Seems like GN got to the bottom of the 12PIN connector debacle, or at least provided the best insight so far

[VicBar]

Summary

GN got an independent failure testing lab to do some really great analysis on failed parts and did some of the most in-depth research being able to even replicate the failure on different models of the adapter cables. They

 

Quotes

From the video's Description:

Quote

The RTX 4090 12-pin cables (12VHPWR connectors) have been in the news for burning, melting, and the theories surrounding them for a few weeks now. We ran our first test piece early because we were at a stage where we needed more cables from viewers -- we got some! With the help of our audience sending in a few failed cables, we were able to reach 3 firm conclusions (or, really, 2+1) of causes for failures of these adapter cables (and unadapted ones, too). Of course, there may be even more reasons -- but we believe these 3 to be the primary ones. Between foreign object debris in manufacturing (seems uncommon, but a cause) and user error, we were able to make a few cables melt. On the user error front, remember one key thing: Even if yours doesn't start fully unseated and is only partly unseated, it can work its way loose over time if you move cables around, do some cleanup, move the system, etc.from the article that explains the news in a bit more detail. Don't quote the whole article, but it should be enough to give the reader a reasonable idea about what's going on. 

 

My thoughts

When I know Gn is covering a topic I always wait to get the final word from them. Also I hope they don't get flack for "sort-of, not really but yeah" calling out Igor's Lab for being so conclusive with it's findings without deep enough testing. TechLinked even followed the story from that perspective, so some of the misinformation rant might be directed at them too.

 

Sources

 

 

[Hairless Monkey Boy]

Nice. I'll have to give that a watch later.

[CommanderAlex]

23 minutes ago, VicBar said:

When I know Gn is covering a topic I always wait to get the final word from them. Also I hope they don't get flack for "sort-of, not really but yeah" calling out Igor's Lab for being so conclusive with it's findings without deep enough testing. TechLinked even followed the story from that perspective, so some of the misinformation rant might be directed at them too.

Reminds me of JayzTwoCents covering the 12VHPWR debacle and the last video he[Jay] said that Igor's Lab found the problem and that he won't be discussing more about the 12VHPWR anymore. At 14:50 of the video...

 

[bramturismo]

22 minutes ago, CommanderAlex said:

Reminds me of JayzTwoCents covering the 12VHPWR debacle and the last video he[Jay] said that Igor's Lab found the problem and that he won't be discussing more about the 12VHPWR anymore. At 14:50 of the video...

 

Honestly, when watching Jayz video and seeing how much the cable was purposely ruined to try and trigger a failure made me think how desperate everyone seems to be to make the adapter catch fire. I don't like the adapter at all, but the extent to which people go to make it fail is ridiculous. You can make any cable fail at this rate.

[Stahlmann]

One word: Copium

 

This is basically exactly what i thought from the start; Idiots blaming others for their mistake, then blowing it up on reddit and twitter.

 

GN really went the extra mile to prove it. Very good work on their end. And they've also proven again that the revenue from patreon and their merch really goes back into making better content. The piece of content alone is probably more in-depth than labs will ever be.

[dogwitch]

35 minutes ago, bramturismo said:

Honestly, when watching Jayz video and seeing how much the cable was purposely ruined to try and trigger a failure made me think how desperate everyone seems to be to make the adapter catch fire. I don't like the adapter at all, but the extent to which people go to make it fail is ridiculous. You can make any cable fail at this rate.

also honest question wtf  is with the plug point.

right where the side panel would be? know said design  cabling would not clear most cases?

prop for evga putting it at end of card(j2c video on unreleased card).

[Alex Atkin UK]

Immediately when I heard about the Igor's Lab theory I added them to my "channels that haven't a clue what they are talking about" list as his theory made no sense at all in context of where the plugs were melting.  I wasn't shocked to see Jay believe that theory, as he is the another channel already in that list.

 

Don't get me wrong, I kinda like Jay but I think he gets carried away sometimes.  He used to be humble about not being an expert but sometimes comes across in his videos as being more knowledgeable than he really is which is not good in situations like this.  I mean we've probably all done that, said something with conviction that turns out we were wrong, but I'm no influencer doing that.  The good thing about this forum is if I say something wrong, people WILL correct me, or if I am right people will agree.  One might say were using best-practice actually, as effectively our posts are peer-reviewed.

 

When a problem is potentially this serious, people making random guesses and putting them forward as facts does more harm than good.

[VicBar]

39 minutes ago, Stahlmann said:

One word: Copium

 

This is basically exactly what i thought from the start; Idiots blaming others for their mistake, then blowing it up on reddit and twitter.

 

GN really went the extra mile to prove it. Very good work on their end. And they've also proven again that the revenue from patreon and their merch really goes back into making better content. The piece of content alone is probably more in-depth than labs will ever be.

I agree with the part about GN making the absolute best of their revenue. I wouldn't discount lab that quick, but it does worry me the attitude the main LTT channel has towards the work Labs has done... "Labs did smth, here are some graphs for 5 secs cuz you don't really care that's for nerds, now listen". I expect the in-depth content to be published in writing which I'm fine with. Linus has remained firm on the fact that they will publish written content. Hopefully they provide the world with Crinacle style fully open or "queryable" databases. They are building a very respectable team and it's been slow going but honestly that's fine if they are able to achieve their lofty goals and really shake up the reviewer scene.

[Kisai]

40 minutes ago, Alex Atkin UK said:

Immediately when I heard about the Igor's Lab theory I added them to my "channels that haven't a clue what they are talking about" list as his theory made no sense at all in context of where the plugs were melting.  I wasn't shocked to see Jay believe that theory, as he is the other channel already in that list.

Nah, the reality is that nobody knows why certain connectors burned because usually the GPU's were repaired and burned cables discarded before any forensic analysis was done. Igor didn't have have "burned" cables to test. GN does.

 

40 minutes ago, Alex Atkin UK said:

Don't get me wrong, I kinda like Jay but I think he gets carried away sometimes.  He used to be humble about not being an expert but sometimes comes across in his videos as being more knowledgeable than he really is which is not good in situations like this.  I mean we've probably all done that, said something with conviction that turns out we were wrong, but I'm no influencer doing that.  The good thing about this forum is if I say something wrong, people WILL correct me, or if I am right people will agree.  One might say were using best-practice actually, as effectively our posts are peer-reviewed.

 

When a problem is potentially this serious, people making random guesses and putting them forward as facts does more harm than good.

There is a certain level of "out of my depth" that Jay sometimes sounds like, and in the case of this connector issue starting with "only 30 insertions" where he did that on stream and noted it got looser with each insertion.

 

Now consider what was observed by GN where contaminations contribute to the heating, and the NTK connector having the plating completely worn off after one insertion, with the copper being exposed. which means that over time the NTK connector is more likely to "burn" once the copper oxidizes if it's ever unplugged long enough to do so (which is 20 years in clean air, much less in humid, contaminated air.) I think that's "potential" problem that would not be encountered unless the damage was made before being shipped and then left in a warehouse. The other connector, the one NVIDIA uses, was slightly more durable on this front, but even then. that "30 insertions" thing does seem to reflect the wear to the connector, and since "incomplete insertion" appears to be the main culprit, it's not a stretch to assume that the connector simply has no margin of error, and installing is error-prone.

 

There should have been a much deeper and/or thicker connector. 

 

[CommanderAlex]

2 hours ago, bramturismo said:

Honestly, when watching Jayz video and seeing how much the cable was purposely ruined to try and trigger a failure made me think how desperate everyone seems to be to make the adapter catch fire. I don't like the adapter at all, but the extent to which people go to make it fail is ridiculous. You can make any cable fail at this rate.

Yeah, his video alone seemed to be more entertainment and mocking Nvidia for using the 12VHPWR port. The adaptor that Igor's lab shredded looked very damn desperate ruling out why some adapters are failing and others aren't. 

1 hour ago, dogwitch said:

also honest question wtf  is with the plug point.

right where the side panel would be? know said design  cabling would not clear most cases?

prop for evga putting it at end of card(j2c video on unreleased card).

You know, that's the thing I don't understand why Nvidia changed the ports position between the 30-series and 40-series, from having it set at an angle of 45^o or so to straight out at an angle of 90^o. They purposely introduced the 12VHPWR design with the 30-series, albeit more of a beta and not a complete adoption. I think Nvidia only tests their cards to fit on test-benches and not inside their consumer's cases. 

[goodtofufriday]

2 hours ago, Stahlmann said:

One word: Copium

 

This is basically exactly what i thought from the start; Idiots blaming others for their mistake, then blowing it up on reddit and twitter.

 

GN really went the extra mile to prove it. Very good work on their end. And they've also proven again that the revenue from patreon and their merch really goes back into making better content. The piece of content alone is probably more in-depth than labs will ever be.

Regular uses should not be able to so easily make a product catch fire from regular use through minor mistakes. 

Fail to power the card or some such? Sure. But not set ablaze. That comes down to an issue in Design of so many people are getting it wrong. 

[Alex Atkin UK]

30 minutes ago, CommanderAlex said:

You know, that's the thing I don't understand why Nvidia changed the ports position between the 30-series and 40-series, from having it set at an angle of 45^o or so to straight out at an angle of 90^o. They purposely introduced the 12VHPWR design with the 30-series, albeit more of a beta and not a complete adoption. I think Nvidia only tests their cards to fit on test-benches and not inside their consumer's cases. 

While its bad design, to be fair the FE card fits comfortable in good cases like the Be Quiet DX500 (though the adapter cable is stupidly short):


 Its the AIBs where it becomes a major problem as most are considerably wider so it wont fit comfortably any more.

[Alex Atkin UK]

37 minutes ago, goodtofufriday said:

Regular uses should not be able to so easily make a product catch fire from regular use through minor mistakes. 

Fail to power the card or some such? Sure. But not set ablaze. That comes down to an issue in Design of so many people are getting it wrong. 

Agreed, we honestly should be passed this sort of thing after all the trauma with bad MOLEX connectors.  These new standards were supposed to make it fool proof.  I just don't understand why they didn't design it around the old connector size but with more pins, if they badly wanted it a single connector.

 

What really got me is no click when I inserted the plug and the design of the card meant I can't visually see if the clip is secured or not either.  I pushed down as hard as I could and it still looks like it might not have clicked into place fully, or maybe it has, who knows.  I was careful not to move the cable at all after insertion and will be swapping to the Corsair cable when it arrives.

 

If they had just used a right-angled plug this would have been less of an issue as you can put a lot more force on the back when inserting to be 100% certain its gone all the way in, and also avoid the bend radius issue.  Having all those cables putting stress on such a tiny connector seems really dumb all round.

[Quackers101]

above something that igor's lab mentioned.

its not that GN solved it either, just proved most he can and was hoping for this type of video.

Hopefully a full statement from others and if it could prove if this was the "only issue" being what GN mentions in his video.

Then again PCI-SIG is already done some changes and leaks which sort of suggest it being a bit similar.
Sadly the newer change (if added) would not let the card run if missing more sense pins?

[Alex Atkin UK]

15 minutes ago, Quackers101 said:

above something that igor's lab mentioned.

its not that GN solved it either, just proved most he can and was hoping for this type of video.

Hopefully a full statement from others and if it could prove if this was the "only issue" being what GN mentions in his video.

Then again PCI-SIG is already done some changes and leaks which sort of suggest it being a bit similar.
Sadly the newer change (if added) would not let the card run if missing more sense pins?

GN proved an important point, that uneven load on the pins causing overheating is not true, as even with only two pins connected at 600W load it didn't overheat them.  That kinda flies in the face of not just Jay but Buildzoid too who claims the pins had zero safety margin.

 

Now granted, the problem of them wearing and/or debris causing heat is still quite a big one, but its very different to past theories.

So if that change by PCI-SIG at least ensures the connector has to be fully inserted before working at all, that's a good thing as it should eliminate the issue of the connector not being inserted correctly as being a problem, which it appears is the vast majority of cases causing melting so far.  The rest may require a change in manufacturing to eliminate wear and defects.

[TetraSky]

The people who designed the connector and said it was safe amidst the worries about it be like : "you're using it wrong"

[starsmine]

11 hours ago, Alex Atkin UK said:

Don't get me wrong, I kinda like Jay but I think he gets carried away sometimes.  He used to be humble about not being an expert but sometimes comes across in his videos as being more knowledgeable than he really is which is not good in situations like this.  I mean we've probably all done that, said something with conviction that turns out we were wrong, but I'm no influencer doing that.  The good thing about this forum is if I say something wrong, people WILL correct me, or if I am right people will agree.  One might say were using best-practice actually, as effectively our posts are peer-reviewed.

I dont think Jay was even wrong to spit ball or theory craft in forums or a chat room or irl. It's the fact that he was putting those claims out in an edited and published work that bothers me with how he went about it. For publishing journalistic work, you need to practice editorial discretion on making claims in the authoritarian manor that he did. In ways that can't be proven false immediately, by doing the bare minimum of research. He may not feel like he is an expert and should get a pass, but no that is not how that works for published works. And then taking Igors work and saying, hey problem has been confirmed, when igor did no testing either. Just like holy shit.

He could have posted it in second channel content if he felt like he must post something, to indicate this is conversational rather then some edited work. 

[Quackers101]

5 minutes ago, starsmine said:

He could have posted it in second channel content if he felt like he must post something, to indicate this is conversational rather then some edited work. 

if it was a rant, sure. but to say "hey it's solved, we found the issue, dont worry!"

*sits in a burning house*

[Alex Atkin UK]

7 hours ago, Quackers101 said:

above something that igor's lab mentioned.

its not that GN solved it either, just proved most he can and was hoping for this type of video.

Hopefully a full statement from others and if it could prove if this was the "only issue" being what GN mentions in his video.

Then again PCI-SIG is already done some changes and leaks which sort of suggest it being a bit similar.
Sadly the newer change (if added) would not let the card run if missing more sense pins?

I'd have to disagree, given they inspected actual failed connectors, that's as solved as you're going to get.  He did the same research NVIDIA will, only he is up-front and honest about the result whereas I can't see PCI-SIG or NVIDIA admitting it was a bad design.

 

What boggles my mind is if PCI-SIG are indeed now re-designing the sense pins to only connect once the plug is fully inserted, why on earth wasn't this part of the original design?  The only reason to HAVE sense pins is for that reason, otherwise they could have just mandated that GPUs detect for voltage on the 12V pins to determine if enough PSU cables are connected to an adapter.  It comes across as both an over and under-engineered connector at the same time.

 

When they first announced it I thought it would be able to let the PSU signal to the GPU that it was overloaded and cut-back its power limit, but it seems its actually a pretty dumb solution in reality that seems like those sense pins did not need to exist to achieve the goal.

 

NOTE: Apparently it kinda does this, but only acts at power on based on maximum current of cards connected, not dynamically based on current load:

Quote

By knowing the number of PCIe cards, the PSU can set its SENSE0 and 1 signals NOT to allow the PCIe cards to draw more power than it can support. So say you have a 1200 PSU, which usually supports a single 600W GPU. If you connect two, the PSU will adjust its power output levels on each of the 12VHPWR connectors accordingly to not get overloaded. 

 

[Stahlmann]

13 hours ago, goodtofufriday said:

Regular uses should not be able to so easily make a product catch fire from regular use through minor mistakes. 

Fail to power the card or some such? Sure. But not set ablaze. That comes down to an issue in Design of so many people are getting it wrong. 

Not plugging in a cable all the way isn't a minor mistake. If you can't even do that, you shouldn't build a PC yourself.

[Quackers101]

37 minutes ago, Stahlmann said:

Not plugging in a cable all the way isn't a minor mistake. If you can't even do that, you shouldn't build a PC yourself.

right... but mine dont fit, so how you deal with that salami lid?

I have used motherboard just fine with an "half correctly" inserted connector, where one end didn't want to go into the slot and nearly broke my motherboard in half in how much pressure one had to push on it and got nowhere, worked fine with one end of the connector being 3/4 of the way in and the other was fully connected. 10/10 connectors, 10/10 usability. So again there can be such a difference in quality, fit and finish, and more.

[Gamer Schnitzel]

16 hours ago, CommanderAlex said:

Reminds me of JayzTwoCents covering the 12VHPWR debacle and the last video he[Jay] said that Igor's Lab found the problem and that he won't be discussing more about the 12VHPWR anymore. At 14:50 of the video...

 

Jay is a joke. He just tries to make drama so he can pump out videos since he literally has nothing to offer unlike GN who is not only extremely intelligent but spends his money on acquiring expensive machinery to produce more in-depth videos. Meanwhile Jay releases 100 videos a day where he builds a water cooled PC. Haven't seen a Jay video for half a year and won't anymore either.

[Kinda Bottlenecked]

2 hours ago, Stahlmann said:

Not plugging in a cable all the way isn't a minor mistake. If you can't even do that, you shouldn't build a PC yourself.

 

You'd be surprised how much this mistake happens even amongst boutique SIs. They however do not usually result in burnt components. 

[CarlBar]

16 hours ago, goodtofufriday said:

Regular uses should not be able to so easily make a product catch fire from regular use through minor mistakes. 

Fail to power the card or some such? Sure. But not set ablaze. That comes down to an issue in Design of so many people are getting it wrong. 

 

This, GN even touched on it themselves. I also commented in the video comments that i've had issues in the past with a dodgy connection, (a prior PSU), where one of the SATA power cables wouldn't click into place reliably. If i wasn't careful when cable managing i could put force on the connector through the cable and yank it partway out as a result. I wonder now if the same thing might be happening here, a connector thats awkward to get full seated and cable management, (or possibly just the weight of the cable), pulling on the cable causes it to pull out far further resulting in issues. We also know one of the example cables GN got showed signs of FOD, not incorrect insertion, thats a completely seperate non-user error related issue, and the delaminating platings are a concern too over the longer term.

 

I'll admit i was surprised it suffered no overheating issues when only 2 power pins were connected.Thats definitely well above spec.

[Doobeedoo]

The whole thing is sus, the solder joints are sus, the rail pins are sus, the plastic connector no-clip yet hard push nonsense, still not sure is it good crap, wiggle around on top of that, pressing cables that are near hard plastic connector potentially breakimg solder even. Quite a mess. Also couldn't/shouldn't those sense pins or urm ground ones be longer so they connect first when inserting?

Really I'm surprised new PSs didn't arrive earlier so at least adapter crap can be skipped already. Yes I know those can have issues but that adapter looks like a lap test thing. Weird choice.

Also ffs move power plugs from front and even side of the card and put them at back end so cable comes from case hole directly straight in card.

 

 

Also what about this?