So, to answer the question “how did Nvidia go from 24 pins (three 8 pin connectors) down to 12 pins (minifit molex)” we would obviously need to start with a donor cable to tear apart and inspect what is going on. I do not have a 3090ti, so this cable will never get used again, and therefore, its sacrifice is so we can learn. This cable is from EVGA, simply because I can not get my hands on any other AIB cables right now. If I can find other AIB cables, I will see how they differ.
For this entire thread, connector 1 will be the 12 pin (11+1, I will explain later), connector 2 will be the right most 8 pin, connector 3 will be the middle 8 pin, and connector 4 will be the left most 8 pin. There is a definitive reason that each connector will be referred to as a connector number, and I will cover that momentarily. This is looking at the cable with all clips facing up, so that you can orient yourself to how the cables will be referred to:
First, I started by cutting off all of the heat shrink, tape, and sheathing from the entire set of cables.
Right away, we can see there is not 24 cables, there is actually 16 cables, with 11 of those cables being 125°C 150v 16awg wires and the remaining 5 being 125°C 150v 22awg . Take note, that is 150v Per Cable Maximum, just going to leave that there, as that is the writing on the cable itself. 4 of the cables from connector 3 and 4 went to connector 2.
From there, I removed the rubber EVGA logo from the connector 4 cable and found a hard plastic:
Removing the connector 4 plastic 8 pin adapter reveals there are 3 different sets of pins inside to the 8 pin hard plastic. Later I can identify the 4 pins in the black box are the negatives with two of the 18awg cables going to the 12 pin, the reds are the positives with two 18awg cables going to positive, and one 22awg cable going to connector 2, and the black and green is a 22awg cable going to connector 2 as well. This aligns with the GPU power connections diagram (although the one in my hand is inverted). The Pins 1 through 3 are boxed in red, pin 4 is boxed in Green and Black, and pins 5 through 8 are in black. Just for simplicity, connector 3 and connector 4 are exactly the same.
At this point, I use a razor to cut straight down the middle of Connector 4 to see if there is anything else in there, and there isn’t.
So with Connector 3 and Connector 4 out of the way, its time to look at Connector 2, since there are five 22awg cables and three 18awg cables going to/from this connector. Sinc the rubber is gone, I started looking much closer at the plastic on connector 2, and realize there is a PCB centered inside the plastic.
I knowed off the number 4 pin trying to cut through the plastic as it is very thick. Once I was able to see the PCB a little better, I went straight down the first side of the PCB, and took the remaining 3 pins off:
This reveals the reasoning that I started with calling the 12 pin connector 1, the right most pin connector 2 (this one) and the other two 3 and 4. The Sense pin on connector 2 was soldered to location 4P4, the 3 pin 12V+ was soldered to VP4. On the front of this PCB, there is a version number (Version 2 in this case). 4 resistors of unknown resistance, 2 capacitors of unknown capacitance, what looks like a tiny MOSFET labeled Q1, and lastly a PSNT location for the single 22awg cable that runs out from connector 2 to the 12pin.
On the backside of the PCB, two more mosfets labeled Q2 and Q3, two more capacitors, one possible diode (unknown color), the 4 connections for Connectors 3 and 4 labeled VP2 and VP3 for 12V+, and 4P2 and 4P3, as well as #7 for the negatives coming into the PCB.
The 16AWG wires are connected with two coming from 12v+ and one from from 12V- to the 12 pin, and a single PSNT 22awg sense cable running to the 12pin.
So the connections from the pig tails are set up like this on the 12pin:
The PCB seems to be a check board to verify that all three 8 pins are connected, and if they are not, then the connection may not work. I labeled the sense pin as 2S on the picture, so because I can not get a good picture from the connector I actually have here. If anyone of curious about what color the diode is when all connections are plugged in, or one is missing, you can move the rubber sleeve without damaging it to see the color on connector 2. Unfortunately, I do not have a 3090ti to plug a cable into for testing this one thing, and I was given this cable by a forum member to help me figure out exactly what it was that NVidia and the board partners had done to accomplish the feat of going from 24 pins down to 11+1 in this case.
So, "how does NVidia go from 24 pins to 11+1 on the 3090 TI?" Joining all of the PSU connections into three separate connections, and only running 4 cables to 1/3rd of the 12 pin, and then having the positive and negative control lines from connector 3 and 4 go to connector 2, and a tiny little PCB, sense if all connections are present, and then sending a "go/no go" signal to the 12 pin. Personally, I found this extremely interesting, and I have the 41s and 52U on the way so that I can look at how those are put together. Unfortuantely, no one really wants to give up their triple 8 pin to 12 pin connectors from other companies, so I can not see if ASUS or any other company is doing the same thing, but now I am very curious if the dual 8 pin to 12 pin used in the 3090 and lower also has the same type of sense pins. If anyone has a dual 8 to 12 pin, I will gladly destroy it to figure out what has been done before.
If you have any questions, please ask, and I will try to answer them.