I'm looking at doing a wiring project which would connect a SMPS in series to a PicoPSU and PCI-E connector in parallel. The idea is essentially to retain all the benefits of the PicoPSU (less wires, more efficient) while at the same time, allowing a single PCI-E connector for a 150W GPU.

In theory, the idea should work without being too taxing on the power system. The SMPS supplies 12V 25A at a single screw terminal output. My idea is to use copper lugs to run 5 18awg wires off the +12v terminal and 5 18awg wires off the ground terminal. This would be 10 wires total which would fill all the pins on both the PicoPSU connector and the PCI-E.

I guesstimated my total load to be about 225W and the SMPS is rated for 300W, or approximately 30% higher - which is exactly what the manufacturer recommends so as not to be too taxing.

Questions: Is there any current loss by running all these wires in parallel?

Remote sense- The SMPS has both DC output connectors and signal connectors. Remote sense ships with the "local sense" jumper, but if I use RS to increase efficiency, should I wire just one sense to the PicoPSU? Or should I also wire remote sense in parallel? I'm sure one RS would be good, but how about 2?

Finally, 5VSB: there is a 5VSB header pin on the SMPS. I'm wondering what I need to do with this, if anything. My thoughts were that the PicoPSU should supply +5VSB, but should I direct wire the 5VSB from the SMPS to the PicoPSU ATX connector pin? I just don't grasp how these things interface...

I'm guessing that it might make more sense to connect the remote sense wire to the PCIe connector because the motherboard probably does not draw as much 12v current as the video card and the picopsu is a dc-dc switcher, the 5v and 3.3v lines are going to be regulated the same unless your 12v line droops a whole lot.

Both points should be tied together through the motherboard so if there is more ir drop between the PSU and the video card than between the PSU and pico and you sense at the pico, your video card will have a slightly low 12v or maybe cause more current to flow through the motherboard than ideal.

Another option some people might have is a volt meter probe point. I have a motherboard with a special header for testing voltage. I still think the PCIe is probably the best sense point but its an option for some.

Running multiple sense lines is not the way things are supposed to work, the slightly different voltages at each point will cause current to flow through your sense line. That's a no-no.

For the 5v, probably just leave it alone or maybe use it to power an active USB hub.

Hmm, I like your idea about the 5VSB, that would be pretty sweet to wire it to an active USB hub! I could definitely use it, because I've got at least 4, if not more USB devices. Thanks!

As far as remote sense, it sure is a useful function. However I'm not so sure it should be wired to the GPU since my GPU is 150W. In theory, 75W are loaded on the PCI-E side, and 75W loads on the PCI-E bus. That means with a 65W CPU, 75W on the PCI-E bus, and some 5-10W on peripherals there's a total 150W load on the PicoPSU whish is twice the PCI-E, so better to wire RS on the higher load no?

That's all I can think of. I suppose if I were to run a 225W GPU, then it would be better to RS the GPU. I really don't know in this specific circumstance and honestly there's zero information on the Internet about doing it in this way. My gut sense is to put RS on the larger load but I could be wrong. Hmm

I would be really surprised if your GPU was broken up in to two separate circuits that each draw 75 watts.

My assumption is that the PCIe connector ground and 12v lines are electrically connected to the ground and 12v traces on the card edge connectors. If the PCIe cable has less resistance than going through the ATX12v connector and then through several inches of motherboard trace, then most of the current flowing to the video card is going to go through the PCIe cable which brings me back to sensing current there.

I didn't explicitly state that earlier but that was part of my reasoning.

Xbitlabs used to have really good power consumption figures with distribution across connectors, but not any more. The latest I found was for a GTX 580. It draws up to 2 Amps from PCIe slot and the rest from the PCIe power connectors.

The PCI-E connector and edge connectors are connected to the same ground, thus, in theory current will flow in the path of least resistance - through the PCI-E connector. Good observation!

But some current *does* flow through the bus connectors. I was doing some research and I found this:

http://www.xbitlabs.com/articles/graphi ... html#sect0

Xbitlabs used a special setup to measure current draw on the PSU rails: 3.3v, 12v, and PCI-E 6 pin/8 pin. With a full load on Crucis Warhead, a 5750 draws only 1.4A on the 12v, and 3.3A on the PCI-E. But on a 5770 the 12V draws significantly more, 2.3A vs 3.9.

So if you calculate the CPU draw on the PicoPSU, in total there will be more current flowing to the PicoPSU then the 6 pin pci-e. Not by a great amount but certainly more.

What do you think about that? If it was your system and you wanted to prevent voltage drop, would you still put RS on the PCI-E side? Of course, I suppose there's also the additional Vdroop function on the motherboard itself, but still.

Well mostly I think any GPU that draws more power through the slot than through the PCIe 12v plug has a very poorly designed circuit board (more resistance between the PCIe 12v connector and the load than through the card edge, motherboard traces and the ATX 24 pin connector).

It looked like the worst case card drew 4.5 amps through the slot, if yours was equally as bad, that still leaves 8 amps for the PCIe cable. If its like some of the other cards and draws maybe 1.5 amps from the slot, that leaves 11 amps for the cable.

If its 11 amps, that's more than half of your 225 watts and very likely to be the point where your 12 v rail sags the most which is why I think you want to sense there.

But that's just a slightly educated guess, it would probably be a good idea to test several 12v points under load to find where it sags the most and use that point.