Multiple Instances of Native ATX 3.0 12VHPWR Power Connectors Melting Have Now Been Reported

Peter_Brosdahl

Moderator
Staff member
Joined
May 28, 2019
Messages
7,960
Points
113
It now appears that NVIDIA may want to reach out to PSU manufacturers as well now that multiple instances of native ATX 3.0 12VHPWR connectors melting are now being reported and the adapters are no longer the only ones being reported on.

See full article...
 
I'm starting to not really buy into these instances with the melted connectors, and it's not just because I own a 4090. I want to see proof on how the card was installed with the cable(s) plugged into the card and how it looked installed with the cable before they pulled it out and checked the connector. I have a feeling that these "instances" are more on the user end than what is being told to the population. People are also blaming Nvidia for not coming out and addressing the issue, but how can they with such a small amount of cards having issues? Maybe just a generalized statement saying they are collecting data may suffice I guess. Some are also spectualting that other means are being used to melt the connector just because of buyers remorse. Not sure about that theory though myself. Just my opinion.
 
I was thinking about the installation process being a factor as well but it also got me thinking that aside from card dimensions the card manufacturers should also be putting disclaimers about how much more clearance from the card is required for the cable as well. I was looking at one picture (couldn't re-find it because I was really deep in Reddit this morning looking at different ones) and noticed the case side panel was fairly close to the card's power connector port. That can't be a good thing if users are expected to not bend it until after a certain length of the cable. These days most people are more focused on the length or height of a case than they are on the width.

The good news is that yes these numbers do seem to be low (maybe one or two dozen?), for now, and that no one has been injured. I do wonder though how many simply are not aware if any damage has happened, yet and so haven't reported on it.

Buyer's remorse is a possibility but considering how insanely good this card is I'd call them idiots if that was the cause. Even more so if they paid scalpers and have remorse for that.

I do agree though that user error could be the dominating factor here and that there's no proof saying otherwise, just people's word on what they said they did. However, if this thing really is that precarious is it really up to the user to be so careful when a previous design was all but foolproof?
 
Just to add that it is odd this could be happening with the cable and not the adapter. I would think that in the absence of the extra 8-pin connectors it would be easy to install without excessive bending. If people are giving it the angles/space it needs then there is more going on here. If they are obsessively bending it to hide as is common for many enthusiasts, that could easily be the root of it also.
 
However, if this thing really is that precarious is it really up to the user to be so careful when a previous design was all but foolproof?
This I totally agree with. I don't believe they should have stayed with what worked.
Oh, here's the one I was talking about. For them it was a 3x 8-pin adapter.
That isn't ideal unless they do not put the side panel on the case. That looks like a Corsair 4000 or maybe a 5000 series case. My 7000 series case side panel hardly touches the cable.
 
I'm starting to not really buy into these instances with the melted connectors, and it's not just because I own a 4090. I want to see proof on how the card was installed with the cable(s) plugged into the card and how it looked installed with the cable before they pulled it out and checked the connector. I have a feeling that these "instances" are more on the user end than what is being told to the population. People are also blaming Nvidia for not coming out and addressing the issue, but how can they with such a small amount of cards having issues? Maybe just a generalized statement saying they are collecting data may suffice I guess. Some are also spectualting that other means are being used to melt the connector just because of buyers remorse. Not sure about that theory though myself. Just my opinion.
Don't be like that. Just because it didn't happen to you doesn't mean it must be user error. These cards are huge, space in some cases is tight, so the cable will bend in different directions and strain will be put on different parts of it, not to mention the different quality cables going around by manufacturer.

The cable should hold up to a reasonable amount of strain. Even if this cable is fine 99.9% of the time but has problems for .1%, it is still a bad cable that is unsuitable for the application.

For example splicing a bare wire into a molex connector also works most of the time, yet it would not be an acceptable solution out of the factory.
 
The Cablemod right angle adapter should solve most of these issues, but for some reason is till MIA. I signed up to be notified, and so far got to select the kind I need, but no way to order one. I do have the "normal" cable coming from them as well to use with my EVGA PSU. I did order an ATX 3.0 FSP power supply that is on its way, but torn between using that or the Cablemod cable for my existing power supply.
 
Don't be like that. Just because it didn't happen to you doesn't mean it must be user error.
Be like what? Honest? State my opinion? Yeah how dare I do that. Like I said it's my opinion so I'll leave it as that. Not sticking up for Nvidia, and I agree it's a poor design overall, but user error cannot be dismissed.
 
I'm glad for any who do not have to deal with this issue but I really feel NV needs to come up with a better solution. For the most part, nobody is saying it's the board or PSU connectors and it repeatedly seems to be coming back to cables now. Perhaps they offer to ship a right-angle adapter? Maybe something similar to EVGA's? I don't know but it's ridiculous that users were not warned ahead of time that cable clearance would be such an issue.
 
The Cablemod right angle adapter should solve most of these issues, but for some reason is till MIA. I signed up to be notified, and so far got to select the kind I need, but no way to order one. I do have the "normal" cable coming from them as well to use with my EVGA PSU. I did order an ATX 3.0 FSP power supply that is on its way, but torn between using that or the Cablemod cable for my existing power supply.
I didn't post it but did come across a Reddit post from Cablemod saying the pre-order registration thing got messed up. It was buried in one of the reply threads I was in and I don't remember where. Evidently, it's about 4-5 weeks out before they'll begin to receive them. I also registered but never heard anything.

Meanwhile, some are saying good things about this: https://www.fasgear.com/ and https://www.amazon.co.uk/dp/B0B7MGTY56/ref=cm_sw_r_apa_i_7N7JTEPN39P12BPHTV6M_0?_encoding=UTF8&th=1

I've never heard of them before so I don't know about them either but some in those threads are saying good things but I did see one review on Amazon that said it didn't work for them, but 30 that did and gave praise.
 
Be like what? Honest? State my opinion? Yeah how dare I do that. Like I said it's my opinion so I'll leave it as that. Not sticking up for Nvidia, and I agree it's a poor design overall, but user error cannot be dismissed.
Don't assume user error. If I have tight space in my case and as such I have to put more strain on the cable that's not user error. I can reasonably expect the cable to still function, just as my PCIE 8 pin connectors don't fail in tight spaces.

This is an unforseeable error, so you don't need to defend your purchase decision. I'd have assumed NV put utmost effort into these adapters to make sure they work fine, because most of us were against using them from the start. NV dropped the ball bigtime.

The PCiE connectors almost have too much of a safety margin, while these apparently have zero.
 
If its this sensitive, its inadequate, its that simple.
Nvidia didn't want a big fat connector, didn't want tons of little ones either, it was clearly to me an optics over engineering choice. Not to worry, they'll be forgiven in no time.... But lets remember AMD drivers they crashed back in 2015 so I never touched an AMD card again.
 
The 4090 must be doing something different with power vs the 3090, if power overall is the same either it is spiking higher, or overloading a pin or 2 more than the rest. Can't be that everyone had perfect installations with the 3090,/with perfect cables. Also, these are failures of the most niche product in existance, so that divider in the equation is relatively small, and really will remain small.
 
If its this sensitive, its inadequate, its that simple.
Nvidia didn't want a big fat connector, didn't want tons of little ones either, it was clearly to me an optics over engineering choice. Not to worry, they'll be forgiven in no time.... But lets remember AMD drivers they crashed back in 2015 so I never touched an AMD card again.

While you got a point, but my AMD card's drivers were crashing last year when I owned a 5700 XT. Just sayin'
 
This is an unforseeable error, so you don't need to defend your purchase decision.
Defending my purchase is not something I have ever done and I'm not doing that here. I agree that things could have been done better with the connector, but nothing has been proven that anyone dropped the ball. I guess time will tell.
But lets remember AMD drivers they crashed back in 2015 so I never touched an AMD card again.
AMD has never been my go to for cards. Back when they were ATI I used them more. Had a few non driver issues with them (though their drivers weren't great either), and I just didn't look at them any longer. When ATI existed and Omega drivers were a thing those were fun times.
 
The 4090 must be doing something different with power vs the 3090, if power overall is the same either it is spiking higher, or overloading a pin or 2 more than the rest. Can't be that everyone had perfect installations with the 3090,/with perfect cables. Also, these are failures of the most niche product in existance, so that divider in the equation is relatively small, and really will remain small.
I was initially all over the connector (still think it's chitzy) - but yeah, after looking more into it: this. Why wasn't this a thing on the 3090 with almost equal power draw?

And Uvilla has a point = the 4090 is getting a lot of press because A) This, and B) it's new, so it may seem like the cards are all over the place in the wild. But not a lot of people are actually going to be owning one - no matter how great it is, it's still very pricey and very much a niche high-end product. That means that if we are seeing this many failures - and these are only the ones that are getting posted, there are likely more -- it's probably higher percentage than you would think.
 
And Uvilla has a point = the 4090 is getting a lot of press because A) This, and B) it's new, so it may seem like the cards are all over the place in the wild. But not a lot of people are actually going to be owning one - no matter how great it is, it's still very pricey and very much a niche high-end product. That means that if we are seeing this many failures - and these are only the ones that are getting posted, there are likely more -- it's probably higher percentage than you would think.
Higher amount got sold than we might think
https://forums.guru3d.com/threads/nvidia-produced-over-100-000-rtx-4090-units-thus-far.445331/
 
Note that 100,000 units do not equal the total number of GPUs made or sold. Instead, it is the number of chips that were sent to partners so they could make RTX 4090 models.

Maybe.

Somewhere in the high tens of thousands would be what I would expect given where we are in the release cycle, and that's pretty much what your article is saying. The "gamer" market segment is somewhere in the hundreds of millions, and that's discounting the super-niche that are buying this card for other reasons and never happen to game on it (I suspect that's an extremely low number though).

That puts it at around 0.1% penetration, and I would expect to cap somewhere in the neighborhood of 0.5%. Based on historical available data 'n such.
 
Become a Patron!
Back
Top