in a position where I can't recover my money on this system.

My gut says PSU, not sure I've come across power spikes (presumably the cause of the issue(s)) being down to the board, I would of thought you'd at least have a PSU tester or meter, or even a spare PSU to swap in for testing purposes ? Given the cost of a tester these days it's a wise investment that can save you loads of time diagnosing mystery issues, but the problem you are facing is a board fault can't be conclusively diagnosed, most of the time it takes a leap of faith, but if the PSU has caused these issues, theres no way to know if it damaged the board in the process, but confirming the power to be 100% is the 1st step IMHO, new testers give digital read outs of the power these days, hope this helps.

Ken, As I was reading, I anticipated for sure that you'd experiment with that third CPU. I almost held out my hands palm first to the monitor, in a gesture of "No, don't do it"… but of course I had no actual "reason", and may have done the same in the same situation. This is likely to remain a puzzle, and may only have a confident solution by replacing PSU, MOBO and CPU, and checking to assure that there are no shorts to the MOBO via mountings. Bummer. But I do have a question about "pulling the plug". The question is "how". The question doesn't seem to be an issue that would be readily detected by a PSU checker or multi-meter. The machine did operate. But failed at the moment of pulling the plug. Therefore, the question of "how'. You might have simply shut-off power by throwing a switch on a surge protector strip. You might have pulled the three-prong plug out of the AC outlet. You might have pulled the plug directly from the PSU socket on the tower. Only the last possibility raises any thought for me. Is it possible that the socket on the PSU has a fault that is shorted when the plug is abruptly pulled? Hope you find an economic solution. Doug