12V DC AT style PC power supply.

It's a fairly crude circuit,but it works for me.
The regulators for the positive voltages,are basically just some 78xx series regulators strapped to some 2N3055's as pass elements.The 2N3055's do get hot,and need to be on a large heatsink,preferably with a fan.Thier respective 78xx regulators probably don't really need heatsinks,but I put a couple small ones on just for good measure. The diodes in the GND leg's of the 7805,and 7812 raise the voltage about 0.7V,to overcome the 0.7V drop of the pass transistors,thus keeping the outputs right about 5.0,and 12.0V. A low dropout +12V regulator might be a better choice than the 7812,especialy if running from a supply below about 15V.Personally,I havn't noticed a problem with it,and have even shorted the +12V 2N3055 with a jumper,effectivly running the motherboard  +12V straight to the battery. The computer was still running decently down to about 11V,at which time I decided to turn it off,and recharge the battery I was using for testing. (A 17AH 12V SLA.) The computer ran for an hour or two from this battery,and during "testing " all day before that.
The 2N3055's are rated at about 115W,I think.If you need more current you could paralell a couple,but be sure to use small value emitter resistors! (this might drop the voltage a bit,but another diode in the 78xx's ground leg could fix that.)

The Negative voltages are obtained using a simple DC-DC converter circuit,and fed to a 7905,and 7912 for the -5,and -12V supplies.The 7905,and 7912 probably don't need heatsinks,but it couldn't hurt.Note the different pinout of the 79xx series! (the heatsink 'tab" is hot!)

The PG (power good/orange wire) can usully be left floating,or connected to an LED,for a visual "OK". Some motherboards may be picky about this?I read that some motherboard may need a +5V pulse applied to the PG pin at power up,in order to boot?
I didn't have a problem,I just connected an LED to it for a visual indicator.

The DC-DC supply came from an old Fluke 8600A DMM.This was a larger "benchtop" multimeter,with a battery powered option. The meter used the DC-DC supply to obtain +/- 15V,from a 4.8V ni-cd pack, for the various op-amps and whatnot inside. The two transistors in the DC-DC supply are 2N4401's,or similar. I'm unsure of the specifics of "T201",as it's potted in a grey plastic case. I assume it's a small ferite E-I core,with the appropriate windings on it. "L201" is a 4-5 turns of approx #24 wire on a ferrite bead.
The 7805 regulator for the DC-DC input supply will get a bit warm,and needs a small heatsink.