Question on ATX12V 4-pin connector amperage requirements (PSU modding).

[Machupo]

I'm working on another update for my uSFF system and have decided to try out the PSU from the Alienware X51 (original thread from the SFF forum over at H). Everything seems to line up with the exception that the 4-pin motherboard auxiliary power connector puts out @ 19vdc.

Once I completed the Willem Dafoe "there was a firefight" at my work bench, I started thinking about some workarounds. I came up with a pair:
- There is a 4-pin male molex connector on the power board, which ostensibly, can put out 11A on the +12V (at least by connector spec, there is so little documentation for this POS, though, so this could be way off). I'm running a (for now) stock 3770k which should put it at or under 100W (after adjusting for an 80% mobo VRM efficiency), this squeeks me under the 132W max delivery through the molex plug (more = meltdown).
- I could use a single (or pair in parallel) of 12A (100W max) buck step downs to drop the 19vdc to 12vdc. Only issue I can see with this (if I go with the headroom of the dual bucks, one per live wire) is: what a tiny discrepancy between my adjustments on the two step downs would do to the VRMs. Obviously a single buck would again run into headroom issues at max CPU loading.

What this all comes down to is: how much amperage actually runs through the 4-pin ATX12V auxiliary motherboard connector on a maximally loaded processor? Is *all* of the processor power provided through this channel and the power that goes to the motherboard through the 24pin ATX connector segregated for other motherboard functions. Or... is the auxiliary connector only "used" when the 24-pin connector approaches some pre-set maximum of 12vdc power delivery? Any software out there that could provide this info (I have AIDA64 extreme and haven't seen it)?

Taking bets on if I cook this entire system at some point in the process Yes, I understand that it is a marginal drop in size from the well-tested ST45SF-G that is in there, but aimless cranking/soldering can be fun too

[edit]
Went and figured it out myself, thanks.

Went and bought a hall-effect multimeter that could do DC current:

- System power values taken from an in-line Kill-A-Watt
- CPU TDP power values taken from a the reading in AIDA64 Extreme
- Amperage values taken with a Klein CL2000, 3 trials averaged per value
- System was a i7-3770k @ 4.0Ghz, 1.15Vcore on an ASRock Z77e-itx with an evga GTX670-4GB

Conclusions:
- CPU 12V current is provided through the AUX 12v plug
- Peripheral 12V current is provided through the 24-pin plug
- Looks like a single 100W buck stepdown will work just fine

[edit 2]
backtracked the vrm datasheet. has a wide V(in) range, so just plugged it up straight. first boot successful and counting.

 

[pelikan82]

This is an interesting read, certainly for someone like me who has to modify his PSU a lot because of trying new GPUs. Now reading this, I can see that since you didn’t require that much power, you just used a single buck step down, however, it can hold up at extreme limits. I guess you'd have tested it by now with maximum CPU loading for longer durations and I'm really curious to see if it worked even then because normally it isn't advised to put all our eggs in one basket or in this case a single buck convertor. Overall I think this was a brilliant effort that you did on your own which make it all the more interesting.

circuit board fabrication

 

[Machupo]

It turns out that the MOSFETs on the motherboard have a wide range of VDC input acceptance, so I didn't even need the buck stepdown. The 6 MOSFETs on the Z77e-itx converted it directly from 19VDC to 5VDC (and then the VRM subsequently converted from 5VDC to whatever was specified by the CPU).

So I spent a couple bucks that I didn't need to, but it was an interesting exploration nonetheless (and it works!)