News Nvidia on Melting 16-Pin Adapter: Connectors Weren’t Plugged in Correctly

[Admin]

Nvidia says that user error is the root cause of the recent power adaptor troubles.

Nvidia on Melting 16-Pin Adapter: Connectors Weren’t Plugged in Correctly : Read more

 

[USAFRet]

Why is it that easy to plug them in incorrectly?

You can't guard against every user error, but problems like this should have been designed out in the testing phase.

 

[spongiemaster]

Why is it that easy to plug them in incorrectly?

You can't guard against every user error, but problems like this should have been designed out in the testing phase.

50 so far out of an estimated 125,000 cards sold doesn't exactly indicate it is that easy to screw it up.

 

[USAFRet]

50 so far out of an estimated 125,000 cards sold doesn't exactly indicate it is that easy to screw it up.

More than previous high power components?
Seems so.

I have NO dog in this fight, either way.

 

[spongiemaster]

More than previous high power components?
Seems so.

I have NO dog in this fight, either way.

Yes, definitely more so than previous implementations. Being easier to do and just plain being easy are not the same. Something can be easier and still be extremely difficult to do.

 

[DRagor]

This is exactly what's troubling me. GPU use power connectors for about 20 years, and never before we heard about so many connectors melting. I dare to ask, is it that suddenly so many GPU users become stupid enough to not know how to plug in the connector correctly? Or is actually some kind of a problem with new connector?

 

[purple_dragon]

I guess those sense wires aren't doing their job

 

[spongiemaster]

I dare to ask, is it that suddenly so many GPU users become stupid enough to not know how to plug in the connector correctly? Or is actually some kind of a problem with new connector?

As someone who has done tech support for the people around me for decades, plugging in cables, and components incorrectly is a common mistake. I've seen all sorts of cabling mistakes, ram forced in backwards into dimm slots, partially installed PCI cards. You name it, I've probably seen it. The difference here is those mistakes usually aren't destructive and cause physical damage to components, like they can in this case.

 

[RichardtST]

This is obviously a dirtbag-lawyer-cya statement from Nvidia. The problem is obviously a connector that is too small for the power rating and too difficult to be plugged in properly, and even then, as GN showed, nearly microscopic particles leftover from manufacturing can fry the thing too. It is, simply, a hideous connector standard and needs to be retracted. The problem is that no one can retract it without becoming a lawsuit target. Good for AMD for not using it in the first place.

I for one will refuse to accept and adopt this defective connector. I suggest people simply refuse to buy anything with this crap on it. Standards tend to die when no one uses them.

 

[hotaru251]

You can't guard against every user error, but problems like this should have been designed out in the testing phase.

exactly.

connections are supposed to be idiot proof.

we did it with keyed designs and then clips on 1 side.

this might be "user error" but its caused by design error.

 

[TJ Hooker]

connections are supposed to be idiot proof.

we did it with keyed designs and then clips on 1 side.

this might be "user error" but its caused by design error.

"If you design something to be idiot-proof, the universe will design a better idiot" -Unknown

Also, the 12VHPWR connector is keyed and has a clip.

Edit: I won't argue what the 'correct' level of fool-proofing is, or if the 12VHPWR meets that threshold. But saying a connector (or anything) should be completely idiot-proof is nonsense, because that would be an impossible task.

 

[Chung Leong]

We're talking about an activity here that voids the warranty of most PCs. It's not meant to be done by idiots. Having said that, graphic card vendors have been actively courting the idiot crowd with their blinking lights and all.

In the old days, when you messed up, you just humbly eat the cost of the damaged hardware. Now, it's not enough the company is providing free replacements. No, no. The online mob demands one night of rehabilitation.

 

[bigdragon]

At least 50 cases of melting connectors already? Time for a recall. The US CPSC and European equivalents need to get off their butts and do their job. Cables should be idiot-proof. We shouldn't be content with Nvidia saying "you're plugging it in wrong." A $1,600+ GPU shouldn't be so prone to melting connectors. There would have been a recall already if melting affected a wire harness in a vehicle or terminal connection in a home appliance.

 

[PiranhaTech]

This is exactly what's troubling me. GPU use power connectors for about 20 years, and never before we heard about so many connectors melting. I dare to ask, is it that suddenly so many GPU users become stupid enough to not know how to plug in the connector correctly? Or is actually some kind of a problem with new connector?

A lot of the last question, a little of the previous one. The adapter is a huge 4-to-1 adapter that's carrying a lot of juice. The only adapters I've seen like this are the likes of SAS to SATA, and those don't have a lot of electricity running through them.

Please note that the following is a hypothetical, but very possible

You have this 4-to-1 adapter in a case, and it takes up a lot of space. You have $1600 to buy a GPU, you lay out the cables, and now you want to tuck them in so that your GPU looks pretty with that clear-sided PC case you bought. 4 PEG cables is a lot of tucking. Doing things like tucking in the cable, moving it around even by plugging in one of the 4 PEG connectors, etc can loosen it. Okay, you laid out the 5 cables (4 PEG + adapter), and now you need to route and tuck in other cables. Did you un-tuck or loosen one of the GPU cables without knowing it? "The card is running as far as I can tell, so nothing wrong here!"

If you have a tight fit and a non-clear window case, did you shove the side panel on? I've done this... and now I'll be getting full ATX cases from now on.

I knew a guy building a PC, took shortcuts even though it had an expensive CPU (the CPU was over $400) and had this "Ahhh, it's fine..." attitude. If it were a budget CPU, maybe. If you are loading it with mid-tier to premium components, yeah, you shouldn't mess around as much.

 

[ThatMouse]

I've never heard of a cable that is designed to melt when not fully seated. Well, until now.

 

[TJ Hooker]

A lot of the last question, a little of the previous one. The adapter is a huge 4-to-1 adapter that's carrying a lot of juice. The only adapters I've seen like this are the likes of SAS to SATA, and those don't have a lot of electricity running through them.

Please note that this is a hypothetical, but very possible

You have this 4-to-1 adapter in a case, and it takes up a lot of space. You have $1600 to buy a GPU, you lay out the cables, and now you want to tuck them in so that your GPU looks pretty with that clear-sided PC case you bought. 4 PEG cables is a lot of tucking. Doing things like tucking in the cable, moving it around even by plugging in one of the 4 PEG connectors, etc can loosen it. Okay, you laid out the 5 cables (4 PEG + adapter), and now you need to route and tuck in other cables. Did you un-tuck or loosen one of the GPU cables without knowing it? "The card is running as far as I can tell, so nothing wrong here!"

If you have a tight fit and a non-clear window case, did you shove the side panel on? I've done this... and now I'll be getting full ATX cases from now on.

I knew a guy building a PC, took shortcuts even though it had an expensive CPU (the CPU was over $400) and had this "Ahhh, it's fine..." attitude. If it were a budget CPU, maybe. If you are loading it with mid-tier to premium components, yeah, you shouldn't mess around as much.

1. The 12VHPWR connector doesn't inherently involve a 4 to 1 adapter
2. The fact that there are reports of native 12VHPWR connectors melting indicates the issue isn't unique to the 4 to 1 adapters
3. If the cables are plugged in all the way, such that the clips have engaged, you don't have to worry about them working loose

 

[InvalidError]

More than previous high power components?
Seems so.

The new connector passes 2X as much current per pin as the old 150W ones it replaces. Seems natural that you'd get at least twice as many failures from having about half as much tolerance to user error and defects.

 

[peachpuff]

I guess those sense wires aren't doing their job

They only check if there's a ground, they don't sense anything.

 

[InvalidError]

They only check if there's a ground, they don't sense anything.

They allow the GPU to "sense" how much power it is supposed to be able to draw from the PSU.

When the talk of "sense" signals first started, I was expecting that to mean the PSU or GPU would be monitoring voltage drop across the cable and connectors from PSU to GPU so the VRM can automatically throttle max current based on total wiring and connector losses. That would likely have prevented connector meltdowns by limiting total wiring losses to 10W or less.

 

[Blackink]

This is exactly what's troubling me. GPU use power connectors for about 20 years, and never before we heard about so many connectors melting. I dare to ask, is it that suddenly so many GPU users become stupid enough to not know how to plug in the connector correctly? Or is actually some kind of a problem with new connector?

You make a great point!

 

[-Fran-]

I just watched GN's video and it seems to be fairly conclusive. If there's a secondary issue, it's probably placebo or "one-in-a-million" type of issue.

This makes me believe PCI-SIG needs to go back to the drawing board and make a v2 of the connector. Just make it a tad bigger/fatter on the plastic side and keep the same cables and overall layout. It's not like having the same overall dimension from the old 6/8 pin (per pin, that is) is going to blow up if you make it 12+4 and just a tad bigger than now. It'll give you enough extra plastic in the connector composition to be sturdy enough to be plugged in correctly, without a doubt. We don't need a "super precise" plastic connector, just let us brutes enjoy something easy to plug in; specially when dealing with a lot of power in transit.

Regards.

 

[InvalidError]

It'll give you enough extra plastic in the connector composition to be sturdy enough to be plugged in correctly, without a doubt.

At what point in the GN video does it say that the connector isn't sturdy enough as-is? Nowhere. There is no structural problem with the connector that twice as much plastic would solve. The closest thing to a structural issue that gets mentioned in the video is not enough of a positive click to reliably tell when the connector is deep enough for the retention clip to latch on. Foreign objects in the plastic itself don't really matter unless they are conductive, could just be leftovers from the plastic melter/mixer between batches of different plastics or contaminants in recycled plastic.

 

[hotaru251]

Also, the 12VHPWR connector is keyed and has a clip.

yes.
but they made it difficult to visually feel/see if its 100% in correctly.

it should of had a better feedback (via a more noticeable click or so on the plug)


also there are proper "idiot proof" that no person can fail.

take custom cables for keyboard/mic/etc.

they have physical mechanisms that hold em together and wont work w/o being done properly.

again it might be "user error" but nvidia is also at fault for making it easy to make error on.

they arent new to the scene. they should of known better.

 

[little_me]

This is obviously a dirtbag-lawyer-cya statement from Nvidia. The problem is obviously a connector that is too small for the power rating and too difficult to be plugged in properly, and even then, as GN showed, nearly microscopic particles leftover from manufacturing can fry the thing too. It is, simply, a hideous connector standard and needs to be retracted. The problem is that no one can retract it without becoming a lawsuit target. Good for AMD for not using it in the first place.

I for one will refuse to accept and adopt this defective connector. I suggest people simply refuse to buy anything with this crap on it. Standards tend to die when no one uses them.

If you watched GN video, you'd notice that they TRIED to get connector to melt by cutting off 4 of 6 12V pins to GPU, meaning full 600W was going through TWO pins, not 6.
hint: it refused to melt.

They only check if there's a ground, they don't sense anything.

Indeed, the way it should have been designed *(hindsight is 20-20) would have been to make sense pins shorter than all others, so that if connector is not fully inserted, sense pins see nothing (and GPU refuses to power up with only two PCI-E's connnected) so GPU shuts down upon connector wiggling loose.
not fun but prevents melting.
smaller tolerances could give a way for possible way to "it's connected but sense pins are not working right for reason X" problems.

and according to GN, connector needs to be relatively much (Was it around 2.5mm) out AND bent for melty stuff. Connector can also seemingly be extremely easily wiggled loose, even from 99% in connection if it's not clicked while being cable managed, which was also shown in GN video. (4 to 5 back and forth wiggles was all it needed)
not "it drops to floor" out but out enough to cause the heating up issue if bent.
and bent it shall be due to connector placement being where it is, if you want to manage your cables.

not necessarily only way to get them to melt but... so far only way found to be reproducible.

 

[elforeign]

Why is it that easy to plug them in incorrectly?

You can't guard against every user error, but problems like this should have been designed out in the testing phase.

After having watched the Gamers Nexus video on the melting issue and the subsequent video on Nvidia's response, it really does not surprise me that improperly connected cables are the main source of the issue. While i've never experienced a cable meltdown on any PC i've built, I have had issues with cables in the past. Especially the motherboard 24 pin cable not inserting properly without some proper force. I can specifically recall on a Gigabyte Z97 board I ran my 4670K on, that connector required quite a bit of force before hearing that satisfying click to ensure the cable was properly connected and latched.

Also, on a pair of 970's I ran which used a 6+2 connector, i've also had to use quite a bit of force to get them to properly connect and latch. Now, I would not have imagined this to be an issue in the past, but ensuring proper connection of the cable for me has always been something to look out for on components. It seems a better design needs to be implemented by PCI SIG to ensure users don't run into this issue in the future. While 50 cables out of over 100,000 seems miniscule, the entire industry will likely move to using this connector in the future and the failure will become more pronounced.