• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

Design analysis discussion of Nvidia-supplied 12VHPWR connectors

You have to remember more data is still required what is the true issues (if any) with the sockets. If it's melting at the GPU end, then one will need to look at the latest PSU which also has 12vHPWR socket. If it melts here, then there is a serious issue with the end-built quality of the socket/plug pins. I'm fairly confident what was test in a lab, did pass. Something must have change from what was tested in a lab to what end users has ended up with.
 
Let's not forget the solder joints that Igor first pointed his finger at. They are beyond horrible, as if made by someone who only had an axe to cut the wire, his own teeth to strip it and a candle for soldering. There's also no strain relief, all the pulling force pulls on the solder, which becomes brittle over time, more so if exposed to vibration. Solder pads made of very thin metal aren't helpful either. An industrial product should never be made like this.
I'm not saying that solder joints caused any of the accidents; they probably didn't. But after a few years, and a few repeated attempts at cable routing, they would become one of the weakest and most dangerous points of failure in the entire PC, with broken off wires also able to cause a short circuit from +12V to ground.
 
As I mentioned in another thread, I'm waiting for a 12VHPWR class-action lawsuit..... :D
 
P4-630 said:
As I mentioned in another thread, I'm waiting for a 12VHPWR class-action lawsuit..... :D
Click to expand...
Too late as of 11/16/22 filing.
 
delshay said:
You have to remember more data is still required what is the true issues (if any) with the sockets. If it's melting at the GPU end, then one will need to look at the latest PSU which also has 12vHPWR socket. If it melts here, then there is a serious issue with the end-built quality of the socket/plug pins. I'm fairly confident what was test in a lab, did pass. Something must have change from what was tested in a lab to what end users has ended up with.
Click to expand...

It could also be simple manufacturing variance. If tolerances are off in both pin and socket of the same circuit, contact could be poor enough to cause an overheat like what we see here.
 
Update in case you missed Gamer'sNexus take on Nvidia's official response from today, rma dedication and debunking warranty rumors about Cablemod 3rd party cables.


Update latest Gpuz monitoring 12 pin cables.

GPU-Z Now Monitors Your 16-Pin Power Connector on RTX 4090, 4080

Now you know (nearly) all the peculiarities.
www.tomshardware.com www.tomshardware.com
 
Last edited:
Wth are these still out of stock at MSRP everywhere? The overclock.net crowd seems to be pushing these things in excess of 600W without issues.
 
mouacyk said:
Wth are these still out of stock at MSRP everywhere? The overclock.net crowd seems to be pushing these things in excess of 600W without issues.
Click to expand...
Welcome to the time period known as "happy holidays."
 
mouacyk said:
Wth are these still out of stock at MSRP everywhere? The overclock.net crowd seems to be pushing these things in excess of 600W without issues.
Click to expand...
My local Microcenter got in 25 units in today FYI. seems like other locations as well.
 
80-watt Hamster said:
The hot-button (ha!) issue of the moment is melting 12VHPWR adapters provided with RTX 4090 graphics cards. Many of us (hi!) downplayed the trouble to begin with, chalking the furor to new-launch hype and anti-NV sentiment. Turns out there is a real issue, and I'd like to discuss it specifically, and hopefully rationally. Igor's Lab has an excellent teardown of the connector, and goes top-level into what the problems with it are. Please read that first.

I'd like to dive a bit further into why the connector is designed the way it is, and fill gaps in my own understanding, of which there are many. I don't have any credentials in electronics engineering, and am in fact an EE dropout. What I do have is a reasonable grasp of DC circuit fundamentals (P=I*V, V=I^2*R, etc.) and a strong analytical bent. Anyway, based on Igor's article, I drew a crude block diagram in MS Paint of how Nvidia's PCIe-to-12VHPWR connector is configured, and it doesn't make a lot of sense at first blush:

View attachment 267488

The questions that pose themselves include:
1) Why 4-14ga rather than 6-16ga, including two 1-into-2 nodes?
2) Why a monolithic contact array instead of individual contacts?

To the 14ga question, there are some conflicting pro/con factors. 4-14 can carry as much or more current than 6-16. The draw back is that it is thicker, and thus necessarily less flexible. One of my hopotheses is that the extra bending resistance encountered when routing cabling is putting more strain on weak solder joints than 16ga would, increasing risk of broken joints. The existence of only four 12V leads helps answer a question I didn't list above, which is why soldered rather than crimped. Well, because you can't crimp one wire onto two connectors, also soldering is faster and cheaper to implement. The other is that given the adapter draws from four 8-pins, combining those four groups of three wires (twice, once for hot and once for ground) into six leads rather than four is unnecessarily complicated given that it's all going to one node anyway.

The contact array/block is honestly my biggest question. Something I originally typed up in this post but deleted was something about the GPU not knowing that a connection had been lost when one or more 12V leads breaks, since all current is running through the same node. I deleted it due to a lack of knowledge of how things work on the GPU side. Igor posted a block diagram of the connector that indicates that all 12V leads run to the same node by design. Why would this be? If a graphics card isn't treating each 12V connection as an individual circuit but all together as a single node, what's the point of a 12-pin connector? It would be just (or almost) as space-efficient to implement a two- or four-pin with nice robust contacts. Two pins is maybe optimistic because it would require 50A capacity, but 25A for a four-pin seems like it would be doable. If, by contrast, those separate circuits are meaningful, why run all the supply circuits through a single node?

This is meant to be a technical discussion. Do not flame, complain, troll, or hijack. Report button will be employed without hesitation or remorse.
Click to expand...

I noticed Solaris hasn't commented on this, except to suggest buying the Cablemod adapter after calling me "mad" merely for saying people were taking too much risk using these poorly made adapters Nvidia supplies. Still think Nvidia's ones are "fine" Solaris? Well, you aren't really owning up to what you said before IMO, but it appears now that you don't. LOL

IMO, especially after what's happened to some of these plugs, they should require a QC sticker to prove they pass Intel's design standards on them. And I'm talking ALL 12vhpwr plugs, on adapters AND PSU's that supply the plugs and sockets. These requirements should include wire gages, pins, and any material type and thickness used as a solder or ground base. Basically any part of the structure that carries that 600w load.
 
Last edited:
Last edited:
You must log in or register to reply here.
Back
Top