12vhpwr Custom CableMod 12VHPWR Power Adapter Melts on RTX 4090

[BiG StroOnZ]

Summary

After many weeks of silence, melting RTX 4090 cables are back on the agenda. This time the problem affects a custom cable designed and manufactured by CableMod. Apparently, at least one of such adapters failed and melted when used with a GeForce RTX 4090 graphics card, according to a Reddit post. CableMod, the manufacturer of the device, promised to replace the product. 

 

 

 

Quotes

Quote

A redditor named Humble-Brilliant-654 posted a picture of his 12VHPWR cable adapter that was clearly melted. He said that he inserted the adapter correctly until he heard the click and the device functioned without problems for two months.

 

The company behind the custom power cable has already responded in the thread. CableMod appears to be very straightforward in their further responses, suggesting that if the cable is at fault and GPU is no longer working, the company might cover the cost of replacement.

 

“Very sorry about that – please reach out to our support and we will make it right with you! Our products come with service and we help every customer out.”

 

“The card is not damaged (we spoke to him) but in case our cables would damage a card, then we would replace the GPU – we have done that in the past in very rare cases where our cables damaged something.”

 

— CableMod

 

My thoughts

This is not good news, as it seemed like the dust was finally settling with the 12VHPWR adapter. There seems to be some arguments on both sides on whether or not the individual plugged it in correctly. Kyle Bennett from HardOCP on Twitter said, "Don't tell me the user that bought and installed that was not aware of needing to seat it properly". Buildzoid responded to him on Twitter and said, "You can actually see that they probably didn't plug it in all the way on the connector.". There are also other Redditors claiming that you could tell the cable was not inserted properly. However, the affected individual claims he did and it was fine for two months. It's possible it came loose after installation, however, who really knows for certain. Regardless, I'm glad CableMod is handling the situation in a sensible manner. Hopefully, more information arises concerning this dilemma. I will update the thread if more people discover the same thing.

 

Sources

https://videocardz.com/newz/custom-12vhpwr-power-adater-by-cablemod-has-melted

https://www.tomshardware.com/news/cablemod-replacement-12vhpwr-adapter-melts

 

 

[Skiiwee29]

https://twitter.com/GamersNexus/status/1616035465601376258?s=20

 

[Kinda Bottlenecked]

It is possible that the user did indeed have the connector plugged in fully. 

 

 

One thing I've noticed on my 4090s is that the clip is on the bottom side of the connector. So if the GPU is horizontally mounted onto the motherboard, gravity pulls the whole cable downwards. This happens even if you have the 12pin fully seated with a "click".

 

So even if the side of the connector is seated flushed, the top isn't. Over time gravity itself might have loosened the connector. Doesn't help that the pins are so short too. 

 

Definitely do check your cables from time to time and make sure it is plugged all the way in. 

[Quackers101]

10/10 connector, 100% worth for 1000's of dollars.

But yeah its pretty one sided, so I guess one could assume it was "pulled" or if the user just wanted to add to the drama. but might assume thats not the case.

would be nice to see at failure or before pulling out pictures. Also if it was plugged in "correctly" or not. But yeah still think the connector kinda sucks or seem cheap for these highend products.

 

*when they want money from manually destroying some card connectors for fun* jk

Edited January 20, 2023 by Quackers101

[starsmine]

I see the top comment is cable mod 
 

Quote

very sorry about that - please reach out to our support and we will make it right with you!

There is nothing to be made right here, the user plugged it in all of 1/4th the way in. 

fucking 1/4th

[vetali]

Its very obvious it wasn't plugged in enough. The connector is gonna need a locking mechanism that seats the connector properly if people gonna keep messing it up even after the failure point has been identified.

 

Or, just go back to 8 pins.

[Haaselh0ff]

Im shocked people even bother posting these anymore and even more shocked that these still get popular on Reddit. It's proven time and time again that this is user fault so seeing it can only prove the idiocy of the user and not the fault of the CableMod.

[dilpickle]

11 hours ago, Haaselh0ff said:

Im shocked people even bother posting these anymore and even more shocked that these still get popular on Reddit. It's proven time and time again that this is user fault so seeing it can only prove the idiocy of the user and not the fault of the CableMod.

This. 

 

Although I'm not surprised reddit keeps up voting this crap. As much as they claim to hate clickbait they only contribute to the problem. 

[Uttamattamakin]

11 hours ago, Haaselh0ff said:

Im shocked people even bother posting these anymore and even more shocked that these still get popular on Reddit. It's proven time and time again that this is user fault so seeing it can only prove the idiocy of the user and not the fault of the CableMod.

This is users fault for not being more careful in inserting the connector than the teams of highly paid electrical engineers whose whole job is to make these things not possible. 

 

A properly designed connector would have it be not possible at all for this to happen.    At any rate now people know and perhaps PERHAPS they should just buy computers assembled by system integrators or have them assembled by services like those available at MicroCenter and similar stores where for a cost they will assemble a computer. 

[starsmine]

1 hour ago, Uttamattamakin said:

This is users fault for not being more careful in inserting the connector than the teams of highly paid electrical engineers whose whole job is to make these things not possible. 

 

A properly designed connector would have it be not possible at all for this to happen.    

The world always invents a better fool.

No amount of "proper design" would have made this not possible. 

You do know people melt 8 pins to. You know people have managed to shove HDMI into a usb port and go "y dont work" to. 

Yes you can minimize that, but to say this isnt a proper designed connector to me is just wild. All you have to do is plug it all the way in till it latches. 

[BiG StroOnZ]

A theory from Kyle Bennett over at [H]ard|Forum:

 

Quote

Some thoughts on this with everyone chiming in that it was not pushed in all the way.

FWIW, I worked in industrial supply for 10 years before I got into HardOCP, and my primary job was selling thermoplastic hose and tubing, so I have a bit of exposure when it comes to melting plastic.

The only failures we have seen actually happen have been "forced" by the tech journos covering this issue.

My thought is that is it quite possible that failures could be occurring with the connector plugged all the way in. The plug gets super-hot, and due to the plug on the plastic expanding, it is pushing itself out of the socket elongating the capture clip or it coming unclipped, then once it has pushed out as far as it can, it sits there and "cooks," showing the line where the plug has moved out to. Once the connector cools off, you would have no proof of that capture clip elongated or failed.

Just thinking through this, I do not think all the investigation we have seen done on this is conclusive until you could actually see one fail that was fully seated.

That's my 2 cents, you may need change.

 

https://hardforum.com/threads/nvidia-rtx-4090-power-connectors-melting.2022862/page-12#post-1045557219

 

 

[vetali]

19 minutes ago, BiG StroOnZ said:

A theory from Kyle Bennett over at [H]ard|Forum:

I highly doubt it. Its been tested by multiple sources ad nauseam, the connectors don't reach a high enough temperature when properly installed to fatigue the plastics enough to do this. If there is a problem with the connector, its in the locking mechanism for not ensuring a secure lock and allowing side to side play while locked.

[Mark Kaine]

i cant get over it how cheap and flimsy these connectors look...

 

And yes, this is user error,  the error was to buy such a poorly designed piece of hardware in the first place.  Still manufacture responsibility to sell a defective/ dangerous electronical plug in the first place. 

 

 

[Quackers101]

but there has been the saying of the connector location can be in a hotspot.

as with evga and others, not about the connector itself, but why it continues to be in that position, sometimes deep inside the card etc.

[Blademaster91]

30 minutes ago, Quackers101 said:

but there has been the saying of the connector location can be in a hotspot.

as with evga and others, not about the connector itself, but why it continues to be in that position, sometimes deep inside the card etc.

Yeah KINGPIN has said the connector is in a bad spot due to heat, a lot of heat comes from near the GPU die, he talks about that starting at about 4 mins in the video. And I would've liked to see EVGA bring their 40 series cards to the market, because if they did I doubt the EVGA cards would have had any issues with the connector melting as they had it away from the heat, and the connector would have less stress on it being in a different area of the card.

32 minutes ago, Mark Kaine said:

i cant get over it how cheap and flimsy these connectors look...

 

And yes, this is user error,  the error was to buy such a poorly designed piece of hardware in the first place.  Still manufacture responsibility to sell a defective/ dangerous electronical plug in the first place. 

 

 

I agree, I don't see why people would want to spend so much on these cards for what they cost, and considering the risk of the connector coming loose and ruining the card. The 12vhpwr connector is a terrible design if the card can still receive power while the connector is loose, and iirc PCI-SIG didn't want to accept any blame for the design so its on Nvidia to improve the connector.

[Mark Kaine]

10 minutes ago, Blademaster91 said:

iirc PCI-SIG didn't want to accept any blame for the design so its on Nvidia to improve the connector.

Thats the thing though if this PCI-SIG did approve this connector then they're to blame too,  and that Nvidia,  a billion dollar corporation doesn't just go and improves the known flaws in their design is baffling. But in the end this PCI-SIG organization looks really toothless and not reliable at all when their "approval" really means nothing.

 

And what happens to the people who's cards basically became paperweights due to this flaw? I cant really see how the responsible parties are not being held responsible in the end... 

 

It all seems so useless too, this design improves nothing,  the old 2x8 design looked better and obviously had to deal with less strain too, such an unnecessary own goal by Nvidia .

 

22 minutes ago, Blademaster91 said:

Yeah KINGPIN has said the connector is in a bad spot due to heat, a lot of heat comes from near the GPU die,

yeah,  the cards also have an option to have the power connector on the side which isnt only not as hot but puts less strain on the connector (theoretically)... just baffling.

[HenrySalayne]

3 hours ago, Mark Kaine said:

It all seems so useless too, this design improves nothing,  the old 2x8 design looked better and obviously had to deal with less strain too, such an unnecessary own goal by Nvidia .

12VHPWR = 600 W

8P PCIe = 150 W

6P PCIe = 75 W

"improves nothing"

[AlTech]

There is a simple solution to this problem: Just buy an AMD GPU instead.

 

If you want to gamble that your GPU cable will melt and that your GPU will live or die, then the 4090 and other GPUs with the same connector are the way to go.

8 hours ago, starsmine said:

The world always invents a better fool.

No amount of "proper design" would have made this not possible. 

No but it would have reduced the likelihood significantly.

 

Nvidia is ultimately in the wrong for not designing these connectors well enough.

8 hours ago, starsmine said:

You do know people melt 8 pins to.

That is genuinely their own fault though. They have all the tools and info they need to not cause that to happen.

 

This is not true of the 12VHPWR connector.

8 hours ago, starsmine said:

You know people have managed to shove HDMI into a usb port and go "y dont work" to. 

And this is nothing like that.

8 hours ago, starsmine said:

Yes you can minimize that, but to say this isnt a proper designed connector to me is just wild. All you have to do is plug it all the way in till it latches. 

If there isn't sufficient audible or visual feedback for it then that's the connector's fault, not the user's fault.

[mr moose]

Honestly a this point in tech evolution users are becoming so dumb we need to go back to D-sub style connectors with screw fasteners on each side. 

 

Personally I'd be happier with a power connector more akin to an automotive grade one that can carry twice the required power, won't come disconnected even in an earthquake and has the added bonus of being IP6X rated.   Does the size really matter if you risk breaking everything and setting fire to your house because you're a moron?

[starsmine]

3 hours ago, AluminiumTech said:

And this is nothing like that.

If there isn't sufficient audible or visual feedback for it then that's the connector's fault, not the user's fault.

I dont know why people keep making this point like it makes sense

The audio feed back could be a fucking gun shot and would not mater. 

Why? because the feedback ONLY occurs when its pushed in all the way. 
none of these melted cables even GOT to that point. 

 

3 hours ago, AluminiumTech said:

.That is genuinely their own fault though. They have all the tools and info they need to not cause that to happen.

They have all the tools and info here that they need to not cause this to happen. 
It is genuinely the users own fault. The user did not plug in the cable all the way. 

 

  

9 hours ago, Blademaster91 said:

Yeah KINGPIN has said the connector is in a bad spot due to heat, a lot of heat comes from near the GPU die, he talks about that starting at about 4 mins in the video. And I would've liked to see EVGA bring their 40 series cards to the market, because if they did I doubt the EVGA cards would have had any issues with the connector melting as they had it away from the heat, and the connector would have less stress on it being in a different area of the card.

I agree, I don't see why people would want to spend so much on these cards for what they cost, and considering the risk of the connector coming loose and ruining the card. The 12vhpwr connector is a terrible design if the card can still receive power while the connector is loose, and iirc PCI-SIG didn't want to accept any blame for the design so its on Nvidia to improve the connector.

I Wish you time stamped your post when he brought that up.
I do remember him talking about that, but I dont remember where

iirc it had nothing to do with the melting cables and everything to do with the power chips that are right after the connection lose efficiency and efficacy the hotter they are. 

The connection get hot and melt when they make a low resistance parallel path for power that should not be there, not from being warm from the gpu being warm. 

[BiG StroOnZ]

13 hours ago, vetali said:

I highly doubt it. Its been tested by multiple sources ad nauseam, the connectors don't reach a high enough temperature when properly installed to fatigue the plastics enough to do this. If there is a problem with the connector, its in the locking mechanism for not ensuring a secure lock and allowing side to side play while locked.

 

Didn't Gamers Nexus show that when the connector is pulled at a certain angle, a few of the connectors can lose congruence, and then all of the current passes through the remaining connections?

 

Then I believe from there, Gamers Nexus showed you can then get a fair bit of heat accumulation. Where the plastic weakens just enough for the clip to give out, it falls to the fewer connected pins, overheats and then as a result of all of this the connector eventually melts.

 

Doesn't this mean, under certain scenarios, it's possible for the connectors to reach a high enough temperature? Meaning, one tolerance fault within the connector, could cause potential issues. As you said, as an example, if the locking mechanism is not secure enough. There could be enough play, simulating the above scenario; where it then can cause it to heat up, eventually leading to it melting.  

 

Just a thought. 

[GoodBytes]

They should just make the connector in metal! Problem solved, no more melting ones. Done! Ship it! GoodBytes the super genius problem solver!

 

 

[vetali]

1 hour ago, BiG StroOnZ said:

 

Didn't Gamers Nexus show that when the connector is pulled at a certain angle, a few of the connectors can lose congruence, and then all of the current passes through the remaining connections?

 

Then I believe from there, Gamers Nexus showed you can then get a fair bit of heat accumulation. Where the plastic weakens just enough for the clip to give out, it falls to the fewer connected pins, overheats and then as a result of all of this the connector eventually melts.

 

Doesn't this mean, under certain scenarios, it's possible for the connectors to reach a high enough temperature? Meaning, one tolerance fault within the connector, could cause potential issues. As you said, as an example, if the locking mechanism is not secure enough. There could be enough play, simulating the above scenario; where it then can cause it to heat up, eventually leading to it melting.  

 

Just a thought. 

Its not going to happen, and I don't think GN ever concluded that. They even cut off pins in hopes of burning up a connector and the temperatures remained normal. 40-50 degrees Celsius should not create any thermal fatigue on a plastic designed for this type of current draw.

 

The core problem with the connector remains best shown in this clip:

 

 

[starsmine]

2 hours ago, BiG StroOnZ said:

 

Didn't Gamers Nexus show that when the connector is pulled at a certain angle, a few of the connectors can lose congruence, and then all of the current passes through the remaining connections?

Yes but the rest of it was built strong enough to handle a couple of pins not sending power. While yes loosing all but 2 power pairs would probably make it fail long term, that has not been what has happened in any of the fail states shown so far.  you can see that test done at time stamp 11:50 of the gamers nexus video posted above

2 hours ago, BiG StroOnZ said:

Then I believe from there, Gamers Nexus showed you can then get a fair bit of heat accumulation. Where the plastic weakens just enough for the clip to give out, it falls to the fewer connected pins, overheats and then as a result of all of this the connector eventually melts.

 

Just a thought. 

I want to be explicitly clear on this point.
No GN did not show that to be true in any of their testing, data or conclusions. The clip has not "given" out on any of these failures. 
if the connector is clipped in it does not get that hot because there are enough low resistance parallel paths that total resistance is near zero and no heat build up happens. 

[BiG StroOnZ]

2 hours ago, vetali said:

Its not going to happen, and I don't think GN ever concluded that. They even cut off pins in hopes of burning up a connector and the temperatures remained normal. 40-50 degrees Celsius should not create any thermal fatigue on a plastic designed for this type of current draw.

 

The core problem with the connector remains best shown in this clip:

 

 

Sorry, I might have structured the sentencing in a manner that might have been confusing.

 

My stance is as follows ~

 

I believe he comments on it here (starts at 7:07, ends at 7:25):

 

Quote

"The last point we had as a potential failure, is improper insertion in combination with a taut wire on one or more pins; causing at least one point of poor contact, that heats rapidly and melts as a result of a parallel high resistance conductive path."

 

And again here (starts at 14:10, ends at 14:42):

 

Quote

"The most likely cause in at least the viewer cable that we're looking at that failed, is heat generated through a parallel high resistance conductive path. In other words, in addition to the normal and intended dimple to pin connections across the mating surfaces, there also was some weak unintentional connection with high resistance along the same path. This can occur on the NTK Tulip style connectors as well. A high resistance means heat and multiple cycles of that means eventually you're going to have a toasted connector."

 

 

And lastly, here (starts at 22:25, ends at 23:20):

 

Quote

"So our conclusion here, our database one, is that cables are melting when the connector is partially unseated. We've finally been able to recreate it, but it required being very unseated and also pulling the cable at an angle because we tested it unseated without pulling it at an angle. It didn't fail even when it was equivalently unsocketed. So pulling it at the angle so that the GPU's metal pins are in contact by the tip of the metal socket rather than the intended mating surface is what caused it on the two failures we did tonight. In our experiments this happened at the inner radius of the bend. This shares some ground with the FOD Theory as well. Which is a weak high resistance path for current and parallel with the intended path. As stated by our lab contact, 'This partial insertion and angling of the pin into the socket could have increased susceptibility for a high resistance parallel connection at the lip of the socket.'"  

 

 

Steve also mentions foreign object debris as a root cause. This is what I meant by faults within the connector. He seems to verify this from manufacturer defects within the connector causing debris. Also, in the previous post. After I said Gamers Nexus shows how and when the heat accumulation begins. The theorizing begun. Probably should have pointed that out better. And Steve does comment that the connector itself doesn't really click, even when "fully" inserted. This means, if there are tolerance issues, and manufacturer defects; there might be a possibility that even connectors that are clicking still can fail. I give a more elaborate explanation to starsmine below.

 

1 hour ago, starsmine said:

Yes but the rest of it was built strong enough to handle a couple of pins not sending power. While yes loosing all but 2 power pairs would probably make it fail long term, that has not been what has happened in any of the fail states shown so far.  you can see that test done at time stamp 11:50 of the gamers nexus video posted above

I want to be explicitly clear on this point.

No GN did not show that to be true in any of their testing, data or conclusions. The clip has not "given" out on any of these failures. 
if the connector is clipped in it does not get that hot because there are enough low resistance parallel paths that total resistance is near zero and no heat build up happens. 

 

Check above reply to vetali. Maybe that clarifies my stance. You are correct though, GN did not show that in their testing. What I meant was they showed when and where the heat accumulation begins. After that the theorizing began. What I meant was: if the connector is being pulled at a certain angle, and even if the clip is connected fully. If there is a tolerance issue. Perhaps it can be just enough for the clip to release or let go. Which then results in not all the pins being fully seated, the connector overheating and eventually melting. Like so:

 

 

Steve mentions part of the potential design oversight is that the cables don't really click (18:28-18:34):

 

 

Thus, he also mentions potential debris can cause a failure as a result of manufacturer defects within the connector. If this is a viable answer, it's very well possible that people might have connectors that do indeed click, but still aren't fully seated due to tolerance issues. This can cause the cable to eventually heat up, combined with the strain or pulling of the cable at an angle. And the fans ramping up and down constantly from load loosening the connector futher. Eventually, it can wiggle the connector completely loose and become "unclipped". From the partially unseated element, pulling at an angle element, and other jostling taking place from the generic vibrations occurring in a computer case. This can recreate the scenarios Steve found in his research, and eventually cause the connector to heat up and melt. 

 

I could be wrong with all of this, but based on Steve's experiments and research, it seems very plausible, IMO. I'm simply trying to take the stance that this isn't only PEBKAC, which seems like is the easy route in this argument.