News RTX 4090's 16-Pin Connector Melted After One Year of Usage

[Admin]

Another RTX 4090 has fallen victim to the 12VHPWR power connector, but this time, it took a full year before the RTX 4090's power connector finally melted.

RTX 4090's 16-Pin Connector Melted After One Year of Usage : Read more

 

[LostFate]

They really ought to be forced to do a recall on the 4090. The ongoing issues and knowing that it's 100% the fault of the connector, it's not fair to the consumers to have to foot the enormous bill for a replacement in a year because of bad engineering.

 

[TechieTwo]

When electrical terminals run hot due to being undersized or having poor physical contact they oxidize which increases the electrical resistance which creates more heat until the connector eventually melts.

 

[-Fran-]

Nah. We all know nVidia and all of its most fervent followers said it's not an issue. People are just connecting the thing wrong. Fake news all the way. Yes. Definitely.

...

Bad sarcasm aside and on a more serious note, I hope nVidia foots the bill for everyone affected in either a direct or indirect way (via AiBs). That's one really good way to use a VERY small portion of their humongous margins of the AI-shekles they've been getting as of late. It would also gain good faith from an already disappointed enthusiast crowd feeling the milking to the bones.

Regards.

 

[Order 66]

Nah. We all know nVidia and all of its most fervent followers said it's not an issue. People are just connecting the thing wrong. Fake news all the way. Yes. Definitely.

...

Bad sarcasm aside and on a more serious note, I hope nVidia foots the bill for everyone affected in either a direct or indirect way (via AiBs). That's one really good way to use a VERY small portion of their humongous margins of the AI-shekles they've been getting as of late. It would also gain good faith from an already disappointed enthusiast crowd feeling the milking to the bones.

Regards.

get your petitions ready and maybe jensen will consider it.

 

[Releximas]

I'm just now in the market to complete my year-long build this Christmas and have enough saved for a 4090.... what is needed is a way to tell the older cards equipped with the 12VHPWR connect from the newer ones with 12V-2x6.... especially before I plunk my money down! 🤣

 

[JarredWaltonGPU]

I hope nVidia foots the bill for everyone affected in either a direct or indirect way (via AiBs). That's one really good way to use a VERY small portion of their humongous margins of the AI-shekles they've been getting as of late. It would also gain good faith from an already disappointed enthusiast crowd feeling the milking to the bones.

Nvidia and it's partners really need to offer to exchange any original RTX 4090 for an updated card with the new connector. All warranties end, eventually, and this incident just proves that it can take a while for problems to crop up. It doesn't matter if your connector melted six months ago, or if it melts five years from now. It's fundamentally a poor design, originally at least, which is the whole point of a recall.

I have plenty of GPUs that are ten years old, sometimes more. Very few of them have failed, and none of the cards I've used for testing over the years have experienced melting connectors. The only thing I ever saw do that prior to the 4090 was GPUs like the R9 290X/290 used in mining farms, which often pulled more power than what you'd see with gaming. These 4090 cards aren't being used for mining, unless someone wants to risk a $1600 card for potentially pennies of "profit" per day. They're single GPUs, melting in gaming PCs.

I'm sure there are class action lawsuits in the works. More critically, it's probably only a matter of time before a 4090 catches fire and burns a house down. Hopefully not, but 450W of power draw is a lot, now matter how you want to slice it.

 

[mac_angel]

So, when these things melt, it looks like it's mostly melting the power connector (wire side). It seems like most of the GPU companies are honouring warranties with these problems, but what about the damage to the cable? Is anyone paying for those?
I think whomever has a 4090 should regularly check the connection (no, I'm not saying they should have to. But whomever does have a 4090, check it regularly to make sure it doesn't melt and then have to deal with all the headaches of warranty, going without a GPU, etc.) just to be safe.
I think this also creates an out for those that 'tinker' with their GPUs. bad overclock or whatever and you mess up your GPU, try to 'melt' the connector, and then blame that as the issue for a warranty.

 

[DookieDraws]

Do we know if any other RTX 40 series GPUs are affected? I purchased an ASUS TUF 4070 Ti back in June and a Corsair Premium 600W PCIe 5.0 / Gen 5 12VHPWR PSU Cable because it just looks better than that stock cable that comes with these GPUs. I don't recall seeing any 4070 Tis being affected, but that's not to say there haven't been any.

Also, I've noticed the ASUS TUF 4070 Ti is hard to find in stock, and can't help but wonder if there has been an issue with them and this being the reason, or maybe they're simply just a popular choice?

I agree, NVIDIA just needs to do a recall before someone loses their life.

Best of luck to all of you with a 4090.

 

[Alvar "Miles" Udell]

Just wait until next year when we see "How to spot original 12VHP RTX 4090 cards on eBay before buying" articles.

 

[SyCoREAPER]

I'm curious if the individual crossflashed another VBIOS and/or upped the power limits. Those 4 split cables also aren't made as well as the 3 split cables from what I've seen.

I'm also curious what the case layout was and if there was stress on the cable. Over time the solder heat cycling could eventually cause bad connections.

None of this excuses the connectors faults but a single picture and no proof that the card is 1 year old screams alarmist news without the facts (again).

Edit: Doing a search I don't see the individual every mentioning owning a 4090 prior.
I'm suspicious of the 1 year claim at this point.

 

[boju]

In wonder if the guy frequently/infrequently removed and inspected the cable thus increasing chance of failure. Paying for one of these would make anyone nervous, poor guy. I just hope people with OG 4090s don't suddenly start checking, just leave it.

 

[SyCoREAPER]

In wonder if the guy frequently/infrequently removed and inspected the cable thus increasing chance of failure. Paying for one of these would make anyone nervous, poor guy. I just hope people with OG 4090s don't suddenly start checking, just leave it.

It personally don't think it's worth ringing the alarm bells with no information and even if it is a genuine 1 year failure, that's 0.000625% 1 year failure rate, maybe even lower. With 160K+ cards sold, 1 failure though it shouldn't** be discounted is insignificant and within manufacturing margins.

**Edit: Typo, had should

 

[instawookie]

No faults here, purchased my Gigabyte OC 4090 on launch day and I've had it Horizontal/Vertical its been in 2 different cases with 2 different motherboards, had it out over the weekend doing an AIO swap, ran stock cables for 2/3 months, been running the cable mod sleeved cable for about 8/9 months now. While I would love to vouch for the consumers that have had failed 4090s, I always feel like the story is missing a piece. I always see the connecter images burnt up after removal, but I can only recall ever seeing maybe 1 photo of the connector actually attached to the GPU in its melting/failing state. Nobody really shows how the connector originally sits before removing it when it fails from what I've seen. I'm on many reddit pages browsing builds and Facebook pc groups and the amount of hack jobs I've seen it pretty astounding, I would suggest looking at the amount of consumers that allow GPU's to sag as a reference...... Common sense should tell you maybe I should support that so it doesn't crack your PCB or damage the PCI-E slot.... Instead consumer applies Sailor Moon toy to top of GPU....

 

[boju]

It personally don't think it's worth ringing the alarm bells with no information and even if it is a genuine 1 year failure, that's 0.000625% 1 year failure rate, maybe even lower. With 160K+ cards sold, 1 failure though it should be discounted is insignificant and within manufacturing margins.

Plenty discuss on forums but there'd be just as many lurkers as well. True, one reported instance shouldn't cause too much concern but there's already been enough hysteria surrounding this connector. Difficult to truly know if there's been anymore instances of this in private circumstances, I don't think that sort of information be easily divulged from Nvidia.

 

[SyCoREAPER]

@boju
Im further suspicious. When I asked on reddit how we knew if it's a 1 year old card, OP provided two screenshots. It shows a purchase from a German retailer. When I wrote back in German the response was "I don't speak German. it was purchased while they were in Germany but don't live there."

That's not indicative of a lie but that further raises my suspicions.

Edit: Ninja posted each other.

 

[JarredWaltonGPU]

Do we know if any other RTX 40 series GPUs are affected? I purchased an ASUS TUF 4070 Ti back in June and a Corsair Premium 600W PCIe 5.0 / Gen 5 12VHPWR PSU Cable because it just looks better than that stock cable that comes with these GPUs. I don't recall seeing any 4070 Tis being affected, but that's not to say there haven't been any.

Also, I've noticed the ASUS TUF 4070 Ti is hard to find in stock, and can't help but wonder if there has been an issue with them and this being the reason, or maybe they're simply just a popular choice?

I agree, NVIDIA just needs to do a recall before someone loses their life.

Best of luck to all of you with a 4090.

I haven't heard any talk of 4070 Ti or 4080 meltdowns, though that doesn't necessarily mean it's never happened. But I suspect a really big contributing factor is having up to 450W (500W on some OC models) going through the 16-pin connector. The step down to the 4080 drops that to 320W–350W, and 4070 Ti further limits it to 285W–300W.

I can't remember when it was reported, but there was apparently a change made by Nvidia to later 40-series cards before the 12V-2x6 was officially announced by PCI-SIG. I think maybe it started with the 4070 initially, but possibly it was also 4070 Ti. If you bought a 4070 Ti at launch, it may have the original 12VHPWR, but probably recent RTX 40-series models (all of them) are using the modified connector with the shortened sense pins.

 

[DavidLejdar]

No faults here, purchased my Gigabyte OC 4090 on launch day and I've had it Horizontal/Vertical its been in 2 different cases with 2 different motherboards, had it out over the weekend doing an AIO swap, ran stock cables for 2/3 months, been running the cable mod sleeved cable for about 8/9 months now. While I would love to vouch for the consumers that have had failed 4090s, I always feel like the story is missing a piece. I always see the connecter images burnt up after removal, but I can only recall ever seeing maybe 1 photo of the connector actually attached to the GPU in its melting/failing state. Nobody really shows how the connector originally sits before removing it when it fails from what I've seen. I'm on many reddit pages browsing builds and Facebook pc groups and the amount of hack jobs I've seen it pretty astounding, I would suggest looking at the amount of consumers that allow GPU's to sag as a reference...... Common sense should tell you maybe I should support that so it doesn't crack your PCB or damage the PCI-E slot.... Instead consumer applies Sailor Moon toy to top of GPU....

Yeah, difficult to tell if user error can always be excluded. To me it seems that a point of failure may be the use of too slim PC cases, where there isn't really room for the cable (which may be squeezed when the side-panel gets closed). Not necessarily the main reason, but technically possible that it is a factor. And lack of pics showing the installation, or lack of specs (as in the linked subreddit), that sure doesn't help to be able to rule out such issue.

 

[durahl]

Would kinda be interesting how the Meltdown Rate for users running their 4090s with a reduced Power Target is 🤔

I only just recently got my RTX 4090 Founders Edition but the second I came across that Video I immediately set mine to 70% and so far ( combined with a 90° Angle Adapter and a PSU with a Native 12VHPWR Cable ) I've had no issues on top of a 30% reduced Power Consumption and decreased Fan Noise Levels 🤨

 

[Mobius79]

Got mine for almost a year, just inspected it, and nothing visible.
I have undervolted mine since almost day one, so it might play a role.

 

[Heat_Fan89]

Apple cut its losses with Nvidia over defective GPU's in their notebooks, so did Microsoft with the OG Xbox, followed by Sony with the Playstation 3. I totally agree and there should be a recall and 4090's should be replaced for updated GPU's.

 

[RodroX]

Its confirmed, it was user error!!!, he bought an RTX 4090 to begin with .

You have to give nvidia some credit here. They designed a $1600 MSRP video card, and <Mod Edit> messed up with the $2 power connector.

Lack of testing, making things harder to plug, not taking in consideratiion usual PC Case measures (just to mention some).... thats all nvidias fault, not the user.

 

[dimar]

When RTX 5090 is released, by then most of the connectors will be completely melted, a good reason for upgrade.

 

[RodroX]

When RTX 5090 is released, by then most of the connectors will be completely melted, a good reason for upgrade.

By the time RTX 5090 Im not sure there will be any PSU affordable enough that can handle it.
Specially if game developers keep writing code the way they have been doing it, and launching games 1 year early than they should.
Just because a game looks "good" doesn't mean its worth to be played, perhaps most developers and publishers should start to focus in the story again, instead of the shinny virtual objects they can show and sell you outside the base game price tag.

 

[edzieba]

Nvidia responded to these issues in November of 2022, when the problem first started, claiming the issues were directly related to user error. It stated that users were not fully pushing the 16-pin power connector into the reciprocating power connector on their RTX 4090s. However, we noted in our coverage that many users reporting RTX 4090 deaths claimed to be experienced system builders, making the likelihood of a failed user insertion potentially not the only cause.

Having had many thousand system-integrator-assembled systems pass through my hands, and found issues far more egregious than "a cable was plugged most of the way in but not all the way in" (e.g. HSF loose and bouncing around inside chassis, because holddown screws were not present), user error is still far and away the most probable cause.
Particularly for 'experienced' system builders/integrators, who are clearly double-plus-good at their jobs and have never had any problems plugging a cable in before, so any post-inspection is clearly a total waste of time and unnecessary because of course it must be some other problem.