Testing advice needed: Is my power supply still safe to use?

[DeathRayLoveMachine]

I have an Antec EarthWatts 650 which I picked up a few weeks ago. The first system I used it in, I had a new video card plugged into the molex connectors (via a PCI-E to molex converter that came with the card). That video card died (total electrical failure) after about a week.

I had other problems with that system, and it was getting near the end of its lifespan anyway, so I spent money on all new components when I replaced the video card. I re-used the same power supply as before.

The case I bought came with front-mounted fan controllers. When I set everything up and powered on, the system worked for a few minutes and then one of the fan controllers exploded (popped/sparked/smoked).

I've been testing the power supply since then. I have it sitting out on a table, with a paperclip connecting green and black pins on the ATX connector. I only have a crappy analog multimeter, but with no load at all, the voltages are stable and within spec on all circuits and rails. The power supply appears to be fully functional, in spite of withstanding two catastrophic electrical failures on the same molex circuit.

I'm still a bit leery of this PSU. Is there anything else I can do to test it before putting it into my new system, or I am just being paranoid at this point?

 

[laserhawk64]

You need to give it a load to test... hook it up to something with a hefty current draw, that is utterly expendable... e.g. a set of several 120mm fans or a really ancient mobo.

Note that if the PSU has a 24pin ATX connector, you'll (probably) need a 24p->20p ATX adapter cable to hook it up to an ancient mobo, but those are dirt cheap and everywhere.

Again, you want something utterly expendable, but with a rather large draw on the PSU. (If you go with the fans, use a large number of them and make sure that they are as cheap as is humanly possible -- cheaper generally means less efficient means more power draw.)

All of that said, it does tend to sound like the PSU is major toast. I'm just suggesting how you find out for sure.

 

[DeathRayLoveMachine]

You need to give it a load to test... hook it up to something with a hefty current draw, that is utterly expendable... e.g. a set of several 120mm fans or a really ancient mobo.

Again, you want something utterly expendable, but with a rather large draw on the PSU. (If you go with the fans, use a large number of them and make sure that they are as cheap as is humanly possible -- cheaper generally means less efficient means more power draw.)

This doesn't jive with what I think I know about the physics of component failure. Adding load to the PSU should decrease the voltage on each circuit (voltage sagging); undervolting like this might cause system crashes or contribute to the early death of the power supply, but it wouldn't account for my video card dying or the fan controller exploding.

Line noise might account for the video card and fan controller failures, but the test you're suggesting wouldn't be able to measure it: line noise is typically caused by varying the load on the PSU, and just running a bunch of fans at maximum speed wouldn't simulate that. Also, it's really hard to measure line noise with just an analog multimeter. (The needle is a physical object that needs several milliseconds to cover a visible distance; pulses of line noise lasting less than a few milliseconds would effectively be invisible, because the needle couldn't respond.)

All of that said, it does tend to sound like the PSU is major toast. I'm just suggesting how you find out for sure.

There is an alternative hypothesis: the fan controller exploded because it was utter crap (it really was crap, just trust me), and the video card failure was unrelated and coincidental. Right now I'm actually leaning towards the alternative hypothesis.

 

[laserhawk64]

You're not looking for noise, if it exploded a fan controller. You're looking for the psu to fart out a big surge under load... think overheating/defective voltage regulator or (maybe) even a sporadic short in the transformer.

You'll only find it, though, when the contraption thinks it's being useful. Think of it as putting Murphy's Law to good use. Just give it a little stress and let it go all wakkawakkawakkaPHTHTHwheeeeeeeeeeeeeKABOOM

That's what I'd do, anyhow...

 

[DeathRayLoveMachine]

 

[Meghan54]

Software monitoring is utter crap, can and is usually wrong and not to be believed.