Power Supply Torture Test

An anonymous reader writes "With the latest batch of power hungry graphics cards, the PSU in your computer is more important than ever. If you're looking for a new power supply, check out this group test. They've tested 19 PSUs - some good, some bad and some downright explosive!"

Comments

Good, I guess...

[Dragoon412]

I always stress the importance of getting a good power supply when I give advice for building PCs, but it seems like something people are largely willing to gloss over, and just go for some cheap no-name with high wattage. It's nice to see a review for these, finally...

But, of the companies on here, I've never even [i]heard[/i] of Tagan, and it seems a bit sketchy that Tagan wins best honors in their review... which was conducted in Tagan's lab.

Furthermore, why aren't Sparkle or Antec power supplies tested? Along with Enermax, they're widely regarded as some of the best around.

Pity they didn't include...

[hcdejong]

noise measurements. A l33t PSU is no good if it sets up a howling gale in my room.

Damn

[GigsVT]

That page is annoying, it keeps jumping up and down while I'm trying to read it, because of that stupid javascript ad.

It's kinda silly anyway, the article's premise is that they got access to this $20,000 power supply testing equipment... A set of simple $10-$20 high wattage load resistors would have worked just as well.

A $1000 scope might help if you want to catch a load dump overshoot, startup transient, or ripple, but it looks like they aren't even concerned with such important specs of a power supply, specs that could burn our your system.

Anyway, some testing is better than no testing, which seems to be the norm for computer power supplies, so I am thankful that someone with access did these tests, but it would have been more useful if they had tested more than simply steady state load.

Re:Damn

[Linker3000]

I spent some of my electronics engineering apprenticeship in Unit Test where I had to test 5V 300A and 5V 600A power supplies.

We had a FET-based active load for testing and we had to bolt the PSUs to the load with 1 inch copper braid, put the units in a thick plexiglass/Perspex cage and run them at 110% of rated load for 10 minutes.

When we got busy, one of the other engineers soldered up a ton of wirewound resistors to act as a load and fitted them into a spare 19" rack with a large red 'emergency stop' button on the top to cut off the power. The whole rig looked like a McGyver special, but it worked and I wasn't too afraid to use it-but when we got VERY busy, I was told to forget about bolting on the braids, just hold them in place, wind up the active load to 110% and keep it there for 20 secs and 'that will do' - I refused, much to the annoyance of the Unit Test manager who couldn't actually force me to do so!

The guy who made up the resistor load was really mad - it was the era of the BBC micro and he'd lashed up his own floppy drive, but discovered that he couldn't copy one particular game or program because the drive he was using wouldn't step properly to one of the 'hidden tracks' (or something like that), so he bought an official external drive (for some crazy price - it was 1982, after all), did his copying and then hooked up the ground wire of our PSUs to the chassis of the floppy drive and dragged a wire hooked to the +5v line around all the chips' legs - end result was tons of sparks and one very smoky unit which he took back for a full refund. Apparently he complained bitterly to the shop about the way the drive had suddenly 'exploded'.

This was also the guy that fell over and broke his leg walking across the car park one sunny day - we really couldn't work out how he managed it! He also removed all the light fittings from a temporary work cabin the night before it was due to go back to the hire shop - when the hire company came to pick it up, they refused to take it so our boss called us all together and said it might be a good idea if the light fittings returned 'tonight' - lo, next day they were all back in place.

Torture? I think not.

[NoseBag]

First - I spent the better part of 20 years designing military and aerospace switching power supplies and systems. Admittedly these toys were commercial products, but I think I'm qualified to say the following:

a) purely resistive loads are a poor model for PC load characteristics, as are purely constant-current loads. But the CC load is a tougher test.

b) transient line and load response (or lack thereof) can preclude operation at the steady-stete levels tested.

c) I just skimmed the article as it was loading strange, but I didn't note any specific cross-regulation tests. These types of tests may yield poorer performance measurements. They almost always do.

d) Testing to see if a PS will run at full load is not the same as seeing if it will START under FL. I didn't notice these tests. Likewise Starting at low or hi input is tougher that running at same.

e) we ps designers had an expression: Power supplies are like assholes - everyone has one and everyone thinks they're an expert. The former is true (some folks have two :) )- the latter is seldom accurate.

OK - flame away. I'm gonna go eat lunch.