PCI-SIG says it isn't to blame for melting 12VHPWR cables

Daniel Sims

Posts: 757   +29
Staff
A hot potato: Various entities are passing around the blame for the melting power cable connectors on Nvidia's new GeForce RTX 4000 series graphics cards. The consortium managing PCI specifications is the latest to deny any problems with its end of the production process, reminding manufacturers to take proper precautions.

The Peripheral Component Interconnect Special Interest Group (PCI-SIG) delivered an official response to the press this week concerning the melting power cables for Nvidia's latest GPUs. The group's boilerplate safety reminder suggests it thinks the problem lies not with it or the end users but with manufacturers.

The PCI-SIG publishes and manages specifications for many aspects of PCs, including the 12HPWR power cables at the center of the recent controversy. It said its specifications include all necessary technical information on interoperability and that the rest is up to manufacturers basing products on those specs – just Nvidia in this case. The PCI-SIG suggested Nvidia is responsible for properly designing, manufacturing, and testing products like GPU power cable connectors.

Since late October, users have complained about melting power connections on Nvidia's RTX 4090 graphics card. They quickly blamed the adapters the company included to make the 16-pin 12VHPWR cables compatible with older ATX 2.0 power supplies. The cables only natively support the newer but less common ATX 3.0 PSUs.

One customer filed a class-action lawsuit against Nvidia, calling its adapters dangerous and defective. However, separate investigations from Nvidia and Gamers Nexus indicate that only a tiny percentage of users encountered the issue and that it stems from improperly connected cables. The PCI-SIG said it's aware of the lawsuit but only directed its safety advisory toward manufacturers, not customers.

The adapters appear safe for most users who follow directions, although Igor's Lab thinks they might be a bit delicate. If you're unsure about the included power connector adaptors for your RTX 4090 or 4080, you could shop for third-party adapters or switch to an ATX 3.0 PSU to remove the need for adaptors. Upgrading the power supply would only add a couple of hundred dollars to the price of either GPU, both of which exceed $1,000.

The whole mess has given AMD more ammunition in its competition against Nvidia. AMD's upcoming Radeon RX 7900 XTX and 7900 XT GPUs are cheaper than their RTX 4000 equivalents and don't require adapters for ATX 2.0 PSUs. The company didn't hesitate to highlight both advantages.

Permalink to story.

 

Uncle Al

Posts: 9,363   +8,581
Separate investigations from Nvidia and Gamers Nexus indicate that only a tiny percentage of users encountered the issue and that it stems from improperly connected cables.

Well, that may be true but I'd like to see an independent investigation by the Govt. or one of it's agencies/contractors. Both Nvidia & Gamers Nexus can hardly be considered independent. If the problem is the connector in that it did not completely snap together, I'd call that an Nvidia issue and they should provide a full refund and pay for any damages as well.

Perhaps it's a bit of an overreach but I'd like to also see a law prohibiting companies from investigating themselves and force them to refer to matter to outside sources that have Govt. approval; just to elliminate any issues of self-promotion ......
 

Ravey

Posts: 383   +177
Safety starts with the design. If it was a manufacturing problem they should have picked it up in the prototyping phase and engineered around it. I see a class action lawsuit coming.

I was about to say, I'm sure the UK's HSE would disagree with the statement by PCI-SIG under the CDM 2015 Regs. Culpability is just as much the responsibility of the Client and Designer as well as the Manufacturer. I would ask PCI-SIG if they conduct physical audits of Nvidias manufacturing process to ensure their guidence has been adequately followed. If not, why not?

You can't just go to court and say "Well your Honor, the consumers' were using a sub-standard adaptor to power their card.. blame them"... That's not going to fly.
 

Neatfeatguy

Posts: 1,094   +2,019
I was about to say, I'm sure the UK's HSE would disagree with the statement by PCI-SIG under the CDM 2015 Regs. Culpability is just as much the responsibility of the Client and Designer as well as the Manufacturer. I would ask PCI-SIG if they conduct physical audits of Nvidias manufacturing process to ensure their guidence has been adequately followed. If not, why not?

You can't just go to court and say "Well your Honor, the consumers' were using a sub-standard adaptor to power their card.. blame them"... That's not going to fly.

It's the norm for companies, shift the blame until their hands are forced. Once their hands are forced they're generally just slapped with a pathetic fine. This only reinforces companies by teaching them they can do pretty much whatever they want because the consequences are trivial. They'll make more money doing something wrong than they will ever be fined for.
 

neeyik

Posts: 2,424   +2,979
Staff member
Safety starts with the design. If it was a manufacturing problem they should have picked it up in the prototyping phase and engineered around it. I see a class action lawsuit coming.
That's precisely what PCI-SIG said - the consortium lays out specifications for interfaces, power delivery systems, signaling tolerances, etc. It's then down to PCI-SIG members to design, test, manufacture, and distribute products to meet those specifications, not the consortium.

While there can certainly be flaws in any PCI-SIG specifications (which are submitted by consortium members for review, comment, and final approval), there are no explicit instructions on exactly how to make and test products based on a given spec.

Take, for example, PCI Express x16 board connectors -- these are mass manufactured around the world, by numerous companies. They can be found using a range of materials, that result in different physical attributes, so some will barely meet the spec's mating cycle limit, whereas others greatly exceed it.
 

bexwhitt

Posts: 622   +362
The latching mechanism should be better and more obvious that you have pushed the connector in properly, making the sense pins a bit shorter, so it knows that the plug is not fully in would help also. FYI not pushing the connector fully home seems to be what is causing burnt connectors.
 
Last edited:

hwertz

Posts: 211   +123
I'll just comment, it probably doesn't matter if it's only a tiny fraction having melting electrical connectors or not. Throughout history (well, back to the 1970s when the consumer product safety commission formed), I think you'll find the products that got recalled all had only a tiny fraction having problems. It should be as @bexwhitt says, the latch should make it obvious when it's fully seated, and the sense pin should keep the connector (and so the card) inoperative if the connector is not fully seated. (Really, I question running that much current through a connector that size, too.)