News User Reports 12VHPWR Connector Melting From the PSU Side

[handsome-jd]

Hey I didn't know how to contact the author.

I just took delivery of a Dark Power 13 850w and the PCIe 5 socket is marred / looks similar to that failure, obviously no melted connectors yet as I haven't used it.

The rest of the connections are completely spotless, its only this connector that looks weird.

Gallery with a few more shots

https://imgur.com/a/JG1pJEB

View: https://imgur.com/a/JG1pJEB


Send me a message if you want more information, I've contacted my retailer to see what they think,

 

[neojack]

@handsome-jd
contact gamer's nexus, they would be happy to buy it from you and make a video. i am sure of it

i can foresee the future dramatic thumbnail with flames and all lol

 

[steve4king]

https://imgur.com/AVyRZHr

https://imgur.com/pVe57Pj

It's definitely the user's fault, they should have poured epoxy into the loose connectors before plugging them in.

 

[Vanderlindemedia]

The PCI-E 8 pin is far more robust then this 16 pin wire.

 

[Dingledooda]

Please let me know if I am wrong.

You're an order of magnitude off, 0.44% of 25.36M is 111K not 1.11M. Outcome is still the same as 0.054% is still a statistically insignificant amount to warrant any action.

 

[Sleepy_Hollowed]

I’m going to let you all test this on your machines, thank you very much.

 

[helper800]

You're an order of magnitude off, 0.44% of 25.36M is 111K not 1.11M. Outcome is still the same as 0.054% is still a statistically insignificant amount to warrant any action.

I knew it didn't look right, thanks for that. I'll make the correction. I don't know where that extra 0 came from in the amount of cards result...

 

[Kamen Rider Blade]

Fix the bloody latch on the 12VHPWR to make a nice "Click!" sound.

It shouldn't be that hard.

 

[Amdlova]

Nahhh keeping my old at power supply. The molex adpter it's only thing I need.

I don't trust in these fancy pins, prefer old 8 pins... 3 wires for 12v and five wires ground one's.
More ground more power baby!

 

[YouFilthyHippo]

Actually, in my 40 years of being an electrician i have seen 100's of outlets burn and in a few cases burn the structure down. Some of them were electrician errors ie. forgetting to tighten a screw. Sometimes the little push connectors get overloaded an burn the wires back. Most of the time it was end users using space heaters in old worn out plugs.

It took 80 years for them to finally come up with ARC Fault protectors to detect bad connections and trip the circuit. Maybe we need ARC Fault protects for these videocard plugs?

Arc fault breakers are satans gift to the electrical industry. I live in brand new house, just built in 2019. The arc fault breakers trip randomly all the time. I asked my electrician about it. He said that he gets calls all the time about people in new houses: their arc fault breakers constantly tripping for no reason. I understand the arc fault breakers from a safety standpoint. But too many false positives. Way too many. It's gotten so bad to the point where I stopped using some of the plugins in my house because they keep tripping randomly. It's not just me. As my electrician said, its widespread.

 

[Amdlova]

Arc fault breakers are satans gift to the electrical industry. I live in brand new house, just built in 2019. The arc fault breakers trip randomly all the time. I asked my electrician about it. He said that he gets calls all the time about people in new houses: their arc fault breakers constantly tripping for no reason. I understand the arc fault breakers from a safety standpoint. But too many false positives. Way too many. It's gotten so bad to the point where I stopped using some of the plugins in my house because they keep tripping randomly. It's not just me. As my electrician said, its widespread.

Never used a single secure device on my wires at home. The one thing my house have is the breaker at The eletric company pole.
After 30 year have the change because it's dropping with low amperage. Can hook 15.870w from the wall after break going up.

Just don't lick the cables and you are safe.

 

[randomizer]

Hey I didn't know how to contact the author.

I just took delivery of a Dark Power 13 850w and the PCIe 5 socket is marred / looks similar to that failure, obviously no melted connectors yet as I haven't used it.

The rest of the connections are completely spotless, its only this connector that looks weird.

Gallery with a few more shots

https://imgur.com/a/JG1pJEB

View: https://imgur.com/a/JG1pJEB


Send me a message if you want more information, I've contacted my retailer to see what they think,

Pre-melted to save you time.

 

[edzieba]

That's an odd failure. Pins 1 through 6 are all 12V, so for them to all melt but 7-12 GND to be unaffected could not be from current draw alone without a short to ground elsewhere (e.g. a cable trapped against a side panel and wearing through exposing the 12V lines to chassis GND). Barring cable damage or a short on the GPU, an unseated connector (latch-side more inserted than sense-pin side) would produce the differential heating seen.

Even rocket scientists and engineers make mistakes. Just because a connector may have been plugged in correctly initially doesn't mean it didn't pull out during cable management, other not necessarily related internal work or transport/shipping.

Connector issues have resulted in vehicle losses multiple times. e.g. SAMOS-3 - the ground power ubilical disconnected before the control umbilical, so power was lost to the booster and the vehicle fell back to the pad and exploded; Mercury-Redstone 1 - where a similar umbilical issue resulted in the vehicle falling back to the pad, though this time the stage did not fall far enough to explode (instead the capsule jumped through its flight sequence due to the unusual loss of power, ejecting the escape tower and firing the drogues, main chutes, jettisoning the mains, and firing the reserve parachutes), and more recently Electron Flight 13 "Pics Or It Didn't Happen" - A connector that had even been potted in place (!) had loosened and heated up during flight, ultimately cutting power to the engine controller for the upper stage.
Doesn't matter how many rocket scientists and millions of dollars you throw at the problem, connectors will occasionally end up not connected regardless of how 'foolproof' the connector is.

 

[wayfarer1]

One user, ah.

 

[pooflinger1]

Nahhh keeping my old at power supply. The molex adpter it's only thing I need.

I don't trust in these fancy pins, prefer old 8 pins... 3 wires for 12v and five wires ground one's.
More ground more power baby!

Two of the "ground" connectors are for sense, not ground. It's used for the video card to determine whether a 6 or 8 pin cable is plugged in to adjust power levels. A 6 pin cable has 2 +12v, 2 GND, 1 sense, and 1 NC or not connected pin. An 8 pin connector adds a +12 to the NC of the 6 pin and then an additional ground and sense to the +2 portion. Also, more ground does not equal more power.

 

[InvalidError]

Fix the bloody latch on the 12VHPWR to make a nice "Click!" sound.

It shouldn't be that hard.

The "click" sound is made from the latch falling hard enough to accelerate and make an impact sound. To do that, the fit has to be loose enough to allow a friction-less fall. A looser fit just to produce a louder click would in no way be any safer.

 

[pooflinger1]

The 8.33A rating per pin needed in this application needs some inspection. And I don't believe blaming the user is a reasonable action. People are not responsible and criticized for the plug on their hairdryer for instance. It is expected to work, and safely.

If I recall correctly, Amphenol is the designer/creator of the connector. According to their datasheets, the connector supports 9.5A per pin with all current and data pins energized. It could very well be that some of the failures may be due to companies utilizing cheap/counterfeit parts or not correctly terminating the cables (such as adding solder or using the wrong tool). I also recall seeing that they mandate no bends in the cable within 30mm of the connector to reduce the chance of side-loading the pins.

 

[PEnns]

Nvidia: "It's ALWAYS the users' fault. Now go and buy another of our overpriced faulty junk."

 

[Roland Of Gilead]

Hey I didn't know how to contact the author.

I just took delivery of a Dark Power 13 850w and the PCIe 5 socket is marred / looks similar to that failure, obviously no melted connectors yet as I haven't used it.

The rest of the connections are completely spotless, its only this connector that looks weird.

Gallery with a few more shots

https://imgur.com/a/JG1pJEB

View: https://imgur.com/a/JG1pJEB


Send me a message if you want more information, I've contacted my retailer to see what they think,

Jeez, it really does look scruffy!!

 

[UWguy]

There's a reason there's only three PSU manufacturers I trust: SeaSonic, Enermax, and Superflower, in that order. Not sure who makes BeQuiet since TH stopped updating the PSU manufacturer list a decade ago, but I wouldn't touch them with a parsec long pole.

I’d recommend the Corsair AX1600i or AX1500i if you want the best PSU.

 

[Lokiwoodplan]

Ive read all the comment and all arguments seems make sense. But i still dont know what the right conclusion was.

Is it possible if intel/nvidia still didnt know 100% what the problem is so instead of making a revision design, they choose to wait n see and say it was user error for the time being? Because doing a revision if they are not 100% sure what the problem is and it still happen, the melting after revision, it will be fatal for the company isnt it?

 

[Puffnstuff]

I am running the Nvidia 4 to 1 power connector that came with my 4090 and I can tell you that when you hear the click it is not fully seated. I used a magnifying glass to verify the latch was seated and to my surprise it was not and did not seat until I kept on wiggling the power connector while pressing down. The second click was the latch locking on the notch and it has not come loose since.

The 12vhpwr design is trash and needs to be redone so it latches like the original pcie connectors which are easy peezy. Why not modify them to add a data lane on a solid connector and be done with it.

 

[kjfatl]

I'll bet they didn't have it fully inserted but it may be it is actually one of the few legit faulty connectors/connections. Could be the wire itself or on the PSU side if that is the case.

You are blaming the user when the engineers who designed this connector, developed and approved the associated specification failed in their design efforts. This is a rookie design problem made by a team of top notch engineers. Sure, it works in the lab under ideal conditions, but if fails in real life applications. There are multiple ways to solve this issue. All of them requires a change to the design. The best solution for this custom connector will probably involve the 4 low current signal pins. Make one or more of them shorter so the system can identify if the connector is not properly inserted and keep the system from powering up until the problem is fixed. I did this on a design I worked on in the late 1990's. A similar thing is seen on every USB- A connector where the signal contacts are shorter than the power and ground contacts.

 

[kjfatl]

Hey I didn't know how to contact the author.

I just took delivery of a Dark Power 13 850w and the PCIe 5 socket is marred / looks similar to that failure, obviously no melted connectors yet as I haven't used it.

The rest of the connections are completely spotless, its only this connector that looks weird.

Gallery with a few more shots

https://imgur.com/a/JG1pJEB

View: https://imgur.com/a/JG1pJEB


Send me a message if you want more information, I've contacted my retailer to see what they think,

The arrows on the left and right appear to where lifters exist in the mold to push the plastic header out of the mold. These do not indicate a design or manufacturing issue. If the lifters don't sit perfectly flat, some flash may exist that needs to be cleaned off. This us very common on injection molded parts.

 

[kjfatl]

That's an odd failure. Pins 1 through 6 are all 12V, so for them to all melt but 7-12 GND to be unaffected could not be from current draw alone without a short to ground elsewhere (e.g. a cable trapped against a side panel and wearing through exposing the 12V lines to chassis GND). Barring cable damage or a short on the GPU, an unseated connector (latch-side more inserted than sense-pin side) would produce the differential heating seen.


Connector issues have resulted in vehicle losses multiple times. e.g. SAMOS-3 - the ground power ubilical disconnected before the control umbilical, so power was lost to the booster and the vehicle fell back to the pad and exploded; Mercury-Redstone 1 - where a similar umbilical issue resulted in the vehicle falling back to the pad, though this time the stage did not fall far enough to explode (instead the capsule jumped through its flight sequence due to the unusual loss of power, ejecting the escape tower and firing the drogues, main chutes, jettisoning the mains, and firing the reserve parachutes), and more recently Electron Flight 13 "Pics Or It Didn't Happen" - A connector that had even been potted in place (!) had loosened and heated up during flight, ultimately cutting power to the engine controller for the upper stage.
Doesn't matter how many rocket scientists and millions of dollars you throw at the problem, connectors will occasionally end up not connected regardless of how 'foolproof' the connector is.

If the power pins fail, but the grounds do not, it most likely indicates that there are additional ground paths in the system, most likely the 20 or so ground pins on the PCIE connector.