Supermicro G34 dual socket dual psu problems

[PoopsPoops]

I'm trying to do a dual psu build for some dual socket and quad socket AMD G34 Supermicro boards.
I have 160W Pico PSU plugged into the ATX 24pin and my idea was to power both the pico from the server psu breakout board and also the mobo's 12V EPS input from the breakout board.

Surprisingly, with this setup the motherboard won't turn on. Some fans spin and I seem to have standby current, but never turns on.

Now, the pico PSU also has a 12V CPU EPS connector. When I power just the pico psu from the breakout board and use the pico's 12V output in the mobo's EPS slot, it turns on just fine. So, this makes me think it can't be insufficient power issue.

I then tried other dual psu setups with the breakout board - typical consumer ATX power supply connected to the 20pin on the motherboard and server psu breakout board handling the EPS 12V, these also fail to turn on the motherboard.

Except, I then tried another dual psu setup but this time with a 220W dell PSU and it does work?!

I'm perplexed what could be going on here and how to get the pico psu to handle the 20pin and the server psu handling EPS 12v?

Anyone run into dual psu issues on the Supermicro boards?

 

[PoopsPoops]

Oh, also to try and isolate the Supermicro board as the problem, I tried the pictured setup (the first one described) on a typical consumer single socket motherboard and it worked just fine. The only difference between them is the Supermicro draws maybe like 300W total but the single socket like 75W total.

Could it be that the supermicro board doesn't want give PS_ON when it senses 12V already on at the EPS plug?

 

[Patrick]

I used SeaSonic single PSUs to power mine. Supermicro, especially in that era, was picky about PSUs.

 

[anoother]

What is the server PSU you are using to power the Pico? I'm assuming it can only provide 12V, hence this setup?

The Pico will be further filtering the power it gets, meaning the motherboard is seeing 2 different voltages for 12V. That's the likely cause of your issues (and could actually cause damage).

 

[PoopsPoops]

The server PSU is a Lite-On PS-2571-5Q 750W PSU rated gold, I believe. It's providing 12V (actually at 12.5V). I thought it might be some issue like this, but with a multimeter and at no load the Pico's 12V pin and the PSU were both putting out 12.5V on the 12V rail, but of course that doesn't mean that was also happening at load when the motherboard tried to turn on.

I don't seem to have caused damage yet as the mobo still works fine with a single psu

I'm getting a power brick for the pico to try powering it from the wall and seeing if that works.

What is the server PSU you are using to power the Pico? I'm assuming it can only provide 12V, hence this setup?

The Pico will be further filtering the power it gets, meaning the motherboard is seeing 2 different voltages for 12V. That's the likely cause of your issues (and could actually cause damage).

 

[pricklypunter]

Probably lacking a common ground. The pico is a switching regulator irc, so if I've got my thinking right here, ground is only seen on the board on the secondary, which is effectively floating, or at least at an offset, compared to the common ground from the atx power supply

 

[anoother]

I thought that at first, but surely the Pico references ground to the input 0V?

 

[pricklypunter]

Easy way to tell, is to simply check for continuity between the 12V input ground to the pico and the output ground on the pico's ATX connector

After having had a closer look, I think I was wrong. It looks more like the 12V input is simply passed through, but used to supply switchers for 5V and 3v3 regulation. Of course if you effectively back feed the 12V that the Pico is using as input, with the EPS connector, it may screw with the power good detector in the Pico, so you might want to stick a meter on that pin and see if you do have a proper power good signal.

 

[PoopsPoops]

Those are good suggestions. I will try them. What I eventually ended up doing that worked (for people coming from search engines) is just buying a separate power brick for the pico. Since the dual psu combination worked with other motherboards, I'm thinking the SM board is picky in some way and the issue likely is no common ground, there was some floating offset between the grounds that the board just didn't like.