NV 12VHPWR issues revisited

[amenx]

GN Steve talked about it again recently.

Possibly the cablemod adapter. Unlike most other adapter-cable types, it has 12VHPWR male & female connections.

 

[aigomorla]

I honestly don't care what nvidia says but the 12VHPWR is a FAILED design.
It should NEVER catch fire, even if you put it in wrong.
There should be Many safety's for stuff like that.
And when you have enough fires going, you should RECALL.

But that 12VHPWR there is no fix for it besides checking it constantly.
Hell i check my connector weekly to make sure its not loose.

Its absolute BS.

Im waiting for a fire to start, and a house to burn down, then Nvidia might get off there high horse and redevelop that connector to something more stable.

But yeah its stresses me out that if i accidentally tug on it, and it lodges loose a little, i could essentially lose everything in my PC, all because of a poorly engineered connector and power delivery.

 

[A///]

The connector wouldn't be so bad if it didn't take industrial levels of force to make sure it's secure. instead it never is and creates an electrical hazard. I remember early 8 pins being stiff to work with and they got better over time. still stupid.

 

[Aapje]

But yeah its stresses me out that if i accidentally tug on it, and it lodges loose a little, i could essentially lose everything in my PC, all because of a poorly engineered connector and power delivery.

Fortunately we haven't seen any actual fires. The material seems to have a high enough flash point that it merely melts.

 

[DeathReborn]

I honestly don't care what nvidia says but the 12VHPWR is a FAILED design.
It should NEVER catch fire, even if you put it in wrong.
There should be Many safety's for stuff like that.
And when you have enough fires going, you should RECALL.

But that 12VHPWR there is no fix for it besides checking it constantly.
Hell i check my connector weekly to make sure its not loose.

Its absolute BS.

Im waiting for a fire to start, and a house to burn down, then Nvidia might get off there high horse and redevelop that connector to something more stable.

But yeah its stresses me out that if i accidentally tug on it, and it lodges loose a little, i could essentially lose everything in my PC, all because of a poorly engineered connector and power delivery.

What fires have been started by this? If you're going to go all "we're doomed" at least point to to actual fires. As for the design, HP & Intel should get your ire yet you leave them out, why?

 

[aigomorla]

What fires have been started by this?

Well i consider that very close to the same thing as starting a fire.
It reminds me of the cheap SATA adapters, which also melted and could catch fire like this.

Do you remember this?

 

[H T C]

The connector wouldn't be so bad if it didn't take industrial levels of force to make sure it's secure.

Not only that: the positioning of the connector, and this is SOLELY on nVidia, combined with that means the side panel of the cases exerts unnecessary force on the cable (it's why there's a need for those 90º cablemod adapters to begin with), thus potentially exacerbating the problem.

As for the design, HP & Intel should get your ire

Good question.

I've not seen this design's creators get called out for this potential flaw AS OFTEN as they should. Maybe they have and we just haven't heard of it ... or maybe not ...

 

[aigomorla]

HP & Intel should get your ire yet you leave them out, why?

IF there responsible for that horrible design, then include them on my hate list.
But I do not think Intel would even come out with a plug design like that, as none of there gpu's would even dream of drawing that much power.

As for HP.... if it was HP, it would only fit HP, because HP likes to pretend its apple and make things only work in other HP.
Well for the enterprise sector that is.

 

[amenx]

Not only that: the positioning of the connector, and this is SOLELY on nVidia, combined with that means the side panel of the cases exerts unnecessary force on the cable (it's why there's a need for those 90º cablemod adapters to begin with), thus potentially exacerbating the problem.

The cablemod adapter doest look assuring to me. It has 2 x 12VHPWR connections (male/female). Which means that a 3rd adapter cable may be needed to connect it to the pcie 8 pins. Lots of resistance points already. Maybe no coincidence that the recent failures in the vids all involved that same adapter.

Safer bet imo is to go for one of these:

ATX 3.0 PCIe 5.0 600W 3 x 8 Pin to 12VHPWR 16 Pin Angled Adapter Cable

Buy ATX 3.0 PCIe 5.0 600W 3 x 8 Pin to 12VHPWR 16 Pin Angled Adapter Cable for $19.99 with Free Shipping Worldwide (In Stock)

www.moddiy.com

 

[DeathReborn]

IF there responsible for that horrible design, then include them on my hate list.
But I do not think Intel would even come out with a plug design like that, as none of there gpu's would even dream of drawing that much power.

As for HP.... if it was HP, it would only fit HP, because HP likes to pretend its apple and make things only work in other HP.
Well for the enterprise sector that is.

Intel, HP & Nvidia were listed at PCI-SIG as the designers, the original 12 Pin with no Sense Wires was on Nvidia though. EPS12V would have been a better bet, 8 Pin & 336W.

 

[A///]

Not only that: the positioning of the connector, and this is SOLELY on nVidia, combined with that means the side panel of the cases exerts unnecessary force on the cable (it's why there's a need for those 90º cablemod adapters to begin with), thus potentially exacerbating the problem.

this is something i blame on atx case standards. there is no width standard when there should have been one implemented at some point. if not the 4090 width the width of some hsfs have been ginormous over the last decade.

 

[H T C]

this is something i blame on atx case standards. there is no width standard when there should have been one implemented at some point. if not the 4090 width the width of some hsfs have been ginormous over the last decade.

I disagree: nVidia orients the connector to the outside of the case, which conflicts with the case's side panel when a cable is hooked up to it without an adapter like cablemod's. Had they oriented the connector up or down instead, this issue would be irrelevant.

 

[A///]

I disagree: nVidia orients the connector to the outside of the case, which conflicts with the case's side panel when a cable is hooked up to it without an adapter like cablemod's. Had they oriented the connector up or down instead, this issue would be irrelevant.

i believe an engineer commented on this and said the sheer force to snap it off or damage it would be less required than the side. to move it to the front like their pro cards would pose no problem and make the fitting both save and secure, especially if nvidia weren't cheap bastards and included and required a 90 deg kit.

you want the connector on a card to be parallel with the board, not 90 to it. the mobo power ports are more securely soldered and attached than the flimy connector on a gpu.

 

[H T C]

you want the connector on a card to be parallel with the board, not 90 to it. the mobo power ports are more securely soldered and attached than the flimy connector on a gpu.

Considering the size of "the brick", that would actually be preferable.

 

[A///]

Considering the size of "the brick", that would actually be preferable.

now you see why your suggestion wasn't great. i would have preferred the on power location to be at the front like the pro cards and some od the real old aib cards from the 2000s. if you refer to a jay or the chunky guy who does the sciencey stuff and looks like a troll doll with his metal hair is they had been given some sample rtx 4000 cards by evga for personal evaluation and evga had moved the power connector to the front.

assuming you are not a young buck you would remember tower designs were made in the past with 5.25 in mind and all had a structural brace near the front to limit warp. that brace also housed the lining for the bays, those would jut up against the wireing of anything going to the board be it power, sata, or even making it difficult for front powered cards if the case length was not big enough. most if not all mid towers had this problem and even some full towers if the brace was too much. often if you had a full tower with eatx support it was easy but you were still twisting wired around until power on the side became standard. now that we've mostly ditched front embelishments we can move the power back to the front between the intake fans and a radiator and still have room except in smaller cases but in such a case a 90 deg connector works nicely.

and what is the point of all this you may find yourself asking? a front connector with or without 90 deg adapters places far less stress on the bulky cable coming in. ideally psus should be designed with a native power port or take corsair's better approach with their beefy design 8 pins able to carry 300 watts each in a slick cable. 2 8pins to the 12vhpwr with their older psus and even the new atx 3 pcie5 compliant rmx shift. the shift is a fantastic bit of kit but I highly recommend picking up some diy simple to put on sleeves to protect the wiring plastic if you've got sharpc orners in your case.

 

[alcoholbob]

It turns out that the cards all came from Cablemod and were sent to this one repair guy for repair. They were also saved up for a while, so it doesn't seem that the problem is as big as the repair guy thought.

According to Northridge's latest video (around 10:40 mark), they said they have around 250 RTX 4090s and 4080s in queue for repair. Only eight of them are from Cablemod. So that debunks this claim that it's a Cablemod issue.

 

[MinervaLaw]

According to Northridge's latest video (around 10:40 mark), they said they have around 250 RTX 4090s and 4080s in queue for repair. Only eight of them are from Cablemod. So that debunks this claim that it's a Cablemod issue.

My fav part of that video is how he's talking about how easy it is to plug in the connector, but they have a video jump-cut around 00:46-:00:49 because he couldn't plug the connector in cleanly, lol. Jeeze, seriously?

Honestly? People really should quit using YouTube as "proof" or "evidence", right or wrong, that as something that supports their facts. The data is complete BS, unless backed up with real data that is verified.

I have no pony in this race as I only own (3) 3090s and a couple of M2 macs, but I'd never buy into a YouTube video as proof of anything. He may have 250, but of how many vendors and millions sold? Insignificant? Who knows the REAL statistics.

 

[alcoholbob]

Double post

 

[alcoholbob]

My fav part of that video is how he's talking about how easy it is to plug in the connector, but they have a video jump-cut around 00:46-:00:49 because he couldn't plug the connector in cleanly, lol. Jeeze, seriously?

Honestly? People really should quit using YouTube as "proof" or "evidence", right or wrong, that as something that supports their facts. The data is complete BS, unless backed up with real data that is verified.

I have no pony in this race as I only own (3) 3090s and a couple of M2 macs, but I'd never buy into a YouTube video as proof of anything. He may have 250, but of how many vendors and millions sold? Insignificant? Who knows the REAL statistics.

Well nvidia claimed this affected less than 50 units worldwide so its at least an interesting data point contradicting the manufacturer.

Of course nvidia may just be refusing to provide service in most of these cases, which may be why they are all going to repair shops, and thus they can claim there arent many problem units. The stores taking the returns then have to eat the loss.

 

[MrTeal]

They could have solved the bending issues as well just by using the reversed connector so the retaining latch is on the PCB side, and the notching the cutout further down. You could still have the stupidly large cooler, just mount the connector further down where you have room for a proper bend radius before it hits the case window.

Bonus points if the AIB actually puts little slots in the backplate so you can easily zip tie the power cable to the backplate.

 

[BFG10K]

So, 250 failures from Northridgefix alone. Yet nVidia claims they're only aware of 50 instances worldwide.

Poor, poor, nVidia. Once again they're totally blindsided and in the dark. Totally. Just like they had no idea how many cards miners were buying during the crypto days. No idea, yo.

Also really disappointed to see GamersNexus still pushing the "user error " angle. They lost a lot of credibility after this. If they really believe that, they should have no trouble showing a single repair shop with 250 melted 8-pin PCIe connectors in the last 6 months.

20+ years of PCIe connectors in the wild, thousands of times more than than the number of 12VHPWR in use. Yet no systematic or repeated melting issues have ever been reported.

Do they seriously expect us to believe that 4090 users have suddenly forgotten how to plug cables into a video card?

My fav part of that video is how he's talking about how easy it is to plug in the connector, but they have a video jump-cut around 00:46-:00:49 because he couldn't plug the connector in cleanly, lol. Jeeze, seriously?

I mean that sort of proves the point. If even someone who's an expert at PCB analysis and soldering has trouble connecting it, what chance does an average user have?

Honestly? People really should quit using YouTube as "proof" or "evidence", right or wrong, that as something that supports their facts. The data is complete BS, unless backed up with real data that is verified.

What are you babbling about? Videos of the card repairs being done are right there, complete with full microscope footage. If you think the videos are fake you need to provide evidence.

Feel free to chime in with your soldering "expertise" and point out the problems in the videos.

He may have 250, but of how many vendors and millions sold? Insignificant? Who knows the REAL statistics.

LUL, wut? So if 250 instances of the same car model or appliance melted within 6 months of release, that's OK because "oh, we have millions of others out there!"

And it's 250 that we know of, so far. One repair shop out of dozens - if not hundreds in the world - for a card that's only been out 6 months.

To say nothing of the broken connectors which weren't even reported before. Obviously users are so scared now they're using excessive force and snapping the thing right off. Again, utter garbage design.

Imagine paying $2500 for a device and having to check every day if it's melting. How can people possibly defend this septic excrement and blame the user?

 

[DeathReborn]

Well nvidia claimed this affected less than 50 units worldwide so its at least an interesting data point contradicting the manufacturer.

Of course nvidia may just be refusing to provide service in most of these cases, which may be why they are all going to repair shops, and thus they can claim there arent many problem units. The stores taking the returns then have to eat the loss.

Look at when Nvidia said 50, November 2022. Now we have people posting like they just said it & using it as a reason to jump the shark.

 

[BFG10K]

Look at when Nvidia said 50, November 2022. Now we have people posting like they just said it & using it as a reason to jump the shark.

Fine, so when can we expect their updated numbers?

I'd wager we won't get any, because they expect the problem to blow over by staying quiet. I mean people almost forgot about this for ~6 months until Northridgefix was in the news.

 

[amenx]

Surprised why people resort to fixing the cards instead of RMA'ing them. Minimum warranty is at least a year virtually anywhere one buys the cards. My guess is that these users may have flashed custom bios's that voids their warranty and so resort to having them fixed. Now why would anyone flash a custom bios on a 4090 other than to get more performance and therefore draw more power than the stock bios?

Cablemod says that of 50,000 adapters sold, only 8 customers have reported them melting.

CableMod Is Giving Away A GeForce RTX 4090 To Celebrate 50K Adapter Sales, How To Enter

The GeForce RTX 4090 is NVIDIA's flagship consumer GPU and if you're lucky, you can win one for free from CableMod.

hothardware.com

In first month of RTX 4090 release, 130,000 cards have been sold. Dont know what the figure is now, but I think it could safely be assumed a lot more sold since.

Furthermore, why wasnt this an issue with the 3090s? Maybe because it draws a 100 watt less than a 4090? I think the connector is fine for under 400 watt usage and even 450w usage. But over that, the connector is thus more prone to failure.

 

[igor_kavinski]

Surprised why people resort to fixing the cards instead of RMA'ing them.

Maybe some people (especially those with previous bad RMA experiences) don't want to receive someone else's repaired card with more hours on it or finicky issues. If RMA always meant you get a NIB card, these people wouldn't feel the need to get their cards repaired at their own expense.