Comparison of 16-pin to 4 x 8-pin and 16-pin to 2 x 8-pin PCI-e cables.
Comparison of 16-pin to 4 x 8-pin and 16-pin to 2 x 8-pin PCI-e cables.
It's really pointless anyway because nobody has identified a clear reason. All the warnings CableMod has given are just precautionary statements, like saying something "could" fail. Various groups such as igorslab, jays2cents, gamersnexus, Paul's hardware, buildzoid and others have tested and speculated, but nothing solid. Gamersnexus says the failed cables don't even look alike—they came from different manufacturers, with varying voltages, which only adds to the confusion. Just apply common sense, don’t criticize the cable, stick with the card as planned, and if it fails, report it on forums like Reddit, Twitter, or Twitch. There’s no solid evidence either way about what, why, or when this might happen.
Except that there is, because I haven't heard of any cases where someone with non-Nvidia adapters experienced issues. This clearly points to a build quality problem with those adapters. The specific nature of the issue—whether it's related to crimping instead of soldering, rough edges on connectors, faulty wiring, or something else—doesn't really matter. When hamburger spoils, it doesn't matter if someone touched it and bacteria got in, if it's just old, or if there was cross-contamination; in any case, it wouldn't be used regardless of whether you return it or not. The same applies here. Just avoid using it if it looks bad. Since most power supplies that support these cards come with adapters, it seems obvious what to do.
But I agree that applying common sense is essential, and that sharing your concerns publicly, returning the card if necessary, are crucial steps in resolving this. In truth, all these cards should be recalled until the exact cause of the problem is identified without hesitation, as it could pose a safety risk. If this were a vehicle, they would already be working on it. Unless you're a Tesla, then just ignore it and move on to other markets.
The volume suggests they might have relied on several suppliers for the cables. Or there could be an unreliable supplier who included early rejected batches with the final product. Nvidia is unlikely to disclose such details. The review samples likely contained a higher-end cable than what was actually sold. Examining the two dissections reveals apparent engineering shortcuts on one adapter, possibly converting a 3-cable design into a 4-cable version, while another design seems more suited for four PCIe 8-pin cables. The choice of different insulated wire classes remains unclear. It may be that their supplier simply chose the cheapest bulk option meeting the specs, or they opted for a more heat-resistant cable for one or both cases.
I didn't see that before, but I'm not sure why I missed it—it seems unimportant. It definitely alters the situation. With what I know already, it doesn't make sense for anyone to purchase an RTX 4090 until this issue is fully addressed by everyone. If I owned one of these cards, I'd return it right away for a refund and switch to something else. This isn't a good deal when you think about the price of the cards and the power supplies needed. Nvidia might be making some questionable decisions here. Luckily, they have the support to back it up.
Completely in line with that. Someone messed up the pooch somewhere along the way. All the buzz from Nvidia about engineers and others claiming the socket and plugs are the 'next' best option, only to run into this problem in practice, they won’t admit it’s a problem. This leaves consumers with just two options. Use the card, or not.
It’s straightforward physics. A smaller connection is less effective. If they’d used the standard 8-pin and expanded it to 12 pins, this issue probably wouldn’t exist. But they did the opposite—reducing the size, using smaller pins and terminals, and expecting the setup to handle higher amperage with more wires. The typical PCIe combo is rated for 5A @ 12V per pin. Molex has consistently produced pins and terminals of the same size, rated for 11A or 13A for years. It’s not difficult to use two wires on a 13A pin; you could easily run 6-pin connections on a 4x8 layout, delivering all 12 PCIe hotspots from a 4x8 and still maintaining 600W with 336W extra capacity.
They might have opted for a simpler design instead of trying to reinvent the wheel.
Manufacturers cannot simply include any terms in a warranty and expect them to be automatically legally binding. A clause stating "you agree to pay Corsair 10% of your gross income for every year this warranty is valid" would not be enforceable, as it conflicts with the Magnussun-Moss warranty act. Steve Lehto often addresses such issues on his channel, where he specializes in consumer protection law.
In theory, a three or four should be more secure. A two will manage the task, though. Why? Because if one of the eight pins has a weak connection (from improper insertion), it creates a high-resistance zone. Using just two 8-pin connectors might lead to such a hot spot. With only two, it could trigger an issue. Opting for three or four provides additional wiring to spread out any hot spots from cable faults.
An 8-pin 18-gauge wire can handle up to 360 watts (before melting), but this is ideal only with a perfect connection—rare in practice.
My RTX 4000 series model has a 12-pin connector, which I believe performs worse than two 8-pin connectors. Why? Because the individual pins are extremely small—about half the surface area of an eight-pin connector. If even one or two of these tiny pins have poor contact, it creates high resistance and forces all current onto other cables. This scenario is more likely with the compact 12-pin design than with two or three 8-pin connectors. These must be designed primarily for appearance. The four-pin extension isn’t a solid fail-safe; it only confirms the power supply can handle the amperage. It doesn’t indicate that a poor connection causes melting.
This is compounded by the questionable design choice to angle cables at a 90-degree turn from the start. It seems like an oversight waiting to happen. While some 4090 models have experienced burnouts, others operate near the edge of burning but don’t make it public. The real culprit here is the tiny connectors.
These same miniature connectors are used on both the graphics card and the power supply. My new PCIe power supply came with a 16-pin NVIDIA connector. A larger-pin connector for the power side or soldering it directly would have been safer. Modular power supplies are less reliable in terms of resistance compared to soldered connections. This adds another point of failure. With newer 12+4 GPU connectors featuring tiny pins, it’s more probable that a failure will occur, especially since older 8-pin designs are more forgiving.
All this is meant to make the wiring look neat, but in reality, it often leads to problems when they catch fire.