The thing about testing hardware these days, is that it is very complex. There are all manner of problems which can be difficult to detect by testing with software like these.
I tend to assume that if a fault is found, it is genuine. If I don't find one, I don't assume there isn't one there. You can really only know if a part is bad. You can't always (ever) be sure if one is good. Not great news I know, but thats the reality I'm afraid.
It will depend the component to an extent. Hard drive faults are usually shown up pretty easily, though you cannot rely on the built in SMART diagnostics. The trick is working out when what look like software or directory errors are being caused by hardware (even then it isn't always hardware in the drive which is responsible.
RAM, logic boards, and CPUs are the ones which are trickiest. The tests simply give different components tasks to do, where the diagnostic knows the correct output from the input it provides. A component will only fail a test if it does not provide the output which is expected. I would go so far to guess that a certain level of inaccuracy is tolerated too.
Often the only way to test these components is to swap them out and see if your symptoms persist. I had a PowerMac G5 the other day which would kernel panic with alarming regularity under any sort of CPU load. It also tended to just freeze sometimes. It passed all tests I ran. Apple provides its engineers with special diagnostics for each model of Mac. I imagine these are essentially the same as the included hardware tests, but they give more detail when errors are detected. Hardware test tends to give a non-descript code which can be deciphered by a technician when you take it for service or possibly call Applecare. The aformentioned G5 stopped kernel panicing after some experimental RAM swapping and I was able to deduce a faulty RAM chip from that. It went on to persist with its freezing symptom, and is waiting on a new CPU. It could well turn out to be the logic board, and there is an outside chance that a weird fault in the power supply could be to blame, but I have no way of knowing without testing one by one. I was able to eliminate a drive issue by disconnecting internal drives and booting from firewire. Symptoms still occurred.
It is customary to run several hours of these tests on loop. Sometimes a part only fails once in a large number of loops.
Occasional isolated crashes or odd behaviour are not unheard of in perfectly healthy systems. Its only when a fault is reproducible that you can really hope to pin it down with a degree of certainty.
Top
