8-pin PCIe to ATX 12VHPWR Adapter Included with RTX 40-series Graphics Cards Has a Limited Service-Life of 30 Connect-Disconnect Cycles

[Solaris17]

and not the same one over and over

what do you mean? I'm assuming your not referencing the pics? Because those are all the same card.

 

[Super Firm Tofu]

what do you mean? I'm assuming your not referencing the pics? Because those are all the same card.

And most of them are the same photo: cropped, colored, etc.

Typical internet mob style.

 

[Mussels]

Sorry that was just from a quick google search, not the same ones i've been seeing

coulda been clearer on that

 

[trsttte]

So many reports of these catching fire and melting on social media - and not the same one over and over


It seems if they're bent in any way they become a massive fire risk, and because the GPU's are huge it's almost impossible for them to NOT bend

View attachment 267006

We'll need bigger cases, or simple 90º connectors...

Nah, bigger cases generates more sales $$$

 

[zlobby]

what do you mean? I'm assuming your not referencing the pics? Because those are all the same card.

Problem is still there! Unless ATX require only certified professionals to install those, abovementioned connectors should be designed as fool-proof.

 

[shovenose]

I hope rdna3 still uses regular pci e power connectors ugh.

 

[Valantar]

I hope rdna3 still uses regular pci e power connectors ugh.

Reportedly they do.

I have to wonder: if the goal of the new connector was higher current in a smaller form factor, why didn't they go the conventional route for something like this and use fewer, larger connectors? XT90 connectors are rated for 90A with just two pins - that's 1080W at 12V. Sure, it likely doesn't have the mechanical characteristics for GPU/AIC use (no locking tabs, insufficient plug retention under stress), but that's fixable. Going with a higher number of smaller pins for more current just seems outright dumb.

 

[trsttte]

Reportedly they do.

I have to wonder: if the goal of the new connector was higher current in a smaller form factor, why didn't they go the conventional route for something like this and use fewer, larger connectors? XT90 connectors are rated for 90A with just two pins - that's 1080W at 12V. Sure, it likely doesn't have the mechanical characteristics for GPU/AIC use (no locking tabs, insufficient plug retention under stress), but that's fixable. Going with a higher number of smaller pins for more current just seems outright dumb.

An XT90 connector uses thicker wires. Not that ATX couldn't move as well to thicker wires...

I agree, this seems like a very unnecessary problem. The previous connectors were small enough to fit within the size constrains of a single slot, why is there any need for a smaller connector, is it just because why not? The common 8 pin connectors were rated by the spec for 150W, but the connector it self can handle more than that.

Not to mention current problem seems to be caused by angling the cables, how did no one thought about 90º connectors when doing the stupid adapters!? or angling the connector on the board like the original 3090 did?

Again, unnecessary problem

 

[jonnyGURU]

So many reports of these catching fire and melting on social media - and not the same one over and over


It seems if they're bent in any way they become a massive fire risk, and because the GPU's are huge it's almost impossible for them to NOT bend

View attachment 267006

Note that they're all the Nvidia adapter. Not any of the native 12VHPWR connectors from Corsair, beQuiet, Cablemod, Gigabyte, MSI, etc. This isn't a connector issue. This is an execution issue.

 

[Valantar]

Note that they're all the Nvidia adapter. Not any of the native 12VHPWR connectors from Corsair, beQuiet, Cablemod, Gigabyte, MSI, etc. This isn't a connector issue. This is an execution issue.

This seems so odd to me - at least from Igor's Lab's teardown of the adapter, it does look pretty well made. But it also has a pretty odd wiring layout, with four 14AWG wire pairs going into an anti-kink boot at the 12VHPWR plug end, where those must be split from 4 to 6 pins somehow. Might this splitting be the cause of the issue? Do you have any idea what the adapter looks like underneath that boot - is there a crossbar, are the individual wires bridged otherwise, are the pins in the plug connected?

An XT90 connector uses thicker wires. Not that ATX couldn't move as well to thicker wires...

Yeah, or they could design a connector that explicitly accepts multiple input wires but bridge them into a single output pin. This would make the connector a tad more expensive to make, but I can't imagine that being a big enough problem to really even consider. I guess one consideration is that XT90s and the like require soldering, while these are all crimped-pin connectors, which are no doubt cheaper and simpler to manufacture (and better suited to being bent and moved around a lot). But still ... there are ways to make bigger crimp terminals too.

I agree, this seems like a very unnecessary problem. The previous connectors were small enough to fit within the size constrains of a single slot, why is there any need for a smaller connector, is it just because why not? The common 8 pin connectors were rated by the spec for 150W, but the connector it self can handle more than that.

Yeah, the 150W rating has tons of headroom - that's just 4.17A/pin, while high quality connectors and pins are rated for 9A or even higher. Safety margins are great, but these might have been a tad excessive. But this new design definitely seems like a step backwards rather than forwards, that's for sure!

 

[ARF]

Yeah, the 150W rating has tons of headroom - that's just 4.17A/pin, while high quality connectors and pins are rated for 9A or even higher. Safety margins are great, but these might have been a tad excessive. But this new design definitely seems like a step backwards rather than forwards, that's for sure!

This is energy - you should never be confident about anything with it. Especially when there are facts that it can lead to fires and burnt homes.

 

[Valantar]

This is energy - you should never be confident about anything with it. Especially when there are facts that it can lead to fires and burnt homes.

... I know. Did saying that the safety margins "might have been a tad excessive" give you the impression that I was arguing for speccing PCIe power cables at 9A/pin?

 

[trsttte]

Yeah, or they could design a connector that explicitly accepts multiple input wires but bridge them into a single output pin. This would make the connector a tad more expensive to make, but I can't imagine that being a big enough problem to really even consider. I guess one consideration is that XT90s and the like require soldering, while these are all crimped-pin connectors, which are no doubt cheaper and simpler to manufacture (and better suited to being bent and moved around a lot). But still ... there are ways to make bigger crimp terminals too.

When they require an active adapter (they have some kind of micro controller there to set the sideband signals according to the 8pin sense pins available) cost already went out of the window

 

[zlobby]

Note that they're all the Nvidia adapter. Not any of the native 12VHPWR connectors from Corsair, beQuiet, Cablemod, Gigabyte, MSI, etc. This isn't a connector issue. This is an execution issue.

Astute observation.

 

[jonnyGURU]

Do you have any idea what the adapter looks like underneath that boot - is there a crossbar, are the individual wires bridged otherwise, are the pins in the plug connected?

Yes.

I don't think it's the crimps. I think it's the way they've taped it up.

 

[ARF]

... I know. Did saying that the safety margins "might have been a tad excessive" give you the impression that I was arguing for speccing PCIe power cables at 9A/pin?

How do they even measure it? Under what conditions and what type of connection? What stress/load on that connection, what ambient temperature, what humidity, what dust/salt particles presence in the surroundings?

 

[Valantar]

How do they even measure it? Under what conditions and what type of connection? What stress/load on that connection, what ambient temperature, what humidity, what dust/salt particles presence in the surroundings?

... these are electrical and mechanical specifications arrived at by industrial components manufacturers, who are subject to a wealth of safety regulations and both national and international standards for electrical safety, fire safety, peroduct testing, and heaps more if they want to sell their products to regulated markets like the EU. If you somehow believe they aren't taking care in specifying the capabilities of their products or haven't tested them thoroughly to meet this specification, you need to think again. Are there unscrupulous OEMs shoveling out over-specced crap? Often, yes. Is this the norm? No - if it was, we'd see a lot more stuff catching fire. Increased regulation has massively improved electrical fire safety in the past few decades.

If you want to see the specific safety standards applicable to something like this you can typically find them listed in the datasheets for said products (look for crimp terminals and housings on Mouser or Elfa Distrelec, for example), and in those standards (which are often puclished openly, at least in part) again often list test procedures and conditions - but at that point you're getting to a level of detail where you'd likely need an engineering degree and a few years' experience in a relevant industry to understand what is really being said, and would need to parse out the useful portions of several hundred page long documents.

 

[zlobby]

Yes.

I don't think it's the crimps. I think it's the way they've taped it up.

The indie review scene misses you dearly, bud!

 

[Valantar]

Yes.

I don't think it's the crimps. I think it's the way they've taped it up.

Seeing Igor's Lab's pics without the tape I'm frankly even more surprised that these are failing - in my experience the kind of anti-kink boot they've got over the 12VHPWR terminals do a good job at strain relief, at least for the XT60s I've had similar boots on. Though I guess you might be onto something with the tape - without that, the wire pairs could bend individually, but the tape forces them to bend as one, which would pull quite hard on the outer wires, and even push on the inner ones. Hmm. Might a preventative quasi-fix for this be to just strip off the tape around the 12VHPWR plug? Would definitely be interesting to see this tested somehow.

 

[jonnyGURU]

Seeing Igor's Lab's pics without the tape I'm frankly even more surprised that these are failing - in my experience the kind of anti-kink boot they've got over the 12VHPWR terminals do a good job at strain relief, at least for the XT60s I've had similar boots on. Though I guess you might be onto something with the tape - without that, the wire pairs could bend individually, but the tape forces them to bend as one, which would pull quite hard on the outer wires, and even push on the inner ones. Hmm. Might a preventative quasi-fix for this be to just strip off the tape around the 12VHPWR plug? Would definitely be interesting to see this tested somehow.

Yeah. It doesn't really make sense. Because, you have that base of the connector. Like you said, that SHOULD act as strain relief and prevent the terminals from going cockeyed. But remember: That piece isn't molded to the connector like the strain relief on a USB cable or something. That piece is only clipped to the sides of the connector to keep it in place. It still moves back and forth. That added leverage could actually exasperate the damage.

I know they initially put that there because people were pulling out the sense pins when they pulled the connector out of the card. So it was there more to give someone somewhere to put their thumb rather than to prevent tight bend radius.

I don't know.... I need to look at it further. But something isn't right.

That said... it's only four examples of "burning connectors" out of how many cards sold? 20K? I would think there'd be more failed cards out there than failed cables. And we don't hear about those.

Here we go..... Igor beat me to it! I wasn't going back into the office until Friday.

The horror has a face - NVIDIA’s hot 12VHPWR adapter for the GeForce RTX 4090 with a built-in breaking point | igor´sLAB

Those who are now beating up on the new 12VHPWR (although I don’t really like the part either) may generate nice traffic with it, but they simply haven’t recognized the actual problem with the…

www.igorslab.de

 

[Valantar]

Here we go..... Igor beat me to it! I wasn't going back into the office until Friday.

Wow. Just wow. That is completely unacceptable, and a clear fire hazard when used in any situation with multiple insertion cycles or bending of the wires. Wtf, Nvidia?

 

[Mussels]

It's a good day when JonnyGuru is showing you links to Igor's lab on TPU



All the best people in the one place

 

[jonnyGURU]

It's a good day when JonnyGuru is showing you links to Igor's lab on TPU



All the best people in the one place

I made a joke about Igor once. He didn't think it was funny. But he's German, so that happens. Bought him a beer. I think we're good now.

 

[Mussels]

I made a joke about Igor once. He didn't think it was funny. But he's German, so that happens. Bought him a beer. I think we're good now.

You aren'y really famous if you haven't got into accidental fights with other famous people

 

[zlobby]

You aren'y really famous if you haven't got into accidental fights with other famous people

There should be an ERB episode 'Igor vs Jonny'!