• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Design analysis discussion of Nvidia-supplied 12VHPWR connectors

Scietist revised his stance in light of new evidence, must be clueless! /s

It probably didn't help his opinion that WCCF was drama farming by outright fabricating parts of their stories on the matter.
Lol wccf aka troll magnet/clickbait site!
 
As long as it's "properly" inserted into the socket nothing will happen. however what I was afraid of happened. If you want to get a 4090 , I suggest wait. don't make a 1700 - 2000 dollar mistake.


 
So if I understand this correctly, 3090ti does not have this problem because it doesnt use as much power as the 4090. Is that the right thinking? I looked at my 3090ti 12VHPWR and it looks fine.

If thats the case, this sounds more like a connector issue. This connector doesnt support that much amps running through it.
 
That's disturbing. It's clearly properly connected.
It seems the manufacturers/nvidia/PSU makers didn't do enough testing....
 
So if I understand this correctly, 3090ti does not have this problem because it doesnt use as much power as the 4090. Is that the right thinking? I looked at my 3090ti 12VHPWR and it looks fine.

That's part of what makes this so hard to solve, and so fascinating. Thousands of users are clicking along with no troubles, yet we have a non-trivial number with failures

If thats the case, this sounds more like a connector issue. This connector doesnt support that much amps running through it.

By spec, it should support twice the experimentally-determined power draw of a typical gaming load, which is 400A400W-ish in testing. Each contact is rated for 13A; 13A * 6 = 78A, 78A * 12V = 936W. So if pin loading is equal, each pin is only drawing a 6A or so. So somehow (I suspect) the pin loading is uneven, which shouldn't be possible if all 12V wires are merged through the "bus bar" in the connector.

EDIT: Typed Amps instead of Watts
 
Last edited:
So if I understand this correctly, 3090ti does not have this problem because it doesnt use as much power as the 4090. Is that the right thinking? I looked at my 3090ti 12VHPWR and it looks fine.
Well, I did notice my EVGA 3090 ti branded adapter is constructed a bit differently from the one in igorslabs report. But yeah I think the lower power levels help too. And ditto, no issues here.
 
It's not how I would have made a connector capable of twice the power of an 8pin.
The smaller and shorter contact areas involved in power transfer have less physical margin for error and less actual contact area, but connectors like these have been in use for some time now, they're not that new in principle but I think they exceeded sensible limits while reducing the over engineering so fails are worse.
I have had issues with 8pin connections tarnishing with heavy use but that used to result in crashes, to me something about Nvidia's new power regulation circuitry is working different, IE perhaps better than it previously was so issues with power spikes and dips are not causing a crash, they roll on exasperating the issue at its power inlet, which in the past would have crashed the GPU.

Possibly dirty contact's or poor coating on the odd one.
 
It's not how I would have made a connector capable of twice the power of an 8pin.
The smaller and shorter contact areas involved in power transfer have less physical margin for error and less actual contact area, but connectors like these have been in use for some time now, they're not that new in principle but I think they exceeded sensible limits while reducing the over engineering so fails are worse.
I have had issues with 8pin connections tarnishing with heavy use but that used to result in crashes, to me something about Nvidia's new power regulation circuitry is working different, IE perhaps better than it previously was so issues with power spikes and dips are not causing a crash, they roll on exasperating the issue at its power inlet, which in the past would have crashed the GPU.

Possibly dirty contact's or poor coating on the odd one.
Everyone who follows my posts knows initially I was a skeptic to all this, but the more that comes to light the more I agree. Just a larger mating surface wouldve given more margin for error or whatever.

Lol wccf aka troll magnet/clickbait site!
Pretty much. I would've believed this much more readily without their shoddy reporting muddying the waters. Pretty confident it did not help Jonnyguru either.
 
For an adapter that is supposed to have limited plug in cycles the current state of things is causing hysteria for everyone to facilitate this limit on constantly checking for burned pins. I am personally running my 4090 at 80% power limit only recently after more and more news came out and already ran the card for 1 month and about 3 weeks without a power limit. Also I have a 750 watt platinum sfx psu. Retuned the Thermaltake 1000 watt gold pcie 5.0 atx 3.0 psu because there was an incident with burnt cable on an MSI 1000 watt atx 3.0 psu as well.

As long as it's "properly" inserted into the socket nothing will happen. however what I was afraid of happened. If you want to get a 4090 , I suggest wait. don't make a 1700 - 2000 dollar mistake.


Gamers Nexus theory is that it might be 150 volt cables vs the 300 volt cables they had. Does your cable say 150 volts?
 
Gamers Nexus theory is that it might be 150 volt cables vs the 300 volt cables they had. Does your cable say 150 volts?

It has nothing to do with the cable. If it was a cable issue, then we would be seeing melting of the cable along any length or part of the cable itself. The problem is the pin contact but most importantly the plastic house not withstanding high temperature.

So the first thing that needs addressing is the build material of the plastic housing. Secondly the build material of the pin contacts themself.

I feel there will be a change, but it may be released without notification to end users. I'm sure most of us have read this before in other products.
 
By spec, it should support twice the experimentally-determined power draw of a typical gaming load, which is 400A-ish in testing. Each contact is rated for 13A; 13A * 6 = 78A, 78A * 12V = 936W. So if pin loading is equal, each pin is only drawing a 6A or so. So somehow (I suspect) the pin loading is uneven, which shouldn't be possible if all 12V wires are merged through the "bus bar" in the connector.
Uneven distribution of current is very much possible. Supposed that each contact point has ~1 mΩ (milliohm) resistance when it's new and clean. The six sockets (hollow parts that mate to the pins) are made of an alloy like bronze or brass, elastic enough to act as a spring and hold the pin with a certain force - until you deform it too much. After some wiggling the grip weakens. One point loses contact completely. Four suffer increased resistance by 3 mΩ, so 4 mΩ in total, and 1 mΩ for all four in parallel. The sixth point remains in perfect condition at 1 mΩ - and it takes up one half of the total current! The temperature rises enough to further deform the thin metal socket, also the process of oxidation speeds up dramatically (compared to cold metal). Poof.

Looking at the pictures I assume that something similar happened.

Thick bus bars on both male and female sides of the connector actually make the situation worse. Equalising resistors on the PCB side could help here - those could be made in the form of short, wide tracks from each solder joint to the common, very low resistance bus on the PCB.
 
Gamers Nexus theory is that it might be 150 volt cables vs the 300 volt cables they had. Does your cable say 150 volts?
Physics wise I don't see why the cables voltage rating would make a lick of difference here. It's always 12 volts, anyways. And even if it was higher, voltage ratings are pretty much irrelevant from a heat perspective, amps are what matter there.
 
Physics wise I don't see why the cables voltage rating would make a lick of difference here. It's always 12 volts, anyways. And even if it was higher, voltage ratings are pretty much irrelevant from a heat perspective, amps are what matter there.
This. Im not sure why people are quoting 300V or 150V. Its 12V coming from the rail....

The important electrical property to measure here is Amps.

Just goes to show how much people think they know.
 
Why didnt they just slant the socket like the 3000 series FE cards? which would have significantly reduced bending.
 
New info....

Great article. Sounds like the Astron made adapter is the most problematic. I wonder who on earth made my weird evga 3-pin with molded evga logos on it?
 
"I have a native cable on the power supply. I already wrote that the adapters should only be a real emergency"

Final statement from Igor.

I read his essay, he avoided blame but dished it out too?!

SIG and users get 90% ,he now gets a middle finger from myself, why because the old 8pin way Just worked.

Whereas the User needs qualifications in plug insertion to not Fffff this up.
 
I mean he's likely trying to avoid dishing out blame too much for lawsuit protection. As you note he's pretty thinly veiling his critique though, I really would not give him a "middle finger" here for covering his ass, his writeups have been very helpful.

Whereas the User needs qualifications in plug insertion to not Fffff this up.
That is the bottom line and I think everyone agrees with this point: this should never have happened.
 
Well, I did notice my EVGA 3090 ti branded adapter is constructed a bit differently from the one in igorslabs report. But yeah I think the lower power levels help too. And ditto, no issues here.

Some even have different wattage numbers too.

It's not how I would have made a connector capable of twice the power of an 8pin.
The smaller and shorter contact areas involved in power transfer have less physical margin for error and less actual contact area, but connectors like these have been in use for some time now, they're not that new in principle but I think they exceeded sensible limits while reducing the over engineering so fails are worse.
I have had issues with 8pin connections tarnishing with heavy use but that used to result in crashes, to me something about Nvidia's new power regulation circuitry is working different, IE perhaps better than it previously was so issues with power spikes and dips are not causing a crash, they roll on exasperating the issue at its power inlet, which in the past would have crashed the GPU.

Possibly dirty contact's or poor coating on the odd one.

What i find funny is the many times re-connections with a 6 or 8 pin connector has never failed on me even after years of use.

Going by how this new connector is my 6 and 8 pin connectors on my 390X should be destroyed by now.

EDIT:
Everyone who follows my posts knows initially I was a skeptic to all this, but the more that comes to light the more I agree. Just a larger mating surface wouldve given more margin for error or whatever.


Pretty much. I would've believed this much more readily without their shoddy reporting muddying the waters. Pretty confident it did not help Jonnyguru either.


Did he not retract his comment on it being user error ?, i believe i heard it some were.
 
Last edited:
Going by how this new connector is my 6 and 8 pin connectors on my 390X should be destroyed by now.
Officially speaking, they are all rated the same (30 connect/disconnect cycles).

But practically speaking, Igorslab notes issues with the finishing quality of the nylon material in the current batch of adapters that do indeed probably make the 8-pins of yore fair better.

I think it comes down to, making things this small just lowers the margins for error way too small to trust to just any OEM. It's like trusting the average PC builder to repair a cellphone: it may not always end well, because lower tolerances etc.
 
Officially speaking, they are all rated the same (30 connect/disconnect cycles).

But practically speaking, Igorslab notes issues with the finishing quality of the nylon material in the current batch of adapters that do indeed probably make the 8-pins of yore fair better.

I think it comes down to, making things this small just lowers the margins for error way too small to trust to just any OEM. It's like trusting the average PC builder to repair a cellphone: it may not always end well, because lower tolerances etc.

If they had done it with the same size pins as the 6 and 8 pin connector were would not be a issue now imo.
 
If they had done it with the same size pins as the 6 and 8 pin connector were would not be a issue now imo.
I agree, edited my posts with some details about why. It's just a tolerances thing. You can't trust any idiot OEM to make something so small IMO. Thus, you either drive costs up unnecessarily, or end up with a fire hazard.
 
If they had done it with the same size pins as the 6 and 8 pin connector were would not be a issue now imo.

Possible or even probable. But we're here to discuss why this connector is failing, not the hypothetical superiority of a different solution.
 
Back
Top