User claims RTX 4090 16-pin power connector melted on both GPU and PSU side, despite running at 75% power

News
By
published

Simultaneously!

Melted 16-pin power connector
Melted 16-pin power connector (Image credit: Quasar Zone)

At this point, 16-pin power connector meltdowns are like a nightmare we can't wake up from. Just when we thought things had settled down, an RTX 4090 owner on the Korean Quasar Zone forums shared the horrors of how the 16-pin power connector melted on one of the best graphics cards and the power supply — simultaneously.

Typically, you hear cases of the 16-pin power connector (12VHPWR) melting on either the graphics card side or the power supply side. But it's not every day you see meltdown occur on both ends, simultaneously. Although the Quasar Zone user didn't share the hardware models, the photographs showed an MSI GeForce RTX 4090 Suprim X 24G graphics card drawing power from an FSP Hydro power supply of an unknown capacity. What's shocking was that RTX 4090 apparently wasn't even running at full speed: the affected user reportedly limited the power limit to 75%.

The RTX 4090 is a powerful beast; however, not everyone uses their graphics card to its full extent. While some play with different power limits to maximize efficiency, others hobble the graphics card's performance with a stricter power limit to mitigate the risk of the 16-pin power connector melting. However, it isn't an infallible solution, as this latest case has demonstrated: the 16-pin power connectors can melt even if the RTX 4090 is only running at 75% of its power limit.

The RTX 4090 arrived on the retail market in October 2022, and it didn't take long before cases of 16-pin power connector meltdowns started to pile up around the world. There have been many theories as to why RTX 4090 owners were facing the issue — including the 16-pin power connector's design, the quality of the materials, the angle where the cable is bent, etc. While Nvidia promised to dig deeper into the problems, there hasn't been any closure on the meltdowns. Yes, PCI-SIG subsequently revised the power connector to improve its safety, and now we have the 12V-2x6 power connector, but many are still not convinced that such a slight revision can make a big difference.

The meltdowns weren't limited to the power cables/adapters included with the RTX 4090. They persisted with third-party adapters, too. Many were not sold on the included adapter and went out to purchase third-party adapters, which worked for a while. However, Cablemod, one of the more popular brands, had to recall its 16-pin power adapter portfolio due to fire hazards.

The whole 16-pin power connector meltdown is like a disaster waiting to happen. Two years later, we're still hearing stories about 16-pin power connector meltdowns. The number of cases has decreased since the initial spate of meltdowns, but we only know about those who come forward — some affected users may decided to find help elsewhere. Repair technician NorthridgeFix claims he fixes around 200 RTX 4090 graphics cards a month. So, yeah, 16-pin power connector meltdowns seem to be a gift that keeps giving.

Zhiye Liu
Zhiye Liu
News Editor and Memory Reviewer

Zhiye Liu is a news editor and memory reviewer at Tom’s Hardware. Although he loves everything that’s hardware, he has a soft spot for CPUs, GPUs, and RAM.

See more GPUs News
38 Comments Comment from the forums
  • Releximas
    Hope my new Gigabyte Geforce RTX 4080 Super Aero OC 16GB GDDR6X Graphics Card is safe! 🤔🤭
    Reply
  • vinay2070
    Why isn't there a lawsuit for this? Else there will be more melted 5090s.
    Reply
  • chaz_music
    Failures of the connectors on both ends is a common cascading failure mode for power connectors with a high number of parallel circuits. The same thing happens to pralleled wires in industrial power wiring. I've seen it many times over my career as a design engineer. As one pin on connector starts to go high resistance (overheats), it drops the current in that wire and the other parallel conductors have to take up the slack. This puts more stress on the remaining usable conductors and connector pins. The weakest pin in the remaining parallel group then gets the hottest, which can be made worse if it is in the center of the housing (heating from the neighboring pins).

    This issue has been going on for a long time, with many users making relevant comments on what is happening (including me). It appears that they did not select enough pins on the connector for paralleling and did not derate what the pins could handle on top of that.

    From June 16, 2023 Tom's article:
    https://www.tomshardware.com/news/12vhpwr-connector-melting-psu-side
    The saving grace is that most GPU and PSU vendors are using UL approved connectors which should have 94V0 plastic. That type of plastic is hard to truly ignite and have a flaming fire, and if it does flame up, it goes right out once the heat is removed.

    Since I started thinking about this issues more, it is much like the problem that automakers had trying to use legacy 12VDC power rails in cars as DC power usage went up (over 6-8KW). Once they moved to 48VDC rails, the weight of the wiring harnesses got smaller, and the reliability went up due to lower connection failures (reduced current stress). Not to mention the alternator efficiency went up (diode bridge losses as part of the total voltage output).

    These GPUs could have the same solution. As far as that goes, why not bring 48VDC into the motherboards as well? It would be ATX 4.0.
    Reply
  • bit_user
    chaz_music said:
    These GPUs could have the same solution. As far as that goes, why not bring 48VDC into the motherboards as well? It would be ATX 4.0.
    @InvalidError has suggested 24V would be a more cost-effective solution. The USB Consortium also stopped well short of 48 V.
    Reply
  • peachpuff
    vinay2070 said:
    Why isn't there a lawsuit for this? Else there will be more melted 5090s.
    The 5090 will come with a free fire extinguisher...
    Reply
  • Zemerick
    "there hasn't been any closure on the meltdowns."

    Well, there kind of has been. To date, there has been only 1 cause found, and it has affected every single one of the melted cables looked at.

    I know, har-har, not plugged in... but that's just the simple truth. Time and time again, people just failed to plug it in all the way. There's talk about it being difficult in some cases, but that doesn't mean you should just stop trying to plug it in. If it doesn't go in, return it. Then Nvidia / OEMs can deal with why it wasn't plugging in, and you don't have to worry about melting your cables, and possibly damaging the PSU/etc.

    In this case, it's a bit hard to say from the few pictures shown, but the burnt cables on the PSU side strongly suggest it was angled, and not fully inserted.

    @vinay2070 There was... and the person suing had already posted photos online that showed the tell-tale signs of the cable not being properly plugged in, and then silence.

    Reminder: People have pushed these cards beyond 1kw with cables barely even getting warm. The cable isn't the main problem. ( Could it be designed better to ensure it's plugged in? Of course. That clearly needs to be done. But at the end of the day, it's up to the user to make sure it is. It's impossible to idiot-proof something. ) Even the "bend-gate" was disproven as insufficient to cause the issue. It has to actually be unplugged an amount to cause any issue.

    @Releximas Just make sure you have it plugged all the way in. If it's fully seated and latched, you're good to go.

    Oh, slight caveat: There WAS the rather ironic cablemod adapter issue. People went out and bought the cablemod adapters for fear of their cables melting. The original cable was fine, but the cablemod adapter was faulty and had to be recalled.
    Reply
  • Eximo
    Zemerick said:
    Oh, slight caveat: There WAS the rather ironic cablemod adapter issue. People went out and bought the cablemod adapters for fear of their cables melting. The original cable was fine, but the cablemod adapter was faulty and had to be recalled.
    It was a bit of both, they had similar design issues as the original 12VHPWR connector, and people could still plug them in wrong. The right angle adapter was there to solve the connector being too close to many chassis side panels.

    They decided to recall them because they were covering repair costs for people that had problems. Cheaper to remove them from the market than it was to keep sending people's GPUs in for repair.
    Reply
  • Pierce2623
    Zemerick said:
    "there hasn't been any closure on the meltdowns."

    Well, there kind of has been. To date, there has been only 1 cause found, and it has affected every single one of the melted cables looked at.

    I know, har-har, not plugged in... but that's just the simple truth. Time and time again, people just failed to plug it in all the way. There's talk about it being difficult in some cases, but that doesn't mean you should just stop trying to plug it in. If it doesn't go in, return it. Then Nvidia / OEMs can deal with why it wasn't plugging in, and you don't have to worry about melting your cables, and possibly damaging the PSU/etc.

    In this case, it's a bit hard to say from the few pictures shown, but the burnt cables on the PSU side strongly suggest it was angled, and not fully inserted.

    @vinay2070 There was... and the person suing had already posted photos online that showed the tell-tale signs of the cable not being properly plugged in, and then silence.

    Reminder: People have pushed these cards beyond 1kw with cables barely even getting warm. The cable isn't the main problem. ( Could it be designed better to ensure it's plugged in? Of course. That clearly needs to be done. But at the end of the day, it's up to the user to make sure it is. It's impossible to idiot-proof something. ) Even the "bend-gate" was disproven as insufficient to cause the issue. It has to actually be unplugged an amount to cause any issue.

    @Releximas Just make sure you have it plugged all the way in. If it's fully seated and latched, you're good to go.

    Oh, slight caveat: There WAS the rather ironic cablemod adapter issue. People went out and bought the cablemod adapters for fear of their cables melting. The original cable was fine, but the cablemod adapter was faulty and had to be recalled.
    This many people don’t just “fail to plug it in properly” for no reason. Even though I haven’t had the issue since I only have a 4080, I fully recognize the plug is NOT a good design with a nice firm click like it should be.
    Reply
  • newtechldtech
    the solution is to add 48V to the 12V as we used 5V and 12V in older systems.

    48V for GPU power and 12V for the rest...

    Plus changing into USB-C types of connectors.
    Reply
  • Trake_17
    vinay2070 said:
    Why isn't there a lawsuit for this? Else there will be more melted 5090s.
    These people are addicts of a sort, I think. They don't care if costs $2,000 they just want another one to date their fix, and Nvidia is their dealer
    Reply