Hardware Insights Forum

The fan (branded T&T, maybe short for Thieves & Thugs?) was seized up and the PSU was horribly damaged by the heat and all of the caps were vented. The insulation on the secondary side wires and on the torroid coils is also hard and brittle. After removing the caps, you can clearly see that the PCB is almost black from the heat, and pulling the blade housing off the fan showed why it failed - it was rusted. Amazingly, it still runs, but with lots of ripple due to the failed caps

Whoops...
That PCB is so burnt!

I have replaced the caps, one of the torroid coils, the fan, and the damaged wires. I haven't load tested it yet, but I paperclip-started it, and all of the voltages were in spec.
The caps used are

12V - 2200uF 16V Panny FM + 1500uF 16V Nichicon HE
5V - 2200uF 6.3V Nichicon HE + 1500uF 6.3V Nichicon HE
3.3V - 3300uF 6.3V Panny FK + 1000uF 6.3V Nichicon HE
-12V - 1000uF 16V Nichicon HE
5vsb - 2200uF Samxon GD + 1000uF 10V Panny FK

I'll post the resluts of the load testing as soon as I get the load tester out.

I have load tested it. With 10A on the 5 and 3.3V rails, 19A on the 12V rail, 0.1A on the -12V rail, and 1A on the 5vsb rail (just over 300W in total), here's what we got.

12V: 11.83V, 40mV ripple
5V: 4.99V, 16.2mV ripple
3.3V: 3.34V, 15mV ripple
-12V: -12.09V, 63.4mV ripple
5vsb: 5.03V, 15.2mV ripple

I think that's quite a good result for a PSU which was just about fried whien I got it.

not bad at all!

I finally put the PSU into a PC (a Pentium E5300 with 7300GT Grapics). It works fine. You'd never know it was previously fried.