Nvidia delivers official statement on melting RTX 4090 power cables

[Daniel Sims]

A hot potato: Power supply-related controversy has surrounded Nvidia's monstrous new flagship graphics card since before it launched. Users recently criticized the company after reports of melting power cables, but Nvidia claims only a handful of incidents occurred and that it was user error to blame.

Nvidia issued an official statement today regarding instances of melting power cords for the GeForce RTX 4090. The company is aware of about 50 cases worldwide, claiming they result from users improperly plugging in the cables.

The trouble started in late October when Reddit users reported power connectors for the 4090 melting on the end of the cable and the GPU's connection port. Suspicion quickly shifted toward the power adaptors Nvidia shipped with the cards to make them compatible with older power supplies.

The RTX 4090's 16-pin 12VHPWR cables are meant for ATX 3.0 PSUs, but Nvidia knew many customers wouldn't want to spend more money upgrading from their ATX 2.0 models. Multiple outlets have criticized the company's adapter solution.

In September, the Peripheral Component Interconnect Special Interest Group (PCI-SIG) raised a potential over-current issue with the adaptors that Nvidia claims it fixed. An investigation by Igor's Lab found that the adaptors are so fragile that plugging or unplugging them can damage the connections. Cable supplier CableMod warned that bending the cables at a short distance is unsafe.

However, a Gamers Nexus investigation arrived closer to Nvidia's conclusion. It said problems only occur when the connector is pulled at an angle or partially disconnected. The issue could also involve foreign object debris. Like Nvidia, Gamers Nexus estimates the 4090 has an extremely low failure rate.

The controversy escalated further when user [name redacted] filed a class-action lawsuit against Nvidia, accusing the company of committing fraud by distributing dangerous, defective adaptors. Despite findings blaming user error, he claims his experience with GPU installation ensured best practices.

It seems that Nvidia's adaptors are safe for most people if appropriately connected. Users who aren't sure about the pack-in adaptors can find third-party options or upgrade to an ATX 3.0 PSU, which doesn't need an adapter for the RTX 4090. After spending over $1,600 for the graphics card, dropping less than $200 on a PSU is a small extra step.

While AMD tacitly admitted that none of its new RDNA 3 GPUs can directly compete with the 4090, it can claim that they don't need adapters for use with ATX 2.0 power supplies. The Radeon RX 7900 XTX and XT launch on December 13 for $999 and $899, respectively.

Permalink to story.

https://www.techspot.com/news/96709-nvidia-delivers-official-statement-melting-4090-power-cables.html

 

[torchyear]

The new gig economy is getting out of hand. Nvidia didn't even have to ask or pay GN, and they got results!

 

[nnguy2]

"Nvidia blames their dumb consumers" FIFY

 

[brucek]

My problem with the user error hypothesis is that it's hard to understand why 4090 purchasers are less capable than those of any other card. They can say what they want but the bottom line is there is a uniquely dangerous condition that is occurring with this one particular video card and not with any others.

They may technically be correct that that additional user caution could have mitigated these issues. The point though is that it is at least partially on them for a) having created the situation where more caution than ever before is now required; and b) not having done anything proactive to alert their customers to this new situation nor apparently to mitigate it, such as safeguards that provide better visual confirmation when the required degree of solid connection has not been achieved or ideally just refusing to power on until it has.

User error is not in short supply in the real world. Product design and consumer safety generally demand the product not fail catastrophically because of it when there is any reasonable way that could have been prevented.

 

[Uncle Al]

And the slide continues. Only the most lame of companies blames their customers when parts of their product are flawed and can easily lead the customer down the path of destruction. And to set their supplier on the clients simply shows that Nvidia is only a shadow of it's old self and probably cannot be trusted in the future.

 

[Faelan]

My problem with the user error hypothesis is that it's hard to understand why 4090 purchasers are less capable than those of any other card. They can say what they want but the bottom line is there is a uniquely dangerous condition that is occurring with this one particular video card and not with any others.

I have lost count of how many PCs I’ve built over the years. I don’t consider myself a pro, but I do feel that I at least know what I’m doing. I got my 4090, plopped it in and gave that connector such a hard push that I was worried about the PCIe slot. It never clicked but it appeared to be fully inserted. Because people already at the time were talking about being careful with bending and handling the adapter in general, I didn’t do the pull and wiggle I sometimes do if I’m not sure if it latched. I carried on plugging in the other cables, arranging them, moved the case around, then checked again and surprise, surprise, the connecter was no longer fully inserted. Half a millimeter was showing. Had to shove it back in again and get out my phone to use as a flashlight to see if it latched because it felt like something may have clicked, but I didn’t hear anything and wasn’t quite sure. Someone just rushing to get his system back up and running would have missed it. That to me says that they need to revise the design.

 

[Lew Zealand]

"You're plugging it wrong."

- maybe Steve Jobs

 

[Neatfeatguy]

My problem with the user error hypothesis is that it's hard to understand why 4090 purchasers are less capable than those of any other card. They can say what they want but the bottom line is there is a uniquely dangerous condition that is occurring with this one particular video card and not with any others.

They may technically be correct that that additional user caution could have mitigated these issues. The point though is that it is at least partially on them for a) having created the situation where more caution than ever before is now required; and b) not having done anything proactive to alert their customers to this new situation nor apparently to mitigate it, such as safeguards that provide better visual confirmation when the required degree of solid connection has not been achieved or ideally just refusing to power on until it has.

User error is not in short supply in the real world. Product design and consumer safety generally demand the product not fail catastrophically because of it when there is any reasonable way that could have been prevented.

I don't disagree with you, they should have a better design to ensure the cable seats correctly. After watching GN videos it become apparent that even with a lot of force, sometimes the cable doesn't seat correctly, but it gives the person the feeling that it's in all the way.

I'd just like to think that if there is anyone dumb willing to spend $1600+ on a 4090 they should be smart capable enough to notice they didn't insert the adapter all the way.

In the end, we all know Nvidia doesn't care. They'll be sure to let the world know it's user error and that there is no problem, just to save face all while they're actually calling their consumers dumb/stupid.

 

[TheBigFatClown]

Well, I think NVidia is so disrespectful to come out and make accusations against their own companies customers when two other sources have discovered that these are not well made connectors.

I won't have to worry about this issue. If I do upgrade anytime soon it will be to a Radeon 6000 series. And that will more than quadruple my iGPU power which has served me well for the past 3 years. No need to go full balls unless your just rich. Then it's okay I guess. Can't take it with you.

 

[DonquixoteIII]

All design errors are the users' fault. However, that didn't end well for Ford or Firestone, and I predict a similar outcome for nVidia.

 

[punctualturn]

So blame their customers....

Why I am not surprised?

Better yet, the influencers are running wild with that BS excuse to defend their beloved and defenseless Nvidia.....

 

[punctualturn]

"You're plugging it wrong."

- maybe Steve Jobs

- I am your god Jensen.

Fixed a bit for you.

 

[godrilla]

FYI their response came just as soon as a lawsuit filing occurred 34 days post launch. Coincidence? Nvidia was more focused on 4080 launch than resolving anything.
Also lines up with Gamer's Nexus thorough objective analysis on the issue. The rate is a 0.4% just in case you missed it at 50 cases so far over 125k shipped.

 

[takaozo]

Too many design errors recently:
Nvidia - will it burn aka new 12v plug
Intel - will it bend aka socket 1700 CPU
AMD - will it stay cool aka Zen4 IHS

None of them admits it's their fault.

 

[Nobina]

-It's only 50 cases worldwide.
-How many have you sold?
-52

 

[hahahanoobs]

I haven't heard of this being tested or mentioned before, but does plugging into the 16-pin connector feel any different than
standard 8-pin?

Even ignoring that, if I had to make an educated guess, the amount of power going through that one connector (and also being an adapter and not native from PSU) makes it even more important to be sure more than ever that it's seated properly.

Reports were of damage with no one taking pictures of the adapter still attached that I saw, meaning we have no idea what these people did. People are/can be pretty.... you know.

It still could have been executed and handled better. The two different adapters and the images of weak solder certainly didn't help.

I personally would only use an adapter for anything as a LAST resort. Always have.

 

[hahahanoobs]

Too many design errors recently:
Nvidia - will it burn aka new 12v plug
Intel - will it bend aka socket 1700 CPU
AMD - will it stay cool aka Zen4 IHS

None of them admits it's their fault.

Except there were no reports of damage from the Intel socket and Zen 4 95C temps.

 

[yRaz]

My problem with the user error hypothesis is that it's hard to understand why 4090 purchasers are less capable than those of any other card. They can say what they want but the bottom line is there is a uniquely dangerous condition that is occurring with this one particular video card and not with any others.

My problem with user error is that a large part of making a connector is to eliminate user error.

the 8pin and 6 pin connectors didn't have this issue, molex connectors didn't have this issue. nVidia released a new power that is less safe than what it was trying to replace.

The other thing, it isn't just the cable being fully seated that's the issue. The issue comes when it is pulled at an angle. The smaller pins aren't as strong as the old connector and cannot handle the physical load being put on it.

 

[DSirius]

Lets be clear. Nvidia customers are clever enough to spend 2000 $ or Euro for 4090 cards and right after, Nvidia claims that the same clients are too dumb to plugin the failed designed 16 pin power connector.
Another pinnacle of Nvidia mental gymnastic.
As I said in a previous post, Nvidia chose to embrace Steve Jobs mantra to Iphone4 antenna gate: "You keep it wrong".
Well, now we have a new motto from Nvidia to 4090 customers - You bought it wrong!

 

[takaozo]

My problem with user error is that a large part of making a connector is to eliminate user error.

the 8pin and 6 pin connectors didn't have this issue, molex connectors didn't have this issue. nVidia released a new power that is less safe than what it was trying to replace.

The other thing, it isn't just the cable being fully seated that's the issue. The issue comes when it is pulled at an angle. The smaller pins aren't as strong as the old connector and cannot handle the physical load being put on it.

They don't make them like they use to.

AMP connector was 8.5 Amp nominal and 11 Amp max per pin designed in 1963. 20 years later Molex had about the same specs.

Wikipedia
8 pin and 6 pin PCIE power are rated ~ 4.2 Amp per pin.

Wikipedia link 2

this 16 pin 12vhpwr could be designed with dual latch since one it's not good and will disconnect on cable routing. Example:

It's not like reinventing the wheel, they just have to look at auto or military designs and copy/paste.

 

[yRaz]

They don't make them like they use to.

AMP connector was 8.5 Amp nominal and 11 Amp max per pin designed in 1963. 20 years later Molex had about the same specs.

Wikipedia
8 pin and 6 pin PCIE power are rated ~ 4.2 Amp per pin.

Wikipedia link 2

this 16 pin 12vhpwr could be designed with dual latch since one it's not good and will disconnect on cable routing. Example:

It's not like reinventing the wheel, they just have to look at auto or military designs and copy/paste.

I know the "amp per pin" design on the 8pin connectors is very low. For some of my electronics projects I use modular power supply cables because I have a literal box full of ones I'm not using. I've pushed 20amps through a single pin, no problem. No recommending people do that, but modern power supply connectors are horrendously under rated. Maybe nVidia is trying to change that by making them worse?

But this problem has been solved decades ago and no one was asking for a new PCI-e connector. They made a product no one asked for and they screwed it up. Good job

*claps slowly*

 

[QuaZulu]

Meanwhile the terribly overpriced RTX 40 series is sold out and being scalped to even further asinine heights. And fools are buying even those, giving NVIDIA and scalpers oxygen.

Someone (Intel? AMD? Anyone?) please, NVIDIA needs a hard kick right in the "RTX"!

 

[Greggs]

I'd just like to think that if there is anyone dumb willing to spend $1600+ on a 4090 they should be smart capable enough to notice they didn't insert the adapter all the way.

That would correlate lots of resources (money) with being smart or capable. The person spending this kind of cash for a gaming card likely doesn't see the price as a barrier. The person who saved up to buy a $300 card likely will treat it extremely well.

 

[Geralt]

Too many design errors recently:
Nvidia - will it burn aka new 12v plug
Intel - will it bend aka socket 1700 CPU
AMD - will it stay cool aka Zen4 IHS

None of them admits it's their fault.

Typical human nature.

 

[Watzupken]

Yes, people are plugging it wrong. But perhaps they need to consider how they can improve the connector such that people get the assurance that it's really plugged in properly. Ultimately, I won't discount the fact that it is a poorly designed connector and early adopters are paying the price for poorly thought design.