The notorious 16-pin connector melting didn't originate with the RTX 50 series, and it dates back to as early as 2022, when we used to see regular reports of connector melting on the RTX 4090. Despite NVIDIA and its AIBs' attempts to mitigate the problem, the saga continues. Almost every week, we hear new reports of connector melting, mostly on the flagship Blackwell GeForce RTX 5090, rated at 575W.
We have been reporting the stories regularly since the launch of the RTX 5090, and even a year after its official launch, the reports continue. Today, we will be deep diving into the issue, examining the common patterns behind these connector melting reports, what NVIDIA and GPU manufacturers have done to mitigate the risks, and what you can do to ensure a melting-free operation.
RTX 5090 Becomes the New Victim of Connector Melting: The Timeline
NVIDIA GeForce RTX 5090 was officially unveiled at CES 2025, launching at the end of January in retail stores. In just 10 days, the first RTX 5090 connector melting report surfaced, revealing how severely the cable connector had melted on the RTX 5090 Founders Edition. While it was no surprise to see the connector melting issue getting extended to RTX 5090, as it was also common on the GeForce RTX 4090, the RTX 5090 appears to be affected more frequently.
With the TGP limit increasing from 450W on the RTX 4090 to 575W on the new flagship card, it was obvious that this increased power limit would also heat the connector quickly. Despite transitioning to the latest ATX 3.1 standard, the new 12V-2x6 connector isn't able to protect the card from damage. Eventually, more reports started to surface on the internet, mostly on Reddit, including reports of melting connectors on both the GPU and PSU.
MSI's yellow-tipped connector also failed to mitigate the issue. We have recently seen how the MSI yellow tip connector was reportedly pushing itself out of the connector after some weeks, which is probably why it fails easily. A lot of RTX 5090s failed in the last year, and while it's difficult to know the exact number of such cases, hardly any week passed that we didn't see someone report it. It should be noted that MSI has two forms of yellow-tipped connectors, one shipped with the GPUs which are entirely adapter-based, and then we have those that ship with MSI PSUs, which are single 16-pin to 16-pin connectors. The latter have shown remarkably lower fail rates or none at all, while the adapters which are officially supplied to MSI by NVIDIA have been prone to these issues.
Sometimes, the connector melted in an overclocking attempt, but most of the time, users saw connector melting while gaming. This happened reportedly despite proper connection, but there were also many reports where the connector remained slightly loose on the GPU side. That said, in some instances, the connector melting was so intense that one could hardly recognize it.
Then there were cases where the connector refused to come off as it melted and got stuck to the GPU. Of course, there were instances where the RTX 5090 was seen catching fire and damaging other parts of the build, but this post is limited to the 16-pin connector melting.
Why Does the 12V-2x6 Connector Melt?
The problem stems from the flawed connector design. The very next day after the first RTX 5090 connector melting incident, popular German overclocker Der8auer demonstrates how the 16-pin connector (12VHPWR) can reach up to 150°C on the PSU side and nearly 90°C on the GPU side as well. While one could argue that the latest ATX 3.1 standard 12V-2x6 connector offers more safety measures and better reliability, the core design was retained.
12V-2x6 connector introduced shorter sense pins (roughly 1.5mm shorter than 12VHPWR sense pins) and longer power/ground pins (+0.25 longer) to ensure a proper contact. This change was made as the first 16-pin 12VHPWR connector was supposedly causing melting problems due to improper contact with the pins. Whenever the connector has some of its pins disconnected, the entire load gets distributed on the rest of the pins.
Unlike traditional 6-pin and 8-pin PCIe connectors, the 16-pin design can carry the entire load through a single contact, even if other pins in the group lose connection. Now imagine sending nearly 600W through a single pin or a few pins, which is already a big jump over what the conventional 8-pin connector can handle. The temperature rise is inevitable. This is why the PCIe 5.1/ATX 3.1 standard was introduced, and while each power pin was designed to handle around 9-9.5 Amps of current (around 108 W per pin), a few pins or sometimes a single pin have to carry the entire load, which results in such a disaster.
As we mentioned earlier, Der8auer's testing included the 12VHPWR connector on the cable-side, but Andreas Schilling's investigation, which had a 12V-2x6 cable revealed a similar finding. The connector reportedly reached 150°C quickly under full load, which caused the melting of the housing of one pin, while others remained intact. This is why some users still witnessed the connector melting despite reducing the power limit of their GPUs.
NVIDIA and AIBs Response: Mitigation Steps Taken by Vendors
It's hard to believe, but NVIDIA didn't do much to mitigate the issues after launching its RTX 50 series cards. The connector remains on the cards as it was a year ago, and no new revision has been made since. As far as what NVIDIA's guidelines are concerned, the company just keeps repeating the same thing. It recommends users to make sure their connectors are fully and evenly seated. Unfortunately for its board partners, they can't do anything but comply. AIBs don't have any choice but to use the 12V-2x6 connector, which is why some of them tried to come up with their own solutions, such as MSI's yellow tip connector, which has been there for a while now.
Many vendors ship their own 8-pin to 16-pin adapters and recommend using them, but we have seen that these adapters are prone to failures more often than a single cable. Many users who use a 4x 8-pin to 16-pin adapter may not be aware of the correct guidelines for using it.
For instance, ZOTAC recommends using one cable for every 8-pin connector for the adapter. In simple words, if your PSU has 4x 8-pin ports, one shouldn't use just two cables with two connectors each, but should instead use "four" cables, each having a single 8-pin connector connected to the adapter.
MSI has also tried mitigating the issue by adding a 16-pin power connector warning on its MPG AI PSUs recently. The new power supply units are equipped with the GPU Safeguard+ feature that will automatically lower the power limit from 575W to 430W when it detects any abnormality. This feature is supported in the latest version of MSI Afterburner (v4.6.7 and above), but only works with the MPG AI PSUs as of now.
ASRock also released a 12V-2x6 L-Type cable, which helps in eliminating any bends in the cable near the connector. It brings a built-in NTC sensor that detects resistance changes and delivers Over-Temperature Protection to keep your connector safe. However, this only works with newer ASRock Taichi and Phantom gaming power supplies.
Then there is Thermal Grizzly, which has the latest WireView Pro 2 with built-in sensors and protection features, and is probably the best device for protecting against connector melting issues. There is another device called Ampinel, launched recently by Aqua Computer, that can perform load balancing and is cheaper than Thermal Grizzly's device.
Guidelines to Prevent Connector Melting
The first and foremost step to mitigate the connector burning risks is to adhere to the manufacturer's guidelines. NVIDIA's board partners have their safety guides for using the 12V-2x6 connector properly on the RTX 5090 cards. Along with those, here are the measures you should follow to ensure a reliable operation:
Choose the Right PSU
It's essential to have a PCIe 5.1/ATX 3.1-compliant power supply unit for RTX 5090 and other GPUs with the 12V-2x6 connector. These PSUs offer better reliability and ship with native 12V-2x6 cables. These are designed to handle strict transient spike requirements defined under ATX 3.x specs and eliminate the need for any adapters. Moreover, choosing the right wattage unit is also crucial. For the RTX 5090, don't settle for anything less than 1000W. However, 1200W or higher is recommended for better headroom.
Apart from these specs, ensure that you are buying from a reputable brand that offers proper 12V rail stability and also meets the modern protection standards. Avoid mixing cables from different PSU brands as much as possible.
Ensure the Connector is Fully Seated
This is a no-brainer, as a lot of connector melting occurred because the connector wasn't fully seated. The connection should have no visible gap between the plug and the housing. Remember that a partially seated connector can cause uneven current distribution across the power and ground pins, increasing the electrical resistance on individual pins. This is a major reason for connector melting. It's recommended to check the connection every couple of weeks because some users have complained that the connector loosens after some time.
Avoid 8-pin to 16-pin Adapters
Even though many AIBs ship an 8-pin to 16-pin power adapter with their GPUs, one should avoid them as much as possible. It's recommended to buy the latest ATX 3.1-certified PSU to have that native 12V-2x6 cable so that a single cable goes from your PSU to your GPU. With an 8-pin to 16-pin adapter, the additional connection points may increase resistance and introduce more failure risks.
Don't Bend the Cable Near the Connector
You shouldn't have any bend within 35 mm on the cable after the connection. Excessive bending or sharp cable angles near the 12V-2x6 connector can compromise the contact integrity, which may loosen the connection and lead to poor contact. This again can cause connector melting and possibly cable damage at the bend.
Buy Utilities like WireView Pro 2
It's better to spend an additional $140 on a device like WireView Pro 2 than risk a $2500+ GPU to a potential meltdown. WireView Pro 2 isn't the only thing you can buy to keep an eye on your connector temperature. You can also buy ASRock or MSI's latest PSUs with the special protective features. At the end of the day, you should try to mitigate the issue as much as possible, because the inherent flaw in the 12V-2x6 connector isn't going anywhere any time soon.
FAQ
Why is the RTX 5090 more prone to connector melting than the RTX 4090?
NVIDIA RTX 5090 is more prone to connector melting since it is rated at a much higher TGP of 575W than any other GPU in the series. Compared to the 450W TGP on the RTX 4090, the RTX 5090 can draw more power, which increases the electrical and thermal stress on the 12V-2x6 connector.
Is the 12V-2x6 connector different from the original 12VHPWR?
The 12V-2x6 connector improves reliability, bringing shorter sense pins and longer power/ground pins for better connection. Apart from these, there isn't any noticeable change, and the core design remains largely the same. This is why the new connector suffers from melting similar to the older design.
Can lowering the GPU power limit prevent connector melting?
Not necessarily. Lowering the power limit may reduce the overall power draw and heat generation, but it will still be high enough to damage the connector if the connection is imperfect and the load is distributed to a few pins. We recently saw how a user reduced the power limit of his RTX 5090 to 500W, but still got his connector cooked.
What warning signs should I look for before an RTX 5090 16-pin connector fails?
Usually, one of the early signs of a 16-pin connector failing includes crashes, system instability, and freezes. Some people may encounter unexpected black screens, but sometimes the graphics card may not show such signs. It's best to check the connection every few months and ensure the connector is properly inserted in cases.
What type and wattage of supply should I use with an RTX 5090 to minimize connector problems?
For the RTX 5090, it's best to have at least a 1000W power supply, but it's better to have a 1250W power supply or more. Moreover, the PSU unit should be ATX 3.1 and PCIe 5.1-compliant to ensure it can handle the temporary GPU power spikes.
