Question How can I get a more stable voltage out of an ATX PSU?

May 19, 2019
I'm starting working with Arduino projects, and in my actual project, I need to run a few servo motors.

I, like many others, turned an old ATX1 PSU into a "bench" power supply (since Arduino can't provide enough current to power all the servos).
I connected an 8.2-ohm 10-watt resistor in the 5v rail of the PSU (which is the one with the most current available) and then tested it.
I noticed that in the 3 rails I connected (3.3v, 5.v and 12v) I read a slight but continuous voltage fluctuation, around +/- 0.02 volts.
I thought that this would be insignificant, but when I connect my servos to the 5v rail, they jitter a lot, they work in a very unstable fashion.
If I connect them to the Arduino's 5v instead, they work like a charm.

So... is there any way to make this ATX's Outputs more stable?

Or is it that the problem is another one that I'm not seeing?

Thank you all in advance for your attention, and excuse my poor English T_T
ATX 1 standard?

Just to say, cheapo PSU's do not have good voltage stabilization , they depends on load, or they are over volted without any load, which I have one from MS.
Maybe 30-40W without load the 12V rails are around 12.5V, on full load overclocked is 11.9V.
Also oscilation, aka ripple which happens under loads on cheap PSU's.
May 16, 2019
With cheap power supplies, there's not much you can do to help with poor electrical performance.

Some use a group regulation scheme at the output. For example, the 5V rail is regulated with the 12V.
When there's a load imbalance, voltages can go out of wack.
When the rails are balanced, voltage regulation is usually fine.

I suggest trying to equally load each rail simultaneously.

If it's a problem with oscillation/ripple, you could recap any dead or blown caps in the unit if you know what you're doing.
Do not go digging in a PSU without any electrical knowledge... It's dangerous.
That old ATX1 PSU may have a high level of ripple (the voltage fluctuation that you're seeing) usually due to insufficient filtering of the DC output rail. Adding a large value electrolytic capacitor across the +5V rail, that you're using, should reduce the ripple.