I want to make a benchtop power supply from a 20 pin ATX power supply from the last millennium. I have seen posts, but still need help.

So far, I have shorted PWR_ON to ground and have connected an LED to PWR_OK. I connected a 12V fan to a 5 pin peripheral connection. When I connect a CD drive to to 5V, the LED lights up and the fan spins. I measured 5.05V between the CD's supply and ground and 0.55A across it. I took this to mean that I should use a 9 - 10 ohm power resistor (2.7775 watt?!) for a preload.

After removing the CD drive, I connected a 50 ohm 25 watt potentiometer to 5V. PWR_OK comes up (repeatably) only when the resistance drops to below 1.2 ohm. I measure 4.5V across the potentiometer and 3.2A going through it (it gets on hot side of warm). When I try to slightly change the resistance up or down the PWR_OK LED goes out.

What accounts for the apparent difference in behavior between attaching the CD drive and this external resistance? Is this getting distorted by looking at the low end of the potentiometer and my meter?

Despite my abuse, the ATX psu seems to work. Would I have better luck building a benchtop supply using a more recent ATX?

  • \$\begingroup\$ The current draw could be non-constant and erratic, in which case your slow multimeter would not work. You would need an oscilloscope with current probe to graph the current waveform. Doesn't answer your problem though. \$\endgroup\$
    – DKNguyen
    Mar 5, 2021 at 18:42
  • \$\begingroup\$ The difference is that the CD drive also loads both 12V and 5V. Depending on the ATX supply and era it is built (no make, model or picture of ratings provided) it may be more sensitive to other supplies than 5V. \$\endgroup\$
    – Justme
    Mar 5, 2021 at 20:06
  • \$\begingroup\$ you don't need a preload on an atx supply made after 2002; basically anything w/o a -5v rail, also virtually 100% correlated with a 24pin mobo connector. I would get a new supply; late-90s bulk capacitors are notoriously unreliable, and you don't want your voltage regulation failing while it's in-use. \$\endgroup\$
    – dandavis
    Mar 5, 2021 at 20:25
  • \$\begingroup\$ @justme: FWIW, I only connected 5V and ground to the CD drive \$\endgroup\$ Mar 6, 2021 at 14:21

1 Answer 1


The apparent difference is that the CD drive loads both 5V and 12V supplies, while your experiment only has the fan on 12V and variable resistor on 5V.

It may simply require more loading on the 12V rail.

And thus it just takes a lot of experimenting what kind of dummy resistor loads should be placed on 12V, 5V and 3.3V output rails and in which proportion.

Not all ATX supplies will work well as a benchtop supply. Since it is an ATX power supply, power must be drawn from ATX supplies according to ratings specified in the ATX specifications. For example changes in load current must not happen instantaneously but with a certain limit how fast it changes, so that the supply regulation feedback loop can react to it without too large overshoots or undershoots in output voltage. Basically the best option would be to build a buck/boost converter which takes in power from ATX supply according to ATX specification and provides an abusable outputs for lab experiments.

Therefore, a dedicated lab power supply works better than expecting an ATX supply to work under non-standard loading.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.