Remember when the most expensive consumer GPUs in the world actually melted? Or, most of the time, when their connectors melted? I do too.But it's not a thing of the past.Although the RTX 50-series wasn't supposed to fall victim to the Nvidia melting saga, it still did.
Within weeks of launch, photos of scorched RTX 5090 plugs were all over the internet.This continues to this day.Nvidia's fix solved a problem, but not the only problem Seating was only half the story Close Thinking back, it's hard to believe that it's been something close to four years since this whole "melting connector saga" even started.
The first victims were Nvidia's RTX 4090s, which, at the time, were the top consumer GPUs available.If you trust any hardware to perform well, it's probably that.Something as expensive as the $1,600 RTX 4090 makes you believe that it can handle anything.
And yet, somehow, reports of issues started cropping up quickly, with cards seemingly failing even when they weren't dealing with any particularly heavy workloads.The melting occurred both at the GPU side of the connector and at the PSU side, forcing RMA processes and replacements.Nvidia investigated, of course.
And it pointed the finger not at the hardware, but at the way people plugged these GPUs in.Makes sense: it turned out that when the connector wasn't fully pushed into the card, it became a fire hazard.Bending the connector was also a big no-no.
Nvidia also said it only knew of about 50 cases worldwide, which is a really small margin.Gamers Nexus backed that assessment, putting the failure rate at roughly 0.04%.Meanwhile, Nvidia promised a fast-tracked RMA process.
A bigger fix followed, kind of The follow-up to the saga came with the arrival of the 12V-2x6 connector, which has sense pins that are 1.5mm shorter and power terminals that are 0.25mm longer than the original (problematic, might I add) 12VHPWR design.The design was meant to fix the entire problem, because if the plug isn't all the way in, the sense pins can't even make contact.That'd solve the whole "not plugged in properly" debacle, as the card can't even pull power through a poorly seated connector.
Nvidia seemed pretty confident that the 12V-2x6 would be the answer to all the problems.It even said that it didn't expect the RTX 5090 to have melting issues.The trouble is that those shorter pins only confirm that the plug is fully inserted, but they can't tell whether the current is spread evenly across the six 12V pins, so a card that sees a properly seated plug will keep drawing full power even if one wire is doing most of the work.
That brings us to the ongoing issue.Related These are the cheapest GPUs that can run DLSS 5 (you probably own one already) Despite its performance cost, you can run DLSS 5 on surprisingly affordable graphics cards.Posts By Goran Damnjanovic The real culprit is where the current goes after you plug it in Six wires, one big pool of 12V, and nothing keeping them in check Let's get a little technical, because to understand why the melting continues even though the new connector was supposed to be the best thing since sliced bread.
It all comes down to what happens to the power once it reaches the GPU, and to make it make sense, we'll look as far back as the RTX 30 series.On the RTX 30-series, Nvidia treated the 16-pin connector as three separate 12V sources.Each had its own shunt resistor, too.
Those shunt resistors allowed the card to balance its load, balancing across each pair of wires and ensuring everything was going well.This changed with the RTX 40-series and continued into the RTX 50-series.These days, all six 12V lines are all lumped into a single input, so the GPU is only made aware of total draw.
That can be an issue, because tiny differences in contact resistance between pins are normal in any mass-produced connector, and current will always favor the path of least resistance.This discrepancy shows up in real-world testing; der8auer tested an RTX 5090 FE, and found that two of the six wires were carrying over 22A, while the rest were chilling at a comfortable 2A.While they were chilling, the connector itself was not: on the PSU-side, it was hitting 150 degrees Celsius.
Oof.Extra context is important here: each pin is specced for roughly 9.5A.You can probably see where the problem lies now.
I have to be fair, though, and say that der8auer's results aren't reproduced by every reviewer, but the fact that they showed up at all is still alarming.2026 has made the problem impossible to wave away Reviewers are melting cards now, and DLSS 5 isn't helping At a time when all GPUs are complete rip-offs, you might want them to at least be nice and stable.But the melting problem hasn't exactly gone away just because the market conditions got significantly worse over the past year.
Reviewers continue to report issues with the RTX 5090, which now costs $5,000-$7,500.Club386 lost an RTX 5090 FE and a be quiet! Dark Power 13 PSU to a melted cable.Nvidia replaced the card.
Meanwhile, Hardware Unboxed caught a cable on its ROG Astral RTX 5090 test bench hitting around 175 degrees Celsius, and curiously, the problem was similar to what der8auer reported previously: one wire was pulling around 22A while the others did next to nothing.Then, there's DLSS 5.Most people complain about the AI slop of it all, but they're missing the real problem; the increase in power draw.
TweakTown saw a jump from around 465W to a maximum of 575W when running DLSS 5 in NBA 2K27.Shortly after, a VideoCardz reader reported that their RTX 5090 connector melted during testing DLSS 5.You can cut your risk without buying a new PSU So, as much as we all wish the melting connector was a thing of the past, it's not.
It's still rare, but not impossible, and with the RTX 5090 priced the way it is, no one wants to harm their overpriced GPU.The best fix remains the same through the years; make sure you push the plug in until it's fully seated, and leave at least 35mm of straight cable before it bends.Check it again after your first few gaming sessions to make sure it's steady and unharmed.
I'd skip extensions and fancy bridges and adapters, as they can void warranty, and they don't always help.You can also consider undervolting your GPU.That may or may not protect it, but even if it doesn't fix the problem, it's a legitimate strategy for better performance and longevity.
Related There's a problem with DLSS 5, and it's not even the AI slop The price of AI is out of control Posts 5 By Monica J.White
Read More
