Discussion Reducing wiring of the 24-pin connector

[protocolsix]

The J-HACK M2426 is a super elegant solution to reducing the unnecessary wires, but seems be out of stock at the moment.

In the meantime, I am wondering if it is possible to make a "poor man's" version without any PCB just by wiring the minimally necessary pins in a standard 24-pin connector.

The idea would be to use this cable with an HDPLEX 250W GaN PSU as well as the standard 8-pin CPU EPS connector to supply +12V.

It was suggested this is possible with only 9 wires:
1x +12V
1x +3.3V
1x +5V
1x +5VSB
1x PS_ON
1x PWR_OK
3x Ground
(total 9 wires)

This thread indicates that some (not all) motherboards require -12V to boot (although apparently some can be fooled by ground), but I think it could be probably be omitted entirely. It seems the M2426 doesn't generate -12V.

The HDPLEX 250W GaN is rated to 10A on +5V and +3.3V and 5A on +5VSB, but my understanding is actual +5V and +3.3V draw is way lower in modern computers so I think a single wire for each of these should be OK. (Notably M2426 supports max 6A on these rails).

According to the molex spec, pins of a 24-pin PCB mount are rated to 5.5A to 8A depending on the wire gauge. So I think it would make sense to use thicker wire gauge for the higher current pins (Grounds, +12V, +3.3V, +5V, +5SB) and thinner gauge for PS_ON and PWR_OK.

I have a few questions:

• Does the motherboard even require +12v from the 24-pin connector if an 8-pin CPU EPS is also used?
• If it does, I am somewhat concerned about a single wire for +12V into the 24-pin motherboard connector. Although apparently almost all draw for +12V actually comes from the EPS connector so it is fine?
• How many grounds would actually be required? It seems reasonable to have roughly the same number of ground pins as rails that draw any kind of current, but assuming the 8-pin EPS connector is used, should't there be relatively little current drawn from the 24-pin connector? If combined draw of the 24-pin is under say 5A, a single ground wire might suffice for all the rails?

 

[vinnyoflegend]

Power Supply - FSP FSP500-30AS Modular Flex PSU from GEEEK, anyone have one?

A few questions: 1) Did you just use 5V or 5Vsb for the Power Good pin or…? 2) Were the two red 5v wires the only soldering you did? 3) Did you just double/triple wire at certain pins at the adapter connectors as needed? ie. It looks like the two red 5v wires you soldered then went into five...

smallformfactor.net

I was running with 11 wires in the post above. No power ok was needed. I could have tried removing ground and the extra 3.3v and 5v

Currently have 12 wires (got a 24 pin housing and wired up the 2nd 12v). But I could try removing one of each rail and the 4th ground to go down to 8 wires if it’s of interest.

EDIT: Curiosity got the better of me, works fine with just 8 wires.

3.3v, ground, 5v, PWR_ON, 5VSB, ground, ground, 12v

 

[k0n]

Sorry if it's known but not possible for some reason, but wouldn't a picoPSU take care of everything and just need 3 wires to it?

12V/GND/Signal to the main PSU with high compatibility. Leave out the pico's EPS and bundle a couple 12V/GND's for the pico supply at the main PSU.

@vinnyoflegend I don't see a graphics card this might be a factor.

 

[vinnyoflegend]

Sorry if it's known but not possible for some reason, but wouldn't a picoPSU take care of everything and just need 3 wires to it?

12V/GND/Signal to the main PSU with high compatibility. Leave out the pico's EPS and bundle a couple 12V/GND's for the pico supply at the main PSU.

@vinnyoflegend I don't see a graphics card this might be a factor.

Yes, the J-HACK M2426 OP was mentioning is basically like a picoPSU that gets 12v from another PSU (for the purposes of wire reduction). Since it was out of stock, I was considering getting another one that was pre-wired to get 12v from a PCI-E 6-pin, it would require soldering PWR-ON to get them to turn on at the same time (the M2426 already assumes you will want this so includes a PIN/wire for it). But after getting 5v out of the PSU like others in the thread I decided just to run it directly but with fewer wires.

In another post in the thread I've run a 2080 ti with the motherboard getting a single 12v from the ATX power before, so no issues pulling 75w (plus more power from the PCI-E slot according to GPU-Z). I don't have a 75w GPU without PCI-E connectors to test that specifically but I expect similar results.

 

[protocolsix]

Wow, that is cool you've reduced it down to just 8 wires!

Does the motherboard even require +12v from the 24-pin connector if an 8-pin CPU EPS is also used?

I did some more reading and it seems the +12V from the 24-pin connector is typically used for RAM, so trying to supply all +12V just from the 8-pin EPS connector probably won't work.

Interesting PWR_OK isn't needed at all! I wonder how many motherboards actually require it. It's wild how bloated the 24-pin connector is...

If the motherboard actually requires PWR_OK, I suppose an alternative would be shorting it directly to +5V on the motherboard connector side instead of a dedicated wire.

 

[b_force]

The issue here are not just the wires, but the pins have a certain maximum rating as well. So be careful with reducing pins!!
Also don't forget that the same return current needs to go over the ground pins/wires.

For some reason PC specs of those pins seem to be very low compared to the values brands like Molex or Würth give in their datasheets.

My guess would be that mixing and matching similar connectors from different brands can result in a less optimal connection.

 

[BaK]

As @b_force above, I would be very cautious cutting down the wire/pin numbers like that.
@Thehack, @REVOCCASES, any advice on that?

 

[vinnyoflegend]

One thing I'm wary of, is some motherboards have a USB-C port that supports USB-PD. above 15w, it seems the potential is there for devices to draw 3-20a@5v. If this is really happening then having the 5 wires/pins for 5v (and more grounds) could be important. It could still be a concern even if the board is converting 8-9a@12v for the PD (some Thunderbolt cards do this but of course they get dedicated PCI-E connectors for power).

 

[b_force]

As @b_force above, I would be very cautious cutting down the wire/pin numbers like that.
@Thehack, @REVOCCASES, any advice on that?

Since I am an electronics engineer myself, I could answer a bit here as well.

Unfortunately, the specs and max ratings PC manufactures are using, are not nearly the same as Molex, Wurth etc etc are suggesting (and I am using on a daily basis).
I have no idea why this is, and it has been puzzling me for years.
The best answer I can come up with, is the very low grade connectors you sometimes find.
Which you have to design for as a manufacturer.
These will result in a much lower power rating, since the contact resistance is much higher, since P = I² * R
Or in simple words, a little bit more current will drastically increase the dissipated power (=heat) inside a (poor/cheap) connector.

In general, I don't find it a good approach to just cut wires left and right even if it has enough overhead.
Unless it's a system that is just never going to change at all in the future and has not other heavy power consumers.
Point is, that is not just a problem of wires, but connectors as well.

If anything, you could maybe just use an adapter PCB/solution, to keep the same amount of connector pins, but cut down on just the wires.
Assuming this will also not cause any addition EMI problems and all grounds are connected (never looked into that actually)
If you want to stay within the same spec you won't win much in cable diameter I think.

@vinnyoflegend With PD, the max current stays the same, just the voltage is being boosted up.
I don't know how manufactures actually do this on the PCB itself.
The nice way would be using the highest voltage available, which is 12V, but I know there are also PD chips that regulate everything from 5V.

In that case you're right that the 5V rail can all of a sudden ramp up the current drastically.
The new PD standards go easily up to around 50W and even above 100W.
I have now idea what USB ports are limited to, but around 25W @ 5V would be around 5A additional load.

 

[vinnyoflegend]

Unless it's a system that is just never going to change at all in the future and has not other heavy power consumers.

Funny you mention that, I believe that's part of the point here with these SFF builds, either selecting hardware with low power draw or configuring for lower power draw (under-volt, capping power limits lower, or even downclocking). Someone on these forums and building with this wire reduction goal isn't likely to be powering a full sized ATX board with 6-7 populated PCI-E slots with cards all relying on full 75w slot power per card.

OEM builds with "proprietary" connectors with reduced pin and wire counts may also fall under your criteria (not expecting or allowing upgrades or publishing documentation that upgrades may require different grade PSUs that may have better wiring and connectors, etc).

 

[SFFMunkee]

It's wild how bloated the 24-pin connector is...

Exactly why I'm so mad that manufacturers and OEMs didn't get onboard with ATX12VO - it's...

Alt text here being "an elegant [solution] for a more civilised age"

 

[SFFMunkee]

One thing I'm wary of, is some motherboards have a USB-C port that supports USB-PD. above 15w, it seems the potential is there for devices to draw 3-20a@5v. If this is really happening then having the 5 wires/pins for 5v (and more grounds) could be important. It could still be a concern even if the board is converting 8-9a@12v for the PD (some Thunderbolt cards do this but of course they get dedicated PCI-E connectors for power).

Usually USB-PD is specified on both ends, and even in the cable there's an eTag that will specify which voltages and currents are supported, plus with PD3.0 and PPS you can do any voltage between 5V and 20V (IIRC) but limited to around 3A or 5A maximum regardless of voltage.

If your board is expecting to do high power USB-PD (e.g. 45+ watts) it might have a supplementary PCIe connector - otherwise you're more likely limited to ~25W (5V3A 15W / 9V2A 18W / 12V2A 24W or something like this). I know many of the TB3/TB4 add-in cards often have one or more PCIe connectors for exactly this reason.

 

[SFFMunkee]

Power Supply - FSP FSP500-30AS Modular Flex PSU from GEEEK, anyone have one?

A few questions: 1) Did you just use 5V or 5Vsb for the Power Good pin or…? 2) Were the two red 5v wires the only soldering you did? 3) Did you just double/triple wire at certain pins at the adapter connectors as needed? ie. It looks like the two red 5v wires you soldered then went into five...

smallformfactor.net

I was running with 11 wires in the post above. No power ok was needed. I could have tried removing ground and the extra 3.3v and 5v

Currently have 12 wires (got a 24 pin housing and wired up the 2nd 12v). But I could try removing one of each rail and the 4th ground to go down to 8 wires if it’s of interest.

EDIT: Curiosity got the better of me, works fine with just 8 wires.

3.3v, ground, 5v, PWR_ON, 5VSB, ground, ground, 12v

I'm actually surprised this worked! I would have expected you'd need to split out each rail for all of the pins the board expects those voltages. I wonder how it impacts the voltage regulation for different components. And I'd be very interested to see how much heat build-up you get in those connectors/pins along the way.

 

[vinnyoflegend]

I'm actually surprised this worked! I would have expected you'd need to split out each rail for all of the pins the board expects those voltages. I wonder how it impacts the voltage regulation for different components. And I'd be very interested to see how much heat build-up you get in those connectors/pins along the way.

I was very surprised too.

Most of the CPU power is surely coming from the PSU's 8-pin EPS connection using 18AWG Cable Matters brand extensions. I have been running the main ATX connection with 12 wires and pins (10 from the original connector PS_ON, 5VSB, 2x 3.3v, 2x 12v, 4 ground + 2x 5v wires soldered salvaged 18AWG crimped with female connectors per the thread). My current "final" cable is made from the same sleeved extensions, with just 12 wires. (see pic)

The colorful short extension harness was what I used for my original testing and features "cheap" 180mm pre-crimped 20AWG wire so I was expecting issues (if you read the wire labels it's even only rated for 80C). But it didn't seem to have any issue even when I reduced each rail and ground for 4 less wires. It is only a Ryzen 7 5700G that at most hits PPT 90W-100W during cinebench23. Wires never seem to heat up but maybe the pins do? I don't have a thermal imaging camera to observe. If anyone can recommend one that has other household uses I'll consider picking one up.

I'm currently putting it to work (with 11 wires going into the ATX 20-pin connector) on an old z97/4790k(4.4@1.17-1.2v) system which tops out at similar 100W power usage during CB23 runs.

 

[b_force]

If anyone can recommend one that has other household uses I'll consider picking one up.

These days there are those cheap Chinese thermal cameras. I forgot the name and model, but they are actually pretty good

In this case you can also use a more subjective method, just feel how hot the pins get. It's all low voltage anyway, so no danger there.

 

[b_force]

Funny you mention that, I believe that's part of the point here with these SFF builds, either selecting hardware with low power draw or configuring for lower power draw (under-volt, capping power limits lower, or even downclocking). Someone on these forums and building with this wire reduction goal isn't likely to be powering a full sized ATX board with 6-7 populated PCI-E slots with cards all relying on full 75w slot power per card.

OEM builds with "proprietary" connectors with reduced pin and wire counts may also fall under your criteria (not expecting or allowing upgrades or publishing documentation that upgrades may require different grade PSUs that may have better wiring and connectors, etc).

Don't get me wrong, do whatever you want to do with your stuff.
(Since electronics is my work as well as hobby, I can guarantee I've done some "interesting" things myself ).

It was more like a disclaimer against just general advice.
More because we don't know how careful people are just reading down the sideline.

The main issue here is still connectors.
I only use the fire retardant type (since that is often also mandatory by regulations), but I have no clue if PSU and mobo manufacturers do the same.
Which is a good example why I don't understand what manufacturers in the computer business are using, because those boiling, melting and burning connectors on GPU's shouldn't be possible to begin with.

Anyway, I am getting sidetracked, lol

My best general approach would be just be either using and adapter board or otherwise just measure the current per pin to get a good sense of what's going on.

Btw,
In the old AT and ATX power supplies, cables of the same color where just going to the same connection on the PCB. I don't know if they still do that, it has been a while that I took a PSU appart.

 

[vinnyoflegend]

These days there are those cheap Chinese thermal cameras. I forgot the name and model, but they are actually pretty good

In this case you can also use a more subjective method, just feel how hot the pins get. It's all low voltage anyway, so no danger there.

What's "cheap" so I'm looking in the right price range? I was thinking of getting a Flir from amazon warehouse as it seems those dip down into the low 300s. Interesting, would I be able to feel the motherboard side pins from the back of the motherboard? Otherwise if you mean quickly powering off and feeling where the cable and motherboard pins mate, those haven't had any trace of heat (to me, 35C-40C would feel noticeably "hot", or if ambient air is 27C+)

It was more like a disclaimer against just general advice.
More because we don't know how careful people are just reading down the sideline.

Very true, the DIY sub on reddit has to always have disclaimers about working on garage door openers (springs kill), asbestos tiles (don't remove them yourself), etc

The main issue here is still connectors.
I only use the fire retardant type (since that is often also mandatory by regulations), but I have no clue if PSU and mobo manufacturers do the same.
Which is a good example why I don't understand what manufacturers in the computer business are using, because those boiling, melting and burning connectors on GPU's shouldn't be possible to begin with.

Anyway, I am getting sidetracked, lol

My best general approach would be just be either using and adapter board or otherwise just measure the current per pin to get a good sense of what's going on.

Btw,
In the old AT and ATX power supplies, cables of the same color where just going to the same connection on the PCB. I don't know if they still do that, it has been a while that I took a PSU appart.

The melting/burning connectors also boggles my mind. The last time I saw a hot pin and charred connector, it was about 20 years ago on a socket A/462 motherboard. The cause I'm fairly certain was some failed capacitors (could see the leak and bulges) in an Enermax PSU, it weakened the 5v rail delivery to <4.6v and the motherboard just kept drawing more.

I've been wary about crimping my own cables too as I'm not sure what is a reliable source of wires and connectors these days (and still haven't decided on a crimp tool either (Engineer PA-21 for $40 or Molex service grade for $80). So everything I have used so far is de-pinned from other cables.

I thought about getting the adapter board originally but I don't see what those PINs get me when the PSU can already generate all the voltages, I might as well just crimp double wires and fill out out the connector if the extra pins are that critical (I don't think they are). I'll have to see if my multimeter can be put in line somewhere.

The cables are all black now but yes, as far as I've seen this is still true that the wires for the same rails/function often terminate in just a few spots on the PSU PCB.