There's not much to it. @dukedevil0 nailed it in his post where the source of the 5v rail was deduced. See https://smallformfactor.net/forum/t...-from-geeek-anyone-have-one.19238/post-284388
I soldered two salvaged cables (atx female pin already crimped) to the spot circled in blue to get the 5v rail. I found it was easiest to snake the cut ends of the two cables underneath the small board and bend them back so they could lay flatter once soldered. I also put a piece of kapton tape over the solder joint as an extra precaution since the cover would be pressing on it slightly (though the cover is painted and also has a plastic flap).
The rest of the cables I de-pinned from the 10-pin connector (3.3v x 2, 12v x 2, gnd x 4, PS-ON, and 5VSB). Combined with the 2x 5v cables. I originally tested them directly in an ATX 20pin connector (omitting one 12v rail) as seen in one of the pictures but settled on rearranging an extension cable as a 12-pin to ATX 24-pin (see https://smallformfactor.net/forum/threads/reducing-wiring-of-the-24-pin-connector.19321/post-285413)
I did notice one oddity since running this full time and on some test systems, sleep did not work for me on two Intel systems (one Asus Z97 with 4790k, another Asus Z790 with a 12700k. I think it may be the lack of PWR-OK/PWR-GOOD but I haven't had time to test further (such as trying to supply 5V on that pin). Sleep did work fine on an MSI B550 with Ryzen 7 5700G. On my Intel systems, instead of sleeping, it seemed like it would try to sleep but then would reboot (and then wake from sleep). I did confirm that both Intel systems could sleep fine with other PSUs I had available with full pins but for all I know it could be an Asus vs MSI quirk. I have a couple of other systems I could try it on but they're coincidentally also Asus with Intel, and MSI with AMD.
Fortunately I don't use sleep anyways, I always leave my systems running 24/7. But perhaps @milesvw has another take on it since he also excluded the Power Good pin but added all the other pins he could try sleep on some systems. If that works, then I wonder what the spec for sleep is looking for...
