Skip to the end for useful advice, the rest is more philosophical rambling on the art of computer diagnostics.
A lot of us here have piles of old parts lying around so swapping in known good components isn't a problem. We usually point out the most likely culprits, the simplest to test, then the least expensive and keep moving down the list to save time and money. Without having access to spares, you can still test things but it's more complicated.
Let's say the PSU is flaky, sometimes it starts and sometimes it doesn't. But you don't know that so you try checking the RAM. You go from 2 sticks to 1 stick, you try it in every slot, then repeat with the other stick. Voila, it finally boots. You shut it off and try again no problems. So you think you've identified a bad RAM stick and slot, but in reality the half dozen boot attempts warmed up the PSU enough to get going. Then you shut it off and the next day it won't turn on.
There's a slew of diagnostic methods for each component, but each can give false positives if the problem is somewhere else. A CPU might overheat because the cooler is at fault, or because the mobo voltage regulation is screwy. You shove a beefy watercooler in and get the temps down, but it's still getting overvoltage and fries itself after a month.
Anyway, you said it's dead until you unplug and plug it back in. Does that mean it starts up ok afterwards? Or that it just tries again for 2-3 seconds then dead? If it does eventually start, it points towards a PSU because they can go bad slowly. Other types of component failures are more binary, they work or they don't. That's not always the case but it's a process of elimination.
Facebook
Twitter
Instagram