PCIe Power Cables and GPU Connections

H

haz_mat

Limp Gawd
Joined
Sep 14, 2012
Messages
326
Hey guys I'm wondering about how my Seasonic 750 handles PCIe power over the supplied cables.

I'm sure other PSUs are like this as well: I'm running my 290x on a single line from the PSU - the line on the GPU-end is split to offer two eight-pin connectors (which can be split again for 2x 6-pin connectors). No problems so far, but I got to thinking about how these were wired up...

I assume Seasonic supplied the proper wire gauge to handle both connectors but the thing that concerns me is the way the 2nd connector is split from the first. The second connector's wires basically nest into the first connector. If that doesn't make sense, here is a picture: Link - imgur.com

I am no electrician, but I can imagine running all the current through one connector like that might not be the best idea. I haven't been having any problems, but due to paranoia I will go ahead and install a second power lead. Initially I thought that less cable mess would be nice, but now I'm thinking this might not be a good idea if I want to exceed 300w on the GPU. Is this something to be concerned about or is this a non-issue? I know my PSU is a single rail, so the power is all coming from the same source - its just this split connector I'm worried about.
 
Last edited:
It's a non-issue. Inside the connector, the power going to the second connector doesn't actually flow through the pins. It goes direct from cable to cable, as the cables are twisted together then crimped to the pin. You don't need to worry about that as causing your connectors to overheat.
 
Dont think i would be the least bit concerned unless it was powering a 295x2 or something like that. But like Tsumi mentioned those pins don't seem to be an issue with the wires twisted together. For most cards its more than enough.
 
as usual going with Tsumi on this

unless you are adapting off of 2, or as I have seen too often 1!, molex wire the pin interface seems to be the weak link

If you were going to clock the nuts off it maybe use a second cable assembly, but for normal use you are probably fine
 
Thanks for the quick responses everyone, I appreciate it. I have the utmost confidence in the Seasonic engineers who design these, thats why I bought this brand. But I still wanted to know for sure and this is the best place to ask. :cool:

I do plan to push this card a bit, as it continues to give clocks with more volts added. I have a bit more headroom from new thermal paste on the GPU and better tape on the VRMs, so its not uncommon to see it pulling over 350w logged in GPUz.
 
it might be worth stringing another cable out and see if you get a bit more clock once you know the card's limits

you know...for science
 
4 x 16AWG conductors would let you pull 1056W before the wires burned. The PSU would shut off way before you got anywhere close to the maximum load of the wires.
 
GiGaBiTe said:
4 x 16AWG conductors would let you pull 1056W before the wires burned. The PSU would shut off way before you got anywhere close to the maximum load of the wires.
Click to expand...

Well, it's actually only 3x16awg, so only 792Watts. The extra two wires on an eight pin connector are for voltage sensing, not power delivery.


The down side of having dual connectors is that the more current that passes through a wire the more voltage drop you'll experience. (V=IR). Since 16awg wire has a resistance of 0.004ohm per foot. If the wire is 24" long, that's a resistance of 0.008ohm.

So if you're drawing 25a (the max for two 8pin PCIe connectors) down two set of wires, you'll drop 0.03v. But if you confine the same load to one set, the voltage dropped increases to 0.07v. You'd still be well within the ATX specs of a good PSU, and the GPU's would still work as intended. However, the VRMs on the card would have to work ever so slightly harder to provide the GPU with stable power, which could mean a slight bit more noise. That might (strong emphasis on might) mean the GPU won't be able to reach as high when being pushed beyond it's spec.

So in the end, it's perfectly safe and well within the limits of the wires to put two connectors on one strand. But it's not guaranteed to be optimal.

Note: some good PSU's only use 18-awg wire, in which case you're have 0.013ohms, 0.054v, and 0.109v respectively.
 
kudos for application of Ohm's law ryan, but do you have a source for the +2 pins being sense?

I was under the impression they were both grounds
 
Great explanation ryan, thanks for that. Those are exactly the kind of metrics that I was wondering about. I figure every little bit has the potential to help, all it takes is one little hiccup at high overclocks to start spitting out artifacts.

I've got the second power lead attached and I haven't done any tests to see if the stability threshold has changed but I know where to start. I'm not expecting much, but there is no fun in not trying.

On a side note, I think there is less physical strain on the GPU connectors with the two dedicated leads. With the way the split-leads are only a few inches, I had to bend a pretty narrow U-turn into the wire to get it to plug-in.
 
jojo69 said:
kudos for application of Ohm's law ryan, but do you have a source for the +2 pins being sense?

I was under the impression they were both grounds
Click to expand...

Sorry, I meant one of the two wires was for sensing (the other is for grounding).


Officially, the 6 pin has 2x +12v, 2x GND, and 1 sense. Pin 2 is not connected according to the spec, but most PSU manufacturers put a +12v there anyway. That's why molex to PCIe adapters usually only have two molex connections instead of three.


The new spec officially ties pin2 to +12v and adds a another ground and a second sense line.
 
You must log in or register to reply here.
Back
Top