1
Looking at a diagram below of a PCIe power supply connector it shows it has 3 12V pins. When it's plugged into the PSU how many 12V rails does it use?
I assume it is just one, but how does that work? How can it provide three separate 12V pins?
Thanks.
RJSmith92
Posted 2015-08-28T01:08:52.230
Reputation: 820
2
You have multiple ATA and/or SATA power connectors coming out of your PSU, no? All of their 12 volt pins come from the same rail. "How can they provide howevermany separate 12V pins?" Easy: All those pins are connected via individual wires to a common connection point back in the PSU.
The reason this is done is that there is a limit on how much current each pin can safely handle. This has nothing to do with how much current the PSU can provide; it has to do with the resistance inherent in the pin-and-socket connection. For the usual "Molex" ATA drive power connector this is 11 amps per pin. It's a little less (7 amps) for the smaller pins in the PCIe, etc., connectors.
By running multiple wires and multiple pin/socket connections in parallel, the effective total resistance is greatly reduced. If you have four +12V pins then the effective total resistance is just 1/4 of that of any one pin. So you can (in theory) safely draw four times as much current without overheating the connector or causing other damage.
Jamie Hanrahan
Posted 2015-08-28T01:08:52.230
Reputation: 19 777
2
Theoretically, all the rails can be seen as one. More pins just allows for more current to flow to the device.
1 garden hose can supply 10 gallons per minute, 3 garden hoses can supply 30 gallons per minute.
One garden hose that is larger in diameter can also move that same 30 gallons per minute but is much heavier and stiffer to move.
Matt Clark
Posted 2015-08-28T01:08:52.230
Reputation: 1 819
Thanks @Jaimie, I've just realised that now. A single 12V rail can have as many connections as it wants coming from it, but it will always be pushing with a force of 12V. A maximum amount of current is then shared between each connection connected to that 12V rail. I believe that to be correct? – RJSmith92 – 2015-08-28T01:27:34.127
Just realised I misspelled your name, Sorry! – RJSmith92 – 2015-08-28T01:33:40.757
1Yes. Current follows all available paths, dividing itself according to the inverse of the resistance in each path. So if you had one path of 1 ohm resistance and another path of 2 ohms, the 1 ohm path would carry twice as much current. Ideally in these connectors all the paths have identical resistance, so the current will be divided equally among them. In the real world things are never quite so identical, but the resistances in the PCIe connector will be fairly close to each other. – Jamie Hanrahan – 2015-08-28T01:38:35.987
1As for voltage: Yes again! Voltage at all points along a conductor is the same (assuming either zero resistance or no current). In practice resistance is never exactly zero, so there are slight differences in the voltage at different points in a supply "rail" when current is flowing. For example the "+12" at the brake light in your car will measure a little lower than it will back at the brake light switch when the brake light is on because the latter is much closer to the car battery. But when the switch is open, no current flows, so the voltage is the same at both points. – Jamie Hanrahan – 2015-08-28T01:46:13.103
Sorry to bother you again, was just looking at this page - http://www.playtool.com/pages/psuconnectors/connectors.html#pciexpress and for example the 6 pin PCIe connector. It says the maximum current per line is 2.083A. What dictates this is the maximum? Are GPUs designed to only draw that much from each line?
– RJSmith92 – 2015-08-28T02:03:10.2831The 8 amps per pin spec is from the connector maker. Like the power ratings on resistors and the like that is likely a "free air dissipation" spec, assuming good air circuilation and nothing else nearby making heat. Whereas 75 watts is the maximum the ATX spec allows for the entire 6-pin connector, with six of those heat-making connector pairs packed into a tight space. Divide it out and you'll find it comes to about 2.1 A per pin. Apparently when they spec'd the 8-pin version they decided 75 watts was too conservative; the 8-pin version is spec'd for 150 watts. – Jamie Hanrahan – 2015-08-28T02:30:32.497
1(Too much for one comment!) On the GPU all of the +12 pins are connected together, and all of the ground pins are connected together. This is true even for GPUs with two power connectors. The GPU draws the total current that it needs and the current divides itself among the available paths, as described before. – Jamie Hanrahan – 2015-08-28T02:32:44.313
ahh that makes sense. Thanks for the quick response. – RJSmith92 – 2015-08-28T02:36:44.053
1One afterthought: Why does the connector maker spec amps, but the PSU specs specify watts? A: Molex doesn't know what voltage their pins will be carrying. But just like wires of a given gauge have a specified "ampacity", independent of voltage, so do connector pins. Now 3x7 amps at 12 volts through a 6-pin connector would be about 250 watts. If the pins don't make perfect contact you could easily be dissipating enough heat in the connector to burn them a bit. They then make more heat...this is soon a runaway condition ending in a charred connector body and maybe even a fire. – Jamie Hanrahan – 2015-09-04T10:45:16.630
Hi @Jamie, Sorry to dig up this question, was re-reading it and just wonder if you can confirm something for me? Is the following correct(ish) - The reason they use multiple pins is because pin/socket connections provide a certain amount of resistance and having a high current (21A for example) going across a single PCIe pin/socket resistance could damage it (overheat perhaps?), but by using multiple pins, the GPU can divide the amount of current it needs between each wire and therefore less current goes across the socket/pin resistance of each wire, reducing the risk of damaging it. Thanks. – RJSmith92 – 2016-01-10T00:19:12.873
1Basically yes, that's it. But it's maybe better to think in terms of power. Power dissipated in a resistance is equal to the current in amps squared multiplied by the resistance in ohms (I-squared-R loss). If you have six pins all in parallel instead instead of one then each pin carries only 1/6th the current. Since the R of each pin doesn't change that means that your power dissipation in each pin is 1/36 what it was before. Across the entire six-pin connector the total power dissipation is still only 1/6 that it would be with one pin. – Jamie Hanrahan – 2016-01-11T13:04:17.260
Brilliant, thanks for that @Jamie. So finally, it's more to do with efficiency than risk to hardware? – RJSmith92 – 2016-01-11T16:50:15.830
1It's really both, since power dissipated in resistance turns into heat... which of course could damage the connector. – Jamie Hanrahan – 2016-01-11T22:24:39.577