It's happening again, melting 12v high pwr connectors

[Degreco]

The solution is simple, just relocate to a zero-G environment and float your cables out in a straight line from the GPU.

Thanks, will do that in the next build

Where are you getting that Nvidia sold "millions and millions" of 4090s?

Where? When reading 35 pages is too much, but you want to contribute anyway

 

[JAS_design]

Jon Gerow (i.e Corsair) thinks it's ok to bend close to the connector, just need to check you haven't popped any pins out.

With respect to Jon Gerow, with a properly formed housing and properly crimped pins, the tangs on the pins should grip the housing well enough that you can't "pop them out." Pin removal tools exist for a reason. In my experience, the crimp itself should fail before the pin pulls out.

If Corsair regularly has pins popping out of the connectors, Corsair is doing something wrong.

 

[ShrimpBrime]

With respect to Jon Gerow, with a properly formed housing and properly crimped pins, the tangs on the pins should grip the housing well enough that you can't "pop them out." Pin removal tools exist for a reason. In my experience, the crimp itself should fail before the pin pulls out.

If Corsair regularly has pins popping out of the connectors, Corsair is doing something wrong.

I would like to see crimp + solder. That's how I do all my connectors. Especially circuits with 10a load over a single wire. But thats how I do mine.

Are these seriously just crimped?

 

[JAS_design]

I would like to see crimp + solder. That's how I do all my connectors. Especially circuits with 10a load over a single wire. But thats how I do mine.

Are these seriously just crimped?

Unless the specifications say otherwise, you're not really supposed to solder contacts that are designed to be crimped. The solder can wick up the strands of the cable, causing it to become brittle and, counterintuitively, actually weaken the joint.

Crimp contacts are designed to have a little bit of flex for thermal expansion and stress relief; soldering can make the the cable more prone to snapping inside the insulation sheath, where you don't see it has broken. It can potentially cause a hidden high resistance nightmare.

 

[STSMiner]

Lets see if we can end the confusion here with the cable and the connectors used with the cable.
First of all, the cable and the connector (12VHPWR / 12V-2*6) are the same, there is no change here.

The change is with the connector that's mounted on the PCB of the GPU and in the PSU (the PCB Header) - ATX 3.1 specs detailing the change.

https://edc.intel.com/content/www/us/en/design/ipla/software-development-platforms/client/platforms/alder-lake-desktop/atx-version-3-0-multi-rail-desktop-platform-power-supply-design-guide/2.1a/pci-express-pcie-add-in-card-connectors-recommended/#_Ref95471394

This is the info from Corsair if anyone wants to go read it.

Evolving Standards: 12VHPWR and 12V-2x6

With the introduction of the RTX 40-Series, the PCIe 5.0 12VHPWR GPU power connector made its debut with the promise to deliver up to 600W of power through a single connector. As with any new standard, things are likely to evolve quickly and we’re now seeing the introduction of a new connector...

www.corsair.com

It's clear here that there is still a problem with this with melting connectors, this is something that Nvidia and the PCI-SIG need to resolve before someone's house burns down or someone dies because of it.

 

[ShrimpBrime]

Unless the specifications say otherwise, you're not really supposed to solder contacts that are designed to be crimped. The solder can wick up the strands of the cable, causing it to become brittle and, counterintuitively, actually weaken the joint.

Crimp contacts are designed to have a little bit of flex for thermal expansion and stress relief; soldering can make the the cable more prone to snapping inside the insulation sheath, where you don't see it has broken. It can potentially cause a hidden high resistance nightmare.

Well, I want to agree, but then again, not really.

For starters, solder is stronger than the wire strands. Yes it could wick up a strand or two, maybe all of them, but then again using a small amount vs just dumping solder would be up the the person doing the soldering. I've never had an issue.

But I have seen wires break right at the solder joint, but on old (years) high resistance circuits such as lighting where you draw 10 amps.

Pages back you can inspect my work if you'd like, I've already posted examples. Obviously not on an NVidia card, but on 10a circuits with much larger connectors, the male side in the modules are actually spades, not pins.

 

[AVATARAT]

Jon Gerow (i.e Corsair) thinks it's ok to bend close to the connector, just need to check you haven't popped any pins out.

Nah, it's too complicated, just use instant glue inside the 16 pin connector and it will never move no matter what you do to it.

 

[Dredi]

Nah, it's too complicated, just use instant glue inside the 16 pin connector and it will never move no matter what you do to it.

It might just work, if used after plugging in the connector.

 

[qxp]

Nah, it's too complicated, just use instant glue inside the 16 pin connector and it will never move no matter what you do to it.

Not really - organic glues come apart when heated

The connector was brilliantly designed to provide excitement no matter what you do.

 

[Gatt]

Exactly my thought. I long for getting drunk with you.

Some people here really think the connector is safe. Yeah, it's perfectly safe, when perfectly plugged, in virtual conditions.
This connector has such low safety margin that it is rendered useless when you take into account connector/cable manufacturing process variation.

Meanwhile, WTF:

New 12VHPWR angled adapter features fan, power and temperature monitoring for RTX 40 Series - VideoCardz.com

Monitor power and temperature levels, and keep cabling tidy A Chinese company has developed a special 180-degree adapter for GeForce RTX graphics cards with a 16-pin power connector. Before you ask, the company only mentions RTX 4090, 4080, and 4070 graphics cards, so one cannot recommend this...

videocardz.com

View attachment 385889

Now every one who owns RTX 4090 and RTX 5090 will get this even when they say the connector is designed properly. You know, just to be sure.

I was almost frying my 4080 with the damned CableMod angled connectors. Both V1 and V2 versions melted. Be VERY careful.

 

[LittleBro]

4080? For real?

 

[Werwulf]

People in the market for a GPU, what's your strategy? Skip this generation entirely in the hope that PCI-SIG have yet another attempt at making a connector that won't melt? Do some napkin math to de-rate the 12VHPWR and get a card under that (maybe with an undervolt)? Wait for AIBs to add better safety mechanisms?

From my rudimentary understanding, in worst case scenarios of cable imbalance like Der8auer saw, you're really only looking at around 300w total board power draw (114w per wire/pin?) before things start getting a bit dicey. Interested in seeing some workings if this isn't accurate

 

[ty_ger]

Where? When reading 35 pages is too much, but you want to contribute anyway

What kind of response is that? Did you read his question?

 

[Degreco]

What kind of response is that? Did you read his question?

I referred to the post he responded to. Context is king.

 

[ty_ger]

I referred to the post he responded to. Context is king.

Eh?
Tell me where in 35 pages someone has provided evidence of millions and millions of 4090s being sold. Don't quote my post as the context. lol
Babaganoosh has been active in this thread. Don't tell him to search harder for something that isn't here. Or provide a link if we both are missing it.
Low quality response.

 

[Knight47]

 

[Degreco]

Eh?
Tell me where in 35 pages someone has provided evidence of millions and millions of 4090s being sold. Don't quote my post as the context. lol
Babaganoosh has been active in this thread. Don't tell him to search harder for something that isn't here. Or provide a link if we both are missing it.
Low quality response.

What a facepalm-worthy post. Are you trying to suffocate me?

First off: I never told Babaganoosh - or anyone - to search harder for something. In fact, I’m on the same page as him, but somehow, you completely missed that.
Second: You are the one who failed to search, read, and understand. Honestly, I had a feeling this would happen the moment I mentioned the oh-so-difficult word 'context'.

Context, by the way, is something you build by actually reading.

I know, reading is nerdy, but do yourself a favor and check these out:

• #849
• #853
• #862 (the post I replied to)
• #863

Then come back and provide your 'high quality' response or troll insult. I'll read it and if it’s any good, I’ll even reply to it. Promise

 

[Bomby569]

People in the market for a GPU, what's your strategy? Skip this generation entirely in the hope that PCI-SIG have yet another attempt at making a connector that won't melt? Do some napkin math to de-rate the 12VHPWR and get a card under that (maybe with an undervolt)? Wait for AIBs to add better safety mechanisms?

From my rudimentary understanding, in worst case scenarios of cable imbalance like Der8auer saw, you're really only looking at around 300w total board power draw (114w per wire/pin?) before things start getting a bit dicey. Interested in seeing some workings if this isn't accurate

you'll be fine if you don't buy the high end, and if you do pay those prices your main issue is not the cable melting, if what is wrong with your brain

 

[Werwulf]

I really wish Youtubers would stop talking like 12VHPWR and 12V-2x6 are different cables.

 

[ty_ger]

What a facepalm-worthy post. Are you trying to suffocate me?

First off: I never told Babaganoosh - or anyone - to search harder for something. In fact, I’m on the same page as him, but somehow, you completely missed that.
Second: You are the one who failed to search, read, and understand. Honestly, I had a feeling this would happen the moment I mentioned the oh-so-difficult word 'context'.

Context, by the way, is something you build by actually reading.

I know, reading is nerdy, but do yourself a favor and check these out:

• #849
• #853
• #862 (the post I replied to)
• #863

Then come back and provide your 'high quality' response or troll insult. I'll read it and if it’s any good, I’ll even reply to it. Promise

Gotcha. You don't have his answer and his answer isn't in those posts. Why tell him to pound sand? Just for fun?

Maybe you should use your words better.

 

[95Viper]

Enough, please post on topic per the Guidelines...

 

[SmookinJoe]

Please watch Jayz video it talks about how, which, and why a particular 12pin connector design is better..Good info..
Don't be thrown by the title.

 

[STSMiner]

Please watch Jayz video it talks about how, which, and why a particular 12pin connector design is better..Good info..
Don't be thrown by the title.

I'm not a fan of the female pins that are used in the connector on the cable (dimple with the two seams vs NTK with only one seam).

I've yet to see an RTX 4090 melt the connector where the NTK female pins are used in the connector, and I've yet to see any RTX 3090 Ti have this issue regardless of the power draw here.

The main change that was done was with the PCB header connector as detailed in the ATX 3.x update and the name change to 12V-2*6, nothing else changed here.

One seam in the female pin here.

vs ..... this crap with "dimples"

 

[SmookinJoe]

I'm not a fan of the female pins that are used in the connector on the cable (dimple with the two seams vs NTK with only one seam).

....

I agree with you about not being a fan, of the whole 12 pin system....

 

[Apocalypsee]

This thread has gone from wire safety margin, to shunt resistor, cooling down the actual cable with a fan then latest how to bend the cable properly. We all need to go on EWIS (Electrical Wiring Interconnect System) course before using this connector