Short answer: the 16-pin power connector on high-end RTX 50 cards is called 12V-2×6. It has 12 power pins and 4 small sense pins, and it is rated for up to 600W. It replaced the 12VHPWR connector from the RTX 40 era, which melted on some RTX 4090 cards.
The redesign fixed one specific problem. It did not end the melting reports. Cases involving the RTX 5090 have continued since 2025. This guide explains how the connector works, what went wrong, and what you can do about it.
What the 16-Pin Connector Actually Is
The “16-pin” name comes from its layout. According to Wikipedia’s summary of the standard, the connector has twelve power pins and four sense pins. Both 12VHPWR and 12V-2×6 are rated up to 600W, and each power pin is rated for 9.5 A. That works out to a theoretical 684W for the whole connector.
A traditional 8-pin PCIe connector is rated far lower, at 4 to 7 A per pin, according to the same source. That is why one 16-pin cable can replace several 8-pin cables on a power-hungry card.
What the sense pins do
The four small pins are not for power. They tell the graphics card how much power the PSU can safely supply. Under the 12V-2×6 rules, a card that sees both sense pins unconnected should draw no power at all, according to Wikipedia.
12VHPWR vs 12V-2×6: What Changed
Many people call 12V-2×6 a new cable. It is not. Wikipedia notes that the cables and plugs stayed the same. The change is in the sockets on the graphics card and the PSU.
In the new design, the sense pins only make contact when the power pins are fully seated. Guru3D’s report describes it as “shorter detection sensing pins” and “longer power pins.” The goal is simple. If the plug sits halfway in, the card should refuse to draw full power.
Other facts worth knowing:
- Marking. 12VHPWR is marked H+. 12V-2×6 is marked H++.
- Standards. 12V-2×6 arrived with PCIe CEM 5.1. The ATX 3.1 PSU standard later said new designs should use only 12V-2×6.
- Compatibility. The plug physically fits both sockets, so older cables can still connect.
Why the Connector Has a Bad Reputation
The RTX 4090 reports
Melting connectors on RTX 4090 cards surfaced in November 2022. According to Wikipedia, NVIDIA attributed the failures to partial insertion and offered expedited replacements. A class-action suit over the issue was voluntarily dismissed in March 2023.
NVIDIA’s RTX 50 promise
In January 2025, NVIDIA staff said at an event in South Korea that they did not expect melting on RTX 50 cards. Tom’s Hardware reported the quote as: “We don’t expect that to happen with the RTX 50 series.” The report relied on a machine-translated source, so the wording is approximate.
What happened next
Reports followed within weeks. The RTX 5090 draws a 575W rated power, compared with 450W for the RTX 4090. Wccftech’s incident tracker lists cases from February 2025 through December 2025, including some involving cables with the yellow-tip adapter. The tracker is a community roundup, so treat individual cases as reports, not confirmed findings.
The Uneven Current Problem
Heat in this connector comes from current, and current is not always shared evenly between the wires.
Simple maths shows why it matters. At 575W and 12 V, the card pulls about 48 A. Spread over six power wires, that is about 8 A each, which sits under the 9.5 A rating.
Tom’s Hardware reported Der8auer’s test in February 2025. The test used an RTX 5090 Founders Edition at about 570W. One wire carried more than 22 A, while the others carried roughly 2 A, 3 A, 5 A, 8 A and 11 A. A hotspot at the PSU end rose above 150°C after four minutes. The report notes a moderate margin of error, since the wires could not be measured at the same moment.
Why the card cannot fix it
Earlier cards handled this differently. Wikipedia’s summary says the RTX 3000 series split the power wires into three monitored groups. On RTX 40 cards, the wires are tied together internally, so the GPU cannot tell if one pin has poor contact. The same page says RTX 50 cards share that limitation, and that some vendors, such as ASUS, add shunt resistors to detect imbalances. Those do not balance current on their own.
What NVIDIA and Partners Have Done
According to the Wccftech tracker, NVIDIA has made no connector revision since launch. Its guidance is to seat the connector fully and evenly. Partners have added their own safeguards:
- MSI sells yellow-tip connectors as a seating indicator. Its MPG AI PSUs include GPU Safeguard+, which cuts the power limit from 575W to 430W when it detects a problem.
- ASRock released an L-type cable with a temperature sensor for newer Taichi and Phantom PSUs.
- ZOTAC recommends one cable per 8-pin connector when you use its adapter.
- Third-party tools such as Thermal Grizzly’s WireView Pro 2 monitor the connector, and Aqua Computer’s Ampinel does load balancing.
What You Can Do
These steps come from the Wccftech tracker and the standards summaries above:
- Use an ATX 3.1 PSU with a native 12V-2×6 cable. It removes the adapter question.
- Seat the plug fully. Look for no visible gap, and recheck it from time to time.
- Avoid 8-pin-to-16-pin adapters where you can.
- Keep bends away from the plug. Stay at least 35 mm clear of the connector before you bend the cable.
- Consider monitoring. A tool like WireView Pro 2 shows what the connector is doing in real time.
For wattage and PSU sizing, read our guide on RTX 50 series power requirements. To see how much power the cards actually pull in games, see how much power an RTX 50 GPU uses while gaming.
Do All RTX 50 Cards Use It?
No. NVIDIA’s RTX 5060 family page lists one PCIe 8-pin cable, or a 300 W Gen 5 cable with an adapter in the box, for the RTX 5060 and RTX 5060 Ti. Those cards draw 145 W and 180 W, far below the RTX 5090. NVIDIA notes the specs apply to its reference design, and board partners may differ, so check the exact model you plan to buy.
So Is It Safe?
The connector works as designed when it is fully seated and the current is shared evenly. The risk is concentrated on the highest-power cards, especially the RTX 5090, where there is little room for error.
None of the sources reviewed here mention a recall of RTX 50 cards. Still, the pattern in the tracker is consistent. Loose seating, adapters, sharp bends and uneven current all raise the odds of trouble.
FAQs
What is the 16-pin connector on RTX 50 cards?
It is the 12V-2×6 connector. It has 12 power pins and 4 sense pins, and it is rated for up to 600W.
Is 12V-2×6 the same as 12VHPWR?
Almost. The plugs and cables are the same. The sockets changed, so the sense pins only connect when the power pins are fully seated.
Does the RTX 5090 melt like the RTX 4090?
Melting reports have continued on the RTX 5090. NVIDIA said in early 2025 that it did not expect them. Community trackers list cases since February 2025.
Do I need a new PSU for an RTX 50 card?
Not always. An ATX 3.1 unit with a native 12V-2×6 cable is the cleanest option for a high-end card. Check your card’s requirements first.
Can I use an adapter?
Yes, adapters ship with many cards. The tracker advises avoiding them where a native cable is possible, since they add contact points and resistance.
Do the RTX 5060 and 5060 Ti use this connector?
NVIDIA lists an 8-pin cable, or a Gen 5 cable with an adapter, for those two cards. Partner cards may differ.