ISO 2400-2600 watt CRPS solution (specific questions about FSP-FC250 PDB)


(please don’t stop reading when you see “ATX”)
I’m looking for something and I also need a little help understanding what I might be looking for. What I do know is that I need to power 4 x RTX 4090s, and that they will almost certainly have to be connected to a standard ATX or E-ATX MB inside a standard ATX case (and, of course, requiring ATX voltages). I’m not concerned at all with physical limitations (I can engineer a mounting solution) but everything else has to be right. And I need this to be a redundant power supply solution as well. What I have managed to find is the CRPS standard by Intel that might actually be the key to solving my problem. The only issue is that I have tried contacting several manufacturers for information (FSP-Group, Bel, RF-point) and it has been like pulling teeth with extremely slow single line responses if I get any thing at all. Now if I understand this CRPS standard correctly (and I have no idea if I do), then I SHOULD be able to plug any 2 x CRPS power supply into this FSP-Group PDB, FSB-FC250, and with the proper wire harness, I should be able to power my 4 x RTX 4090 build and if one of the two CRPS dies, the other should take over. But there are a few things I don’t understand and I’m hoping somebody can help me…

  • The rated power of the FSP-FC250 seems to only be 2400 watts. I don’t know if that’s enough. It would be right at the border and maybe a tiny bit beyond. BUT it’s the ONLY CRPS PDB that I’ve been able to find that looks like it has ATX out
  • That 2400 watt limitation of the FSP-FC250…I don’t know how much I can fudge that. Could I pop two 2600 watt CRPS power supplies in there and it function and be safe?
  • Where in the world do I get the wire harness for the FSP-FC250??? The specsheet states, “Wire Harness is NOT included in standard shipping” and yet there is no mention on where to get it, or any part number, or anything.
  • I’ve linked the PDB below from the Digikey website. If there are any solutions in the digikey store that will solve my problems (or anywhere else, failing that), you would be helping to shatter a wall that I’ve been stuck on for the past few weeks at this point.

Thank you so very much!

Welcome to the Technical Forum. I have requested additional help on this post. We will let you know when there is a response.

1 Like

Thank you and thank you!! :slight_smile:


The FSP-FC250 unit mentioned looks like a cage assembly designed to accommodate a few CRPS supplies and provide ATX-like output connectors. There are off-the-shelf rackmount chassis available that appear to accommodate such things, and which may well include that same unit with purchase. The island from whence many such things issue forth is not a large one.

Insofar as it’s a device made to provide 1+1 redundancy, each 1 would need to be rated to carry the full rated load in case of a single failure. What happens if the load decides to ask for 1.1, given that there’s an available capacity of 2 in the normal state? Fair question. I’m not familiar enough with the standard to know whether any distribution equipment is expected to provide any monitoring or over-current protection.

Current carrying capacities of the interconnects are likely to be a limiting factor, considering that the rated output current on the 12V rail is on the same order as that used for single-pass welding of 1/4" steel…

This also bears on the question of cabling. The attached document (source link) from Intel contains a lot of useful info regarding connectors, pin assignments, and such in relation to supply cabling. I’d infer that the connectors marked P2 on the FSP-F250 are of the 12VHPWR sort, though they’d appear to lack side-channel connections, which may throw a wrench in things. I’d assume there’s documentation (somewher, in some language) on the FSP side of things as to what some of the other connectors are; perhaps a wrench removal tool would be found there.
336521_Rev2p01.pdf (2.7 MB)

As for “redundancy” a person ought to think carefully about what exactly is meant by that. Being serious about the matter would mean running separate 240V/15A branch circuits from the service panel, each with their own 3kW+ UPS. (And make sure to do routine maintenance on those UPSs by the way…)

I don’t have a more complete answer to offer, but perhaps this will help you move the ball past the chain for a first down…


Thank you very much for the information! Every bit absolutely contributes to the puzzle and helps get me to my goal. So I’m definitely appreciative!

A thought though, or more truthfully, a question…don’t redundant power supplies have some mechanism that decides which power supply provides all the power, while the other power supply draws very very low power until the other one fails? If that’s the case, would one really need two separate 240v/15A circuits with one for each power supply? And for the same reason, wouldn’t you also only need a single appropriately sized UPS if only one of the two power supplies is drawing the full load at any given time? I certainly can easily take care of having two separate 240v/15A circuits if need be. That part will be easy. But an additional UPS would be an unexpected expense that might not be cost effective. So what you said is a little bit concerning considering a single UPS could cost a quarter of what this entire build will cost haha. If you could elaborate on that a bit, I would most definitely appreciate it!

Thanks again!

Generally no; a typical redundant supply scheme attempts to distribute the load evenly across all participating elements, with sufficient reserve capacity to support the load in case of a single-element failure. Stress factors contributing to failure often have a non-linear correlation with temperature (see “Arrhenius equation”) so best reliability tends to follow from distributing rather than concentrating the burden.

It depends on how serious one is about the redundancy idea, and what one desires to protect against. Feeding a pair of redundant supplies from the same branch circuit or UPS still leaves one with a single point of failure, just a little further up the supply chain than might have been otherwise. If one considers the very real possibility of a failing supply triggering over-current protections on its way out, the case for redundant UPSs and branch circuits becomes clear pretty quickly. Same for the realization that a UPS is itself a device capable of failure.

Redundancy is costly by its very nature and definition. But if one isn’t careful about spending enough and in the right places, it’s not hard to end up worse off than if one had skipped it altogether…


Well what do you know, you learn something new everyday! I will take these extra costs into consideration for my financial projections on calculating an ROI and properly balance redundancy and reliability with what I can expect to earn. I’m trying to build a system with VERY high reliability so that my hardware will be more attractive, but maybe it isn’t worth it. I’ll have to make some considerations and I’ll definitely be looking much more into the idea of a properly redundant system.