Technically Working

I promise this will be the final post about my PSU. I have already given this too much attention. For those who are interested in the personal experience (which was awful), you can read the previous posts: 1, 2, 3, 4. This is a technical post in which I will details how I diagnosed and isolated the problem, which I had promised to some people. Some background: My PSU crashed completely. It died. I took it to the store for fixing/replacement. While it was at the store, I had a replacement PSU that I got from my workplace (for which I am thankful). During that time everything worked perfectly (this was an important detail for later on). After a month, I got my PSU back from the store. That's when things started to get really frustrating.

First Wave

The PSU crashed during games. It took a non-constant amount of time to crash (15-30 minutes on average). The crash was a shutdown of the computer for 5 seconds, followed by a reboot. My main suspect was, of course, the PSU. However, since it just got back from the store, I had to make certain. Also, because it only crashed during games, the graphic card was a very likely suspect. So to isolate the problem, I conducted a few tests.

Cuddly Muscles

The first thing I wanted to do is to rule out the graphics card. I had conducted a stress-test on the graphic card using FurMark, which stresses the graphic card to its limits using fur rendering, and is usually used to check overheating in overclocked graphic cards (my card is not overclocked). In my understanding, FurMark does not stress the CPU. It ran for 40 minutes without crashing. So while the graphic card reached its highest working temperatures (ever), it did not give in.

Think and Remember!

Next thing was to rule out the CPU and Memory. This also called out for a stress test, and the tool I that I used was ORTHOS, which stresses both the CPU and the RAM using some heavy FFT calculations. I let it run for two hours. It did not crash.

Balls of Fire

So now that I had ruled out the CPU, RAM and graphic card, I wanted to find a controlled way to crash the computer. Playing games and waiting for the crash was out of the question. I downloaded an nVidia demo called GeoForms. This demo utilizes both the CPU and the graphic card, although not as extremely as the stress tests I used for them separately. GeoForms uses the CPU for procedurally modifying 3D models and the graphic card for rendering them in real time. However, it was enough to crash my PC in less than 20 minutes on average. 

Conclusion

There could have been other things causing the crash. It could have been a faulty motherboard. However, given that I had just replaced my PSU, and that while it was in repair the replacement PSU (which had lower wattage) worked flawlessly, I placed my “bet” on a faulty PSU. Given the events that followed, it turned my bet was solid enough.

I hope this post can help someone who is experiencing a similar problem.