Running two mainboards from one ATX PSU

[Granitehead]

I want to build (buying a finished one is too expensive) a little server to use for stuff like NAS etc.

I looked at different solutions. I especially liked the thought of an extremely low power arm board. But in the end they aren't cheap enough and don't have enough USB/SATA ports (would need adapters, costs again...). So I think it's going to be just another normal x86 computer.

To keep costs low I plan on putting it in the same case as my main computer (a case which I got for free which has previously housed another computer ), modding it a little to fit the second mainboard and lots of hard drives.

To the actual topic: a second PSU would probably fit, but I don't think I'd ever need more power than the one I have provides. I was going to buy a second one, but then I saw this ATX 24 pin Y cable on dx.com and started investigating whether I could use my PSU for both mainboards (because I don't really trust that cable to work and not damage my hardware). Most information about this which I found on the internet was people in forums who didn't seem to really know what they were doing and who just did it by trial and error. Not something I would want to risk, then I'd rather buy another PSU.

Wikipedia was more useful: It seems the lines to which one must pay attention are PS_ON and PWR_OK and maybe the sense line on pin 13: It seems to me that when powering two mainboards, PS_ON should be shorted to PSU ground when at least one mainboards wants to power up and I should connect PWR_OK on the mainboard which isn't supposed to be powered up to PSU ground (thus telling it it can't draw any power from the PSU) and connect the PWR_OK of the mainboard(s) which should power on to the actual PSU pin PWR_OK.

The sense line could probably be a problem, but I'm not sure about that. I just read another Wikipedia page on how it works. So it seems to be about the voltage drop in the cable. I guess I could connect it to pin 13 3.3V power line at the point the Y cable splits to measure the voltage drop in that part of the cable and keep the split ends of the cable as short and low resistance as possible. That would mean the voltage would end up regulated a little too low because not all of the voltage drop would be measured. If I split it like the other cable I think the PSU should sense a voltage drop value somewhere in between the two real values (which /will/ differ when current differs, and current will probably differ most of the time as the two mainboards will seldom draw the same amount of current from that pin), which would lead to one end of the Y cable getting too much and the other too little voltage, which I guess is more likely to damage something than too little voltage on both of them even if the difference is smaller.

It seems to me that as the mainboard power lines would be connected in parallel they would both get the full (correct) voltage and as much current as they want, as long as they don't draw too much, as long as they aren't as sensitive as that 3.3V line seems to be.

Do mainboards react well or notice at all when there is voltage on their (not standby) power lines while they didn't request the PSU to turn on and while PWR_OK is low?

Does anyone here know whether more is involved?

Am I right in my assumptions?

Are there other side effects/risks to doing this?

 

[Linux-SWAT]

Sounds risky.

Aside this, how will you tell the boards to shut down at the same time ?

Maybe PXE boot the slave and let the master doing the job ?

 

[Granitehead]

PXE booting the one which I wouldn't need permanently is what I had in mind.

Shutdown: I guess I would press the normal button on the mainboard or shut it down via software, the mainboard would then shut down and as the last step try to power off the PSU by removing the connection between PSU Ground and PS_ON. If the other one is still running then the pins should stay shorted anyway. I think on this line I wouldn't need to do more than just splitting the wire.

I guess it might be risky, but whether it's really risky and if yes why is what I want to find out.

 

[T.T.]

Here's my two cents.

To me, it looks like you shouldn't have to touch the sense line from pin 13.

For the PWR_OK, I would put a diode in series on each line coming from the board, the diodes should prevent the two boards from interfering with each other.

The PWR_ON line may be tricky. It's function is simple, all the board does is pull it low when it wants to have power, but each motherboard is going to try and control that line individually, so you may have problems there.
I might test this now, I'm curious