ATX 3.0 12VHPWR connector type concerns

[Ranayna]

Does anyone know to which Molex connector type family the new ATX 3.0 12VHPWR connector belongs?

It is obviously a hybrid of some sort, with it's 12 power and 4 data pins. I am most interested in what the equivalent for the power pins would be. It has a pin pitch of 3 milimeters, so i think that Molex Micro Fit would be the equivalent? https://www.molex.com/pdm_docs/ps/PS-43045-001.pdf

One reason i ask is, that there are increasing reports about melting connectors. This 12VHPWR connector is supposedly rated for 600 Watts at 12 Volts, and the new nVidia 4090 can draw 600 watts peak power according to preliminary tests.

A Molex Micro Fit would not fit the bill then. If i understand the datasheet correctly, in a 12 pin configuration, the rating per pin is just 6.5 amps.
If i have not miscalculated, supposing equal load sharing between all pins, you would get more than 8 amps per pin at 600 watts. Maybe not a gross overload, considering that 600 watts is peak power, but still concerning.
Combine that with the low mating cycle count of just 30, and this is not a connector that i would like to use in a PC, where any layman can build one. Not at these power levels...

I suspect, or at least hope, that this connector will not be part of the standard for long, at least not on this power level. But what could alternatives be, besides going back to the larger and higher rated Molex Mini Fit?

[Ranayna]

Ok, i was actually pointed towards the actual connector:
https://www.digikey.de/en/product-highlight/a/amphenol-fci/minitek-pwr-cem-5-pcie-connector-system

So it is an Amphenol connector, of the Minitek Pwr 3 series.

Does not change much, according to this datasheet it still seems to be underrated for 600 Watts:
https://www.digikey.de/de/pdf/a/amphenol-icc/minitek-pwr-3-product-specification

[Algoma]

I get the feeling they decided on the connector before Nvidia decided to push its limits. One single bad pin connection will cause specatular and expensive failures, especially with the overclockers pushing things FAR beyond the limits..

Next PC accessory market will be a automatic fire supression systems.

[Someone]

Next PC accessory market will be a automatic fire supression systems.

audio power-phile grade power distribution kits with exotic coatings and burn in periods, "upgrade your power connectors to unleash the full capability of your hardware"

[Kahooli]

I would laugh if the problem was some rube who just divided 600W/12V/12 Pins and said "4.16A, that a great safety factor for a 5.5A rated pin!"
Forgetting that it's half as many pins I in and half again I out.

Though it's more likely Molex designed a customized set of changes for the higher capacity.

[Algoma]

So far, the cards seem limit themselves based on the number of cables are plugged in, by tying the data pins to ground though each of the adapter legs. I fear the Knockoff cheap replacement adapters will be an actual real fire hazard with all tied to the same ground.

 Otherwise reviews seem to put the RTX4090 power draw within reasonable limits of this connector (so far) .. let's see if the overlockers might burn up a connector after a few mods.

When you figure all that power is dropped to low voltage at the chip, Thats a LOT of current being put through that piece of silicon.

[Ranayna]

I have to retract my statements above a little bit. Sorry for that.

First: It is not a Molex connector. It is made by Amphenol. I blame the ubiquity to call each and every connector in PCs a "Molex" connector :p

https://www2.mouser.com/new/amphenol/amphenol-minitek-cem-5-pcie-connector-system/

Second: This thing has the proper ratings.

But nonetheless: Considering the power capabilities of current high end power supplies, i am still a sceptic here. There are supplies recommended for the new nVidia GPUs that can deliver more than 100 Ampere on 12 Volt. A *lot* of power, with the potential for a lot going wrong without any protection circuits reacting to a lingering fault.

[mariush]

They could have easily fixed some of these issues by adding 20v +/- 5-10% to power supplies.
 
Drop 3.3v, and make psu output 20v and 12v (still need for cpu and backwards compatibility with pci-e cards and support for fans and hard drive motors etc) and use dc-dc converters for 5v (for usb, sata etc)

As a bonus we would also get USB power delivery as that already has 20v defined, so would have been nice to also have USB ports that can push 20v 100w to a monitor or an external video card.

[Ranayna]

Well...
The first reports are coming in:

https://www.reddit.com/r/nvidia/comments/yc6g3u/rtx_4090_adapter_burned/

Considering that such an RTX 4090 costs more than 2000 Euro, that is just bloody ridiculous.

[mariush]

Apparently the problem is often caused by bad contact due to wires being pulled out of the housing at an angle.

The female pins are supposedly made out of bent metal with a vertical cut towards that ends up towards the top of a video card.
When the connector with male pins is inserted, and the cable is pulled down, the pin inside tends to press up into that vertical cut, potentially stretching the female pin stretching it out and therefore making the contact looser

He explains better below :

[Ranayna]

Yes, i realize know that this is a common issue.
But why then, considering the size of that graphics card, did nVidia not use an angled connector, allowing the cables to go straight down without strain? Or just add enough strain relief? Well regarding the latter: With a 3.5 cm strain relief, the card would not fit into many cases anymore 

The card is so large that in many cases there is quite little space between the card edge with the connector, and the case sidepanel. The power cable has to be bent, and since there is no a long enough strain relief, it will bend close to the connector.
For cards that are mounted vertically, so that the powerconnector is at the top, this is even worse. Gravity alone will bend the cable relatively tight since it does not have strain relief.

[Algoma]

I suspect that it's most likely due to poorly crimped wires at the pins that's generating the initial heated point of failure, instead of the pins themselves. One badly crimped wire and it all goes poof, not much margin for such a fragile process.

[Ranayna]

The adapter that nVidia itself includes with their 2000 bucks GPU seems to be utter and total crap:

https://www.igorslab.de/en/adapter-of-the-gray-analyzed-nvidias-brand-hot-12vhpwr-adapter-with-built-in-breakpoint/

[HwAoRrDk]

Yikes, that's some horrendous construction on that adapter. That they decided to use solder connections on something they must have known would be subjected to bending is reprehensible. And this is an nVidia first-party accessory cable, not some shonky third-party no-name budget vendor.

Even if it's this particular cable adapter to blame, and not the 12VHPWR connector format, I still wonder why the PCI-SIG decided to specify such a small connector. What's the point of a small connector if you still have a humongous cable bundle attached to it?

[coromonadalix]

What bothers me on this connector,  why oh why did they no use  bigger pins and or an 2 or 4  biiiiiiiig  pins connector

this new standard is plain stupid

do we need to have copper power bars now     0 awg wire gauge  now   loll  seriously

[Ranayna]

And on goes the show 

First reports of melting connectors without any adapters are coming in:
https://new.reddit.com/r/nvidia/comments/yltzbt/maybe_the_first_burnt_connector_with_native_atx30/

Of course this also happened with the old connectors, so one has to wait and see. One swallow does not make a summer. 

But since there are still very few power supplies on the market that natively have this connector, that one early swallow may be a concerning sign. 

[Algoma]

Looks like square pins on the card side, certainly any slight rotation of the wire when seating the cable could certainly limit contact area with the pin. almost an all or nothing type connection.