@przem its more a case of, as you say, the feedback loop only being on one of the outputs so the other outputs' regulation for changing loads on them is not that good. If you're needing spot-on voltages you need something better anyway. Usually I've found that while the 5V is spot on, the 12V can be anything from 11 to 13 V, but it's usually good enough for testing anything that wants 12V.
The author Ben Jordan makes this point in his article:
I have seen others converting ATX power supplies to benchtop lab supplies and putting minimum load resistors in them – in my experience this is not necessary since the cooling fan is generally enough of a minimum load, and most modern switching power supplies are designed to run down to zero load anyway.
Many ATX PS designs require some loading on one of the outputs, most often 5V. Usually a 100 Ohm 1/4W resistor should suffice as a standby load. The feedback loop is usually from just one of the outputs, usually also 5V, so other voltages are usually ratiometric to 5V; I have never seen 12V drifting off while 5V is solid.
I'm wondering if more modern PSUs handle this better. I honestly haven't run into this problem at all, but like I said, I'm generally running very forgiving LED strips, as long as you don't push over 6v to them they'll survive.
Most supplies I find are not toooo fussy about loading as long as you are not too fussy about voltages....if you are loading the 5V but not the 12V the 12V line may be a little high. But most supplies are surprisingly good about it, considering the conventional wisdom that switch mode PSUs need a minimum load. Small light bulbs make good base loads and you can use one as a "power on" light.
Keep in mind that many ATX supplies have cross loading requirements. If those requirements are not met, the supply becomes unstable or simply switches off. For example, on many supplies you can't draw a lot of current on +5VV and have no load on +12V, or can't draw significant load on +3.3V without having some load on +12V and +5V, etc. This varies greatly on the particular supply - some more expensive supplies have no dependencies at all. I believe the cross loading requirement is related to the fact that many cheaper ATX supplies use a single transformer inside to generate 12V and 5V.
Would be nice if this adapter had "dummy" load resistors incorporated to keep a minimum load on each rail.
Blog Doing Math in FPGAs Tom Burke 7 comments For a recent project, I explored doing "real" (that is, non-integer) math on a Spartan 3 FPGA. FPGAs, by their nature, do integer math. That is, there's no floating-point ...