Nvidia RTX 4090 power connectors melting?

Kinda a odd issue. A decent 8 pin(6+2) can push 300w all over 3 12v pins. Nvidia could have made a 12 pin of the same form factor that could do an excess of 600w. I'm also tired of seeing 4 6+2 going into these harness. Its unessisary, if you're willing to push a small custom connector close to its limit no reason to be so ridiculously cautious on the other end of the connector especially when all the 6+2 are normally tied to the same 12v rail.

If you are making a custom connector use copper tabs like server nodes. That can push over 1kw reliably and could be fed with just 2 high strand count heavy gauge wires. Then sell adapters over a foot or so letting you hide the 6+2 pin connector stupidness.
Well, see if AMD has a better design for their version of the 12VHPWR connector as their new generation of cards is using it too.
 
For those using this type of connector, check the temperature. At power, gaming, 3dMark Royale looping etc. pushing the power -> Use your senses, smell, sight and touch. Go along the cables towards the connector (to prevent a burn) and see if the plug is getting excessively hot. Surprised not seeing thermal imaging of this online at power with time vs temperature at different power levels, bend radiuses etc. Maybe that will come.

Now I would make up some form of support to keep the force off of the plug/pins, if force is pushing one way on the pins, the pins would be making contact harder on one side (smaller area) and less on the other. If cable/plug, is secured, no lateral/up or down force and not going to move around while opening and closing the case and other things we like to do inside the case and the connection is cool or relatively cool (what is the expected temperature, max temperature of this design? Ridiculas not having this information). I would say not to worry, just check the temperature periodically.
 
For those using this type of connector, check the temperature. At power, gaming, 3dMark Royale looping etc. pushing the power -> Use your senses, smell, sight and touch. Go along the cables towards the connector (to prevent a burn) and see if the plug is getting excessively hot. Surprised not seeing thermal imaging of this online at power with time vs temperature at different power levels, bend radiuses etc. Maybe that will come.

Now I would make up some form of support to keep the force off of the plug/pins, if force is pushing one way on the pins, the pins would be making contact harder on one side (smaller area) and less on the other. If cable/plug, is secured, no lateral/up or down force and not going to move around while opening and closing the case and other things we like to do inside the case and the connection is cool or relatively cool (what is the expected temperature, max temperature of this design? Ridiculas not having this information). I would say not to worry, just check the temperature periodically.
Maybe somebody will release a water block to cool the connector.
 
Kinda a odd issue. A decent 8 pin(6+2) can push 300w all over 3 12v pins. Nvidia could have made a 12 pin of the same form factor that could do an excess of 600w. I'm also tired of seeing 4 6+2 going into these harness. Its unessisary, if you're willing to push a small custom connector close to its limit no reason to be so ridiculously cautious on the other end of the connector especially when all the 6+2 are normally tied to the same 12v rail.

If you are making a custom connector use copper tabs like server nodes. That can push over 1kw reliably and could be fed with just 2 high strand count heavy gauge wires. Then sell adapters over a foot or so letting you hide the 6+2 pin connector stupidness.
The issue is they have to factor in every 8 pin pcie cable in existence. If it fits, it has to work and not burn your house down. Afaik there aren't any definitive PCIE specs other than 75w for 6 pin, and 150w for 8 pin. It's why ~400w+ Ampere cards all had 3x 8 pins (~150w each as a safe limit). Some setups can push 200-300w with fresh crimps/pins and 16awg, but then you run into the issue of buying specific cables/PSU's for a certain GPU to go beyond the 150w "limit". Which is a different headache of itself.

As someone with a ton of real world experience (mining), typical 18awg 8 pins will start to get warm around 150-200w. Dangerously warm? Generally no, but enough to feel and see on a thermal camera. If gpu mfg's started pulling 300w through 8 pins, they would melt down left and right.

That said, 4x 8 pin pcie for a 525w+ card is another headache, so I get the appeal of a new/smaller connector rated to 600w. But it comes at the cost of safety factor, and when operated slightly outside of "ideal" conditions, it turns into a thermal negative feedback loop and melts down. It's clear the wiring/pins can handle the amps at their rated load, in ideal conditions... but if the cable is susceptible to having poor pin contact when bent too tight (or close to the pins), they should be designed to prevent pin shift/extreme angles on the pins (ASSuming that's the issue at hand here, with these failures).

Side note, another one melted: https://old.reddit.com/r/nvidia/comments/yeb9e7/i_think_my_asus_4090_tuf_oc_adapter_is_melting_too/
 
Last edited:
I just don't understand why they didn't make the pigtail thing a hardened right-angle connector from the get-go. You're telling me no one discovered this in testing?

...), so take them with a grain of salt—we don't know the exact configuration of either user's PC setup. The specific model of graphics card (a Gigabyte RTX 4090 Gaming OC for one user, an Asus RTX 4090 TUF Gaming OC Edition for the other), the power supply, and any number of other factors could have contributed to the connectors overheating.....

Article even states they really dont have enough info to make any real claims....
But people dont read articles to...
 
My Rosewill Challenger-S case has a detente on its side panels which provide adequate space for the cable -- at least in my judgment. Taking a quick look, it doesn't look like a lot of other cases have this. I'm grateful to not have to be too concerned about this issue. I like this Rosewill case and can vouch for it, having built at least four systems with it. It's pretty basic I guess but has always done the job for me. Not too expensive at $79.99.

rosewillchals.jpg


I just don't understand why they didn't make the pigtail thing a hardened right-angle connector from the get-go. You're telling me no one discovered this in testing?
It is somewhat baffling. Everything must just be on a bench in testing or something.
 
Shitty design.

A card that pulls that much power needs meatier pins.
 
Shitty design.

A card that pulls that much power needs meatier pins.
or new 24v standard which would decrease the current by 1/2 and thus heat. It is not the power the cabling is handling that heats up the plug but how many amps is going through it. The wire, pins, sockets all have resistance which restricts current flow and converts electrical energy into heat. The more current flow, the more heat generated.
 

That was more a plug (no pun intended) for cablemod. He talked more about cablemod then actual tests. He bent the cable by hand, the wires at the plug, did not put tension on the cable, as in bending coming from hitting the side of the case putting a constant force on the pins. I don't think he understands bent wires is not the issue, it is force on the pins from the wires that maybe the issue, which in his example he did not do.
 
Not sure if anyone 's seen this

https://www.igorslab.de/en/adapter-...ers-and-makes-damage-an-absolute-chief-issue/

Adapter gate? NVIDIA briefs all board partners this morning and makes damage an absolute boss issue​


The adapter in question is said to come from Astron, the company that also produces and sells such plugs itself. This has a stale aftertaste in that the engineers there should have known better. Such soldering processes from thick to thin, as well as pressed wires without real strain relief in the plug, are technically not only mischief and grossly negligent, but actually also forbidden, because they contradict all rules of electrical engineering. NVIDIA has not yet spoken publicly, but has already contacted all affected board partners. Since I have the opportunity to directly ask the responsible persons in the large companies, I could already catch one or the other statement.

nVIDIA just notified all AIC this morning… All damaged cards need to be sent directly to HQ for failure analysis, this is first time… Even a few years ago when 2080 Ti got issue with Micron, they didn’t do this.


So it will be interesting to see exactly what happens now in response. A recall and replacement of the faulty adapter would actually be logical, because this problem can no longer be sat out. And no, it is not because of the propagated 12VHPWR, but because of the lousy implementation by NVIDIA’s supplier.

Earlier analysis by Igor:

NVIDIA’s hot 12VHPWR adapter for the GeForce RTX 4090 with a built-in breaking point​

https://www.igorslab.de/en/adapter-...hot-12vhpwr-adapter-with-built-in-breakpoint/
 
If I get a 4090 or 4080, I'll be using a straight cable like I have on my 3080Ti.
this is the EVGA one that goes straight into my EVGA PSU.
IMG_3088.JPEG
 
solution: be me. don't buy new cards. use GT 1030 that can be powered by PICe power only. overclock. play on 1080p screen. only play AOE.
benefits: don't catch on fire. play only game worth playing. extra monie. hail AOE.
 
Yeah, even if I was in the market right now for a card, I am not installing a fire hazard into my PC, that I often leave running while I am doing other stuff. The card should just be drawing flames on my computer screen, not actually creating them while it's running.
 
It won't have all those thick cables attached to the plug like you get with the adapter. From pictures I've seen the cable looks more flexible. It will hopefully not put the same strain on the plug.
View attachment 521654View attachment 521655

If nVidia had made the wires longer on the adapter, another 3 to 4 inches, or about 7 to 10 CM, it would have helped. People are going to want to tuck away the 4 connectors if possible. With more wire to do that, the wire bends would not be very severe if there was more cable to work with.
Anyone who owns a 4090 of any brand, get the right angle cable from Cablemod once it comes out.
The ones that have melted did so because the cable was pulled on unevenly, which makes some of the connectors make weak contact. A 12vhpwr connector made from thicker plastic that grabs better I think could have helped with that too.
 
If nVidia had made the wires longer on the adapter, another 3 to 4 inches, or about 7 to 10 CM, it would have helped. People are going to want to tuck away the 4 connectors if possible. With more wire to do that, the wire bends would not be very severe if there was more cable to work with.
Anyone who owns a 4090 of any brand, get the right angle cable from Cablemod once it comes out.
The ones that have melted did so because the cable was pulled on unevenly, which makes some of the connectors make weak contact. A 12vhpwr connector made from thicker plastic that grabs better I think could have helped with that too.
Increasing the length of the adapter is going to increase resistance, which would potentially make the issue worse.
 
If nVidia had made the wires longer on the adapter, another 3 to 4 inches, or about 7 to 10 CM, it would have helped. People are going to want to tuck away the 4 connectors if possible. With more wire to do that, the wire bends would not be very severe if there was more cable to work with.
Anyone who owns a 4090 of any brand, get the right angle cable from Cablemod once it comes out.
The ones that have melted did so because the cable was pulled on unevenly, which makes some of the connectors make weak contact. A 12vhpwr connector made from thicker plastic that grabs better I think could have helped with that too.
I would argue that would cause people to pull on it more as they try to get it to the ideal position in the back to tie them down. The adapter itself is poorly designed and they need those solid right angle connectors ASAP, preferably with a longer lead.
 
My personal experience is not necessarily with nvidia 12pin power connector, but with wide molex micro-fit connectors in general. Correct me if I'm wrong, but this looks like a micro-fit connector. I believe i was using a 10pin or maybe 12pin microfit. My problem with them is that if you aren't careful plugging/unplugging them and get them at an angle is that very edge connectors can be become extremely loose/damaged and make very poor contact. I had this occur with 3 separate cables.

To be more specific the female metal recepticle is bent and "expanded" by the male metal connector when the male side is angled and forced in/out.
In my case it was very obvious because those specific connections were actually psu-motherboard "PS_ON" and/or "PWR_OK" signals, which means if i touched the cable my whole system would shutdown.
The fix wasn't very difficult, required some tweezers and squeezing the female parts back together.

I wonder if this is a similar issue, it seems like the edge connectors are overheating?
 
I never understood them out on the side like that. I've always had to kick them down sharply. You'd think it'd be top or bottom.
 
I never understood them out on the side like that. I've always had to kick them down sharply. You'd think it'd be top or bottom.
Or on the side facing the front of the case. Literally anywhere is better than what the standard has been.
 
Id like to order the [H] edition please? With solder pads on the side instead of connector, k thanks!
 
Guess a benefit of the 3000 series FE cards was the connector was angled at 45 degrees.
 
Id like to order the [H] edition please? With solder pads on the side instead of connector, k thanks!
I think every GPU is [H] edition. Just ignore the connector (desolder it if you want) and solder straight onto the pads :D. If you find a couple of GND and +12V planes you can also scrape off the solder mask and solder straight to the planes too to really minimize resistance.
 
If nVidia had made the wires longer on the adapter, another 3 to 4 inches, or about 7 to 10 CM, it would have helped. People are going to want to tuck away the 4 connectors if possible. With more wire to do that, the wire bends would not be very severe if there was more cable to work with.
Anyone who owns a 4090 of any brand, get the right angle cable from Cablemod once it comes out.
The ones that have melted did so because the cable was pulled on unevenly, which makes some of the connectors make weak contact. A 12vhpwr connector made from thicker plastic that grabs better I think could have helped with that too.

While a better short adapter would be an improvement; if you have a modular PSU a new full length power cable that plugs into 2 or more of the outport ports would be a better option and eliminates the chunky and ugly extra connections entirely. That's dependent on your PSU maker or a trustworthy third party offering one.
 
Increasing the length of the adapter is going to increase resistance, which would potentially make the issue worse.

Resistance from an extra few inches of wire is going to be minimal; and spread out over the entire length of the wire in any event. The plug to plug and (if poorly made) plug to wire joints are going to have significantly more resistance; and will be the worrying hotspots if anything goes wrong.
 
Increasing the length of the adapter is going to increase resistance, which would potentially make the issue worse.
To me that seem a vast misunderstanding of the issue (load imbalance), but it is not like things are particularly clear.

1 meter of copper AWG16 would add around 0.0132 ohm (13.2 mohm a 5 cm more would be quite little (0.66) and the same among all electrical path) according to this:
https://www.daycounter.com/Calculators/AWG.phtml

Bent connector pin can go in the 15 to over 800 mohm and sent all the current via a single one that achieved to stay under 10mohm.
 
Last edited:
Back
Top