My brother’s computer at one point was encountering strange power-up issues. You could power it on and it was completely unresponsive. Sometimes, it would eventually start up but immediately fail after a minute or two. Occasionally the computer could be left “on” for about a half hour and it would eventually start up. Interestingly, my brother reported it to happening only on “cold and windy nights”. During the winter season it almost always failed to boot in the morning, when the computer was off for a long period of time and when it was coldest.
Computers generally don’t fail from relatively subtle differences in room temperature. Considering the age of the house, I suspected that maybe the old wiring was going bad behind the walls and the “windy nights” may have shook the house enough to affect this. So we tried plugging the computer into multiple different outlets, none of which made any difference. Although the outlet didn’t prove that this was caused by power failures, it did prove that the windy nights were not the cause. Nonetheless, I was still convinced that this was a power related issue.
Next up to look at was the PSU. At the time, his was about 7 years old and gave poor voltage output results in BIOS. Seven years is about the normal lifespan of the average PSU, and he had a very mediocre model. I didn’t have any new PSUs on hand compatible with his motherboard, so I didn’t have a way of testing a different product. So, we both agreed it was time for a new PSU. After ordering a new one, I opened up his old one and found a severe burn mark on one of the walls of the PSU. There was so much heat applied to it that the stain was burned into the side wall. The cause was, interestingly, a blown resistor. This resistor was a thick one and was located near the beginning of the circuit. I was amazed to find that the computer could boot up at all with such an issue, so we were both feeling confident we just solved the problem…
…But, the problem was still there. Thankfully, buying the new PSU was not a waste of money since his old one was still going to die out very soon anyway. So at this point it was a process of elimination – take out every piece of hardware one at a time and see if it helps. We started out with what we could easily work with by unplugging all devices, drives, and PCI cards. That did not help. He only had 1 stick of memory, and I didn’t have a spare so we couldn’t test that. When the computer eventually started up, we tried a few CPU and RAM stress tests to see if either would fail. Neither of them did; they were perfectly stable for hours and the CPU didn’t even breach 45C.
Since everything else proved to be working ok, we figured it was a motherboard issue. So, we got a new one with a different memory type (but everything else remained the same). At this point, everything except for the CPU was replaced. I figured the CPU was not the problem because when the computer is running, it runs great and it passed stress tests.
Disappointingly, the new motherboard and memory still did not fix the problem. The computer was almost entirely new and it still failed to boot. All this left us with was the CPU – it worked fine, but it was the only component that wasn’t replaced. Now, it was a matter of how was it the problem. The only variable we left out was coldness. When the computer is running, it usually stayed running because it was warm. The room temperature in the house was maybe around 60F (heating is expensive). Sure that’s uncomfortable for humans but processors can operate at temperatures beyond -150C. So, there had to be more than just simply temperature, so I had one last idea…
Wattage is often a direct correlation to how hot an electronic device can potentially get. With processors, the harder they work, the hotter they get. So my idea was to overclock the CPU a safe amount of about 400MHz, increase the core voltage by 0.2V, and the northbridge voltage by about 0.15V. Success! The computer immediately turned on and stayed on! But the computer was already running and warming up by the time I finished configuring it, so we weren’t convinced. So, we brought the computer to the back porch and left it outside in the below-freezing weather for about a half hour to ensure it was very cold. When we brought it back in and hooked it up, it still turned on with no problems at all. There was only one last thing to prove at this point – was this just a motherboard fluke or could the old board work as well? So we put the old mobo and RAM back in, made the same overclock settings, and it still turned on! This fix was applied around January of 2011 and worked several years later (until the computer was upgraded).
So, how often do you get to hear overclocking actually fix a computer and expand it’s lifespan? Although overclocking proved to successfully fix this computer, it is still a very confusing situation as to why and how the problem occurred in the first place. I would suspect that (a) voltage input location on the silicon die got burnt up a little bit in a possible power spike caused by the faulty resistor in the old PSU. Much like how the old PSU was still functioning with a completely crisped component, the CPU’s defect is at a microscopic level. Due to such a small breach, the extra voltage was probably enough to allow the electricity to arc past the damage and kick-start the CPU. The increased voltage and frequency adds more heat. Heat expands metals, which may have helped close the gap of the damaged circuit.
Since the CPU doesn’t really heat up much beyond 50C (even with the thermal paste removed), the minimal amount of voltage and frequency added to it is highly unlikely to decrease it’s lifespan.