Igor's Lab figured out why NVIDIA adapters are melting!
2022 ж. 26 Қаз.
1 191 316 Рет қаралды
Read Igor's Labs finding about the NVIDIA 12VHPWR adapter design here - www.igorslab.de/en/adapter-of...
Learn more about the Canvas lineup of Monitors from NZXT at nzxt.co/Canvas_Jay
Get your JayzTwoCents Merch Here! - www.jayztwocents.com
○○○○○○ Items featured in this video available at Amazon ○○○○○○
► Amazon US - bit.ly/1meybOF
► Amazon UK - amzn.to/Zx813L
► Amazon Canada - amzn.to/1tl6vc6
••• Follow me on your favorite Social Media! •••
Facebook: / jayztwocents
Twitter: / jayztwocents
Instagram: / jayztwocents
SUBSCRIBE! bit.ly/sub2JayzTwoCents
Honestly if I am spending this much on a graphics card I shouldn’t need to worry about silly stuff like this.
I like it, kinda spices things up a little
With gpus getting bigger I’m thinking of moving to a custom loop build these card sizes in 40 series is ridiculous
Like i said people buying Nvidia products deserve all the problems, people continue to buy their crap even after all the terrible way they treat their customers
not that it matters because they dont fit in computers.
u pay for the performance,if you want reliability get that extra cable or don't pull the cable sideways
Good to see Nvidia letting the tech community do their R&D for them. Big brain move.
8999 IQ
Its like EA with their games: why release a game with functional audio, controls and video, if the stupid buyers can fix for us?
They're following Microsoft's lead with Windows 10 when they fired their entire QA team of 2500. Since then we've been the tester.
This cable malfunction seem to be extremely rare, and things that are this uncommon, is not always easy to detect in testing. There have been popular brands of cars that are called back in the tens of thousands to fix things like airbag issues.
@@FabledGentleman How does Nvidia know which batch is $h!t tho? Gotta recall them all!!
I think the reason that you couldn't recreate the failure is that you completely removed the wire. The melting is probably caused by a wire that is barely making contact and creating a high resistance connection.
Yeah, it doesn't make sense that the ends are melting unless what you have is a bad connection as opposed to a non-existant connection. If it was that latter you'd see the melting in the middle of the connector.
its definitely arcing caused by the high potential wattage, if it can find a way to the pin, but has to jump a tiny crack, its enough power to do it where a smaller current couldn't jump the arc. Arcing is what is creating the high enough heat to melt these.
Rtx4090 Scalpers have a big problem since they have already bought the gpu with the shity connector & cable.
This! I was getting frustrated when I saw him completely remove the wire and pad. The heat comes from the resistance caused by the loose connection, making the termination point act as a resistive load now, basically a shitty heater lol. This was obvious to me because I work in electrical, but I don't know this guy's background or experience when it comes to electrical.
@@BushyBrowsHD also, he's removed a lot of the plastic/rubber that would be keeping that heat in
You all should check out Gamers Nexus analysis. They did a very deep dive
You have to admit, for someone who didn't want to dive into this, you sure are very persistent. much respect Jay.
Can't hate a man for milking it instead of just making 10 videos just repeating "I told you". He deserves it.
Gotta create content somehow.
Ill let ya in on a little secret, he loves this. probably made more money off this Nvidia shit than any other one fuck up in his entire career. If a youtuber tells you they don't want to talk about a topic, then proceeds to make video after video about it all monetized, they were lying. its stupid not to.
He took that email personally.
That's why I always watch Jay's videos --- he goes into more technical depth on hardware issues than Linus does
Please bear in mind that your FLIR can't accurately read the temperature of bare metal, since its emissivity is very low. You'll get good readings off the nearby plastic, but the traditional thermographers' hack is to stick a bit of matte black tape (I like Taeyoung electrical tape) onto a metal surface and then measure the temperature of the tape. A spritz of black spraypaint also works well.
BLACKBODY RADIATION FTW!!!!
Or something like a thermal pad since they’re designed to conduct
regulary black vynal tape works pretty good to.
Yup except the issue with measuring the plastic is it's such a poor conductor that you're not seeing the temperature around the pins. The surface temp of the connector will almost certainly read much lower. Even if it's only reading ~80C, the pins are likely hot enough to start melting around them, especially if connections are loose.
If the connector plastic is melting, I think the accuracy of any FLIR device is somewhat academic. It's what's IMPORTANT that matters, not the odd degree or two of temperature.
It's been over a decade since I worked at NV, but I vividly remember Jensen ending the last quarterly meeting I attended with the words, "What can you do to improve gross margins?" Clearly, saving a few cents was the goal when they sourced these cables.
Was he polishing his leather jacket that was hung on a golden stand while he was saying that?
@@silverhawkroman Back then, he had more of a "science teacher trying to be cool" vibe than the "low-budget Bond villain" style he's embracing now.
every company is like this. why are you so surprised?
Thats normal for a very successful company, but the goal is how to improve while not failing the product. you could spend 10$ more on the cable or spend 1$ for a foil solution to save and risk the failure. OR spend 5$ to create a practical safe solution instead of either, which is why some people get paid the big bucks is to figure out that much.
This is an example of the problem with modern capitalism. Henry Ford once said, "There is one rule for the industrialist and that is: Make the best quality of goods possible at the lowest cost possible, paying the highest wages possible." Today that phrase is "Capitalism is; the highest profits possible, while paying the lowest wages possible, while producing the lowest cost product acceptable." NVIDIA decided to save a few pennies on their power cable so they can make a bit more profit. The result is that they produced something that has a safety issue in some situations, if they would have left things alone (increasing cost slightly and therefor lowering profits) there would be no problem. The bad publicity and having to deal with repairs for the issue as well as the cost to replace all of the cables before they sell the cards all that haven't been sold yet add up to LESS PROFITS! This is not just an NVIDIA issue, it is an economy wide corporate management issue. Profits are good, trying to squeeze every last penny of profit out of the business is a bad idea. The world will be a better place when people realize that Henry Ford had the right idea and we can get back to that form of capitalism. Sorry for the rant, this is a pet peeve of mine,
Great video Jay! It was a pleasure to collaborate with Igor and contribute to the article.
Hey! Be quiet!
@@OsX86H3AvY No you!
Loving your dark base pro v2 btw...gorgeous (and QUIET) design 😍
Hey Be quiet! I really want your Dark rock pro 4, but man it's so difficult to find it in stock anywhere in South Africa.
Reading Igor's article and seeing the effort that went into your cables, re: crimping, instantly put Be Quiet! to the top of my list for when I buy my next PSU. It's something that's been on my mind for a while, as I know very little about them and have to rely on reviews. But that detail shows a massive level of..well, detail lol, in your construction, which leads me to believe you put that much effort into the rest of it too.
It scares me that a multibillion-dollar operation can miss such a crucial issue in the design process :/
Apparently not every business learned from the Ford Pinto fiasco
Hey now, we can save $0.01 per cable if we do this... that's a new leather jacket.
Its not Nvidia that designed the cable though. Nvidias problem is they want to always be first and bleeding edge to adopt new tech standards, unfortunately this time its bitten them a little.
You've obviously never worked for a large corporation in a design capacity ..... I worked for a major industrial controls corporation in the late 90's and it was a constant battle between the engineers and the Beancounters who wouldn't know an electron from a proton, a bit from a byte or their own ass from a hole in the ground ..... They were always looking for ways to cut corners and cut costs .... I eventually just quit because I got tired of all my designs being re-specced by people who didn't know what they were doing. There is a reason I chose a particular capacitor type and manufacturer Like I used to say: I can do it fast, I can do it cheap or I can do it right .... Take your pick because you only get one choice
Greed knows no bound fam.
I miss the old days when cables were terminated with a crimped on connector that was nestled neatly deep inside the plastic housing where this sort of damage wasn't possible
Honestly why did they do this? I don't see it being cheaper.
crimped connectors are also connected to the isolated wire which acts like a strain release.
In aerospace all connectors are crimps for this reason. Crimps can survive extreme vibration levels that will cause solder to fail. An issue with crimps in commercial applications is that some manufacturers will cheap out and use generic tooling which doesn't crimp to spec. Pretty much all crimped connectors have an associated official crimp die that is only available from the manufacturer. For example the very common Harwin M20 connector (sometimes called a "dupont connector") has an official hand tool that costs $800 and as anyone who has tried to crimp these with off brand tools can attest generic tools make very poor crimps with them.
@@nevoyu Well it is cheaper just putting some solder and also they don't want to make a product that will last ages
@@nevoyu my guess is they pulled a Microsoft and engineered it to fail so you have to buy something else. Microsoft did it with the Xbox 360 video cable made it so it would fail and you would have to get a new one
If you get your hands on a melted connector I would suggest having it x-rayed before trying to cut it open. I'm pretty sure the heat doesn't come after a wire is completely off, instead it happens when the bridge is almost broken increasing the resistance and causing much higher voltage drop at the crack.There also wouldn't be this sort of issue if they had actually bridged all 6 pins together.
Thanks to the wonderful work on GN, it appears that all of these theories (including Igor's Lab) are unfounded and not the issue at all.
Honestly, both the video title and the thumbnail should be changed, because they're blatantly false and misleading. Though, to be fair, they were always false, because Igor just THEORISED. To confidently proclaim that he "figured out why NVIDIA adapters are melting" and that he "found the issue" was always misleading. But to leave it as it is now after GN have published their thorough research moves it from being an excusable error in judgment and critical thinking on Jay's part to outright disregard for truth and accuracy. edit: Ok, now that I've decided to rewatch the video, I think perhaps the entire video should be made unlisted. He mentions that THEIR testing didn't manage to make the adapters melt, but completely glances over the fact that Igor's testing didn't either. He goes on to shred into people who jump out of the woodworks proclaiming that they "know" what the issue is when they actually don't, even though that's exactly what he's doing throughout this entire video as well... The whole video is just one big giant OOF.
@@Veeno00 He is just riding the "nvidia bad" bandwagon.
@@avilashmohanty4786 GOOOOOOOOOOD MORNING SIRS
Igor’s lab is legit!
Noice
Since the connector itself holds everything tightly together, if the pad/solder joint were to break in usage, it would likely be a situation where it either has less contact on that point than before (increasing the resistance), or it is completely broken and arcing, both of which would cause an increase in temperature. Just a thought for why testing with a complete breakage like you did may not produce the same results.
thats exactly what igor stated. wonky contacts inside the solder joints, buzzing and arcing induce temporary high resistance that lead to melting heat.
Yea his understanding and explanations were way off actual logical problem which is arching/heating on damaged outer connections. He shows how outer connections burn and then proceeds to tell how inner ones get more load and overheat...
He was so close...
Exactly! That's the only thing that makes sense and would explain melting on the outer pins. A complete break means no current and no heat. A partial break means increased resistance and heat.
Exactly.. high resistance equals more heat. That's why the middle isn't burnt, has nothing to do with drawing too much from the middle pins.
Hey, this was already completely disproven.
Maybe it's not completely disconnecting that causes the melting, but the solder breaking and having a thinner/loose connection. With a thinner connection, the higher the resistance and therefore more heat.
good point
Thats exactly what makes cables melt, a too thin connection.
From my understanding of Igor's article, it's not actually the solder breaking but the "thin plate" the wire is soldered to. The wires are soldered to a metal foil pad (aka solder pad) that is part of the overall metal foil material going to the pins inside the plastic housing. These solder pads are glued to the connector plastic base, but the soldered wire is peeling the foil pad off of the connector plastic and slowly tears it off from the rest of the foil material. As the tearing gets worse, then of course you get the higher resistance. Not only that, if you tear the pad off completely, the thin foil is so thin that the bridges between the pins heats up a lot. The fact that the solder pads rip off so easily already shows how shoddy Nvidia's connector's build quality is.
yes. see the article in the link.
Exactly how tf did Jay not even comprehend that. I'm watching this video like wtf is he even doing. He even said the same idea when he thought it was the pins pulling out slightly when bending the cable causing more resistance that was melting it. Jays not the smartest tool in the shed.
Like a few have responded, a loose connection would cause more heat. This is especially the case if there is a small air gap and there is enough current to arc across.
Exactly, in fact thats a spike (Ha) in heat. Short circuits if they do not trip a breaker, WILL cause a fire. A poorly made high current connection behaves exactly like a short
I agree
Yea, he's getting it a bit backwards. If it completely snaps off on the outside pins, then more current is going through middle connections and they should heat up. But what is happening is they are cracking and having a smaller area to pass the current and thus heating up. Then it most likely completely snaps off from tearing it apart to look at it.
1000% correct. I do electronics work in an industrial environment and an open circuit wouldn’t cause that pin to melt but a poor connection would. (Well except for the enough current to jump an air gap part. It’s voltage and higher ambient air temperatures that allows for current to flow across air.)
Can confirm, I am heat.
"and they're all saying that they are the ones that know". Including Jay and Igor.
The 'piece of metal' the wires are soldered is commonly referred to as a Bus Bar 😉 This is nuts. Insane that this construction was ever approved.
A penny saved is a penny earned.
@@Mr.Canuck Actually, to make that type of connector would cost more than just the normal crimped connector.
@@dodgem259 Also the added cost of recalls and RMA and lost sales due to bad PR. When you try to be overly stingy you end up paying double than you normally would.
Might as well call them Sus Bar now
@@dodgem259 Yes, exactly my point....they cheaped out. Their complete lack foresight is astounding and will cost them far more than splurging an extra 50 cents on a proper connector to begin with.
Considering that the solderjoint broke but it was still lodged firmly in there, that stuff probably caused intermittent connections (possibly even arcing), which are a LOT worse than a completely open circuit
Totally agree. These tests were done under ideal load, infinite resistance = fine very high resistance=heat. Probably a better test would not need the connector at all. Get a 600w 0.25 resistor and measure the temperature of the resistor. Could probably calculate something close with a watts to heat calculation.
It won't arc if there's a lower-resistance path, and there is because there are still wires connected.
@@Matthew-ou2ie - Sorry, that's not how electricity works. Every complete circuit has SOME power flow. This is why resistors in parallel of different values still work. If I have a 5 ohm and a 500 ohm resistor in parallel, with 12 volts driving through both, all the current doesn't go through the 5 ohm. MOST of it does....but not all. There's a 12v drop across each, so the current through the 5 ohm resistance will be 2.4 A, and the current through the 500 ohm will be 24 mA. While an air-gap has high resistance, the shorter the gap, the lower that resistance becomes......and at very small distances, it can even act as a capacitance as well. Add in thermal expansion of the metal causing the gap to shrink/grow as the plug heats up, and suddenly you can have some serious arcing/heat issues in a connection that is intermittently making connection/losing connection.
@@Matthew-ou2ie the "arc" is on perhaps micrometer crack which still happens but not visible yet still creating enough heat.
You guys with your electrical know how going back and forth haha bottom line don’t buy this crap 😅
It's not just that the outer wires breaking are increasing the load of the center wires. The outer wires may also be experiencing electrical arcing between the broken pieces. Edit:. For anyone that doesn't know, arcing creates heat.
Arcing does seem likely with the way some of those plugs are melted. They're less "Plastic got hot and melted" and more "Plastic caught on fire and melted", which arcing absolutely would do.
I had one of those molded cheapo Molex to Sata (lose your data) that arced on the first use. Can confirm it generates heat and can melt stuff. Edit: It's specially worse when the PSU doesn't cut power and keeps going as if nothing is going on while everything else is burning (I was testing an extremely generic PSU).
Forgive me if I'm wrong as it was a few years ago now that I studied electricity and electronics but don't you only start to get noticeable arcs at high voltages? I seem to remember 22KV arching less than 1cm, so 12V wouldn't even arc a fraction of a fraction of a mm, let alone really carry much current in these cables for it to even be that hot
Can you get arcing when the wires are essentially the same phase, the electrical potential should be too similar? You certainly could get increased contact resistance which makes heat which makes even more resistance..something which is known to create enough heat to start fires. If there is a hair break in the foil it essentially turns into a badly crimped phase wire, something known to cause destructive heat but I've never heard of a single phase, single electrode arcing.
If it's not arcing it might be just not enough contact between the broken wire and the metal plate...causing it to overheat
As an electronics engineer this is really useful to learn from. Had to deal with poorly manufactured cables in the products i develope at work after our purchase department decided to save some bucks. I even implemented an automated cable tester that we use during final assembly of our products.
Boneless chicken.
" I even implemented an automated cable tester that we use during final assembly of our products." were you accordingly compensated for your efforts?
@@LiveBenchmarks Unlikely, from my own experience, but it does make your job easier and gets your work done faster and properly. Having to wait for customer-feedback about how well a fix worked or find a round-about way of testing something takes quite a bit of time.
Honestly I love videos like this. So if a friend asks what he needs for a build, a new cable/certain PSU, is required
Jay, worst than no contacct at all is a bad contact (aka high resistance, but not open). If they cold solder barly touch the foil, then a hot spot is created. It's like having a tiny tiny wire with a lot of current going to it. In your case, there are other ways for the electricity to travel and this shouldn't allow for over heating on the bad contact. On the other hand, if all the wires from a single phase (rail) are cut due to desoldering or breaking pins, then there might be heat created through the bad contact from that rail trying to push its current though a small wire. For exemple, having 3 wires from the first rail on the 3 first solders spots and the last solder spot is being fed from another rail. If this last solder fails and make bad contact, then the heat would spread through pin and then the plastic and melt the connector. This is my theory... Not sure if I'm clear. Englsih isn't my first language.
Spot on! It's the high resistance and intermittent connection of those broken outer wires. I remember as a kid I'd play with the light switches at home, trying to balance the switch between on and off... (made some nice spark noises) I melted 3... biggest hiding I ever got. This is amplified on DC.
I was about to say this. A bad connection is worse than no connection. If only a couple of strands of the wire are connected they can glow due to the high resistance. This can happen if the solder breaks but is held in place by the connector.
@@TheSeriph Yep, and high resistance will increase the total current required for a given amount of power. So the card will pull its 300-600 watts, whatever. But it will probably also pull however much current is required to overcome a poor connection.
Yup, pretty much like an incandescent light bulb
Nope, bad connections carry little current in this situation, and have the *least* heating. Review the PCI-SIG report in Steve's video if you don't believe that.
The issue is the pad/foil thing is breaking but maintaining physical contact allowing it to still carry current but leading to high contact resistance and heating. Break the pad off and like tape it back together to simulate it still being constrained physically by the connector.
This! I don't know much about how this stuff works, but to me it's logical that if the plastic on the sides is melting, it's clearly not an issue of all the current going through the center wires, but rather an issue with the current going through the wires on the sides.
x-actly :)
Yes, a lost joint won't melt because it's not pulling any power to heat up, but leaving a thin connection essentially a very thin wire attached drawing upwards of 100w will heat up like a light filament and melt the plastic.
It’s hard to watch him flail around and miss the obvious isn’t it?
@@THECOD3Y I mean, he's not an electrical engineer. I was hoping Gamers Nexus or LTT would have weighed in on this, because they have people with the skills and equipment to check all of this stuff.
Hi, on the soldering in the top left corner of your opened plug (9:58) it looks like the wire is a bit retracted. This is a possible point of failure too, as this may break under mechanical stress...and it is less surface contact as well as higher possible resistance. So even as it is way better manufactured ,it still could be possibly problematically.
If your motherboard supports the use of a thermal sensor, I suggest you'll strap it to that connector and set your bios to sound an alarm if it exceeds a certain threshold.
Thermo sensor for a connector?
The issue is not that the power goes through the middle wires, but that the corner ones (while not being fully connected) are arching inside the plastic, and that is increasing the temps.
That actually makes perfect sense!
It's called increased resistance.
What doesn’t make sense to me is that the melting is occurring at the pin connection point and not inside the connector where the problem is claimed to be. Hopefully Nvidia will share its findings. It would make sense that the melting would occur at the area with the most resistance which would get the hottest.
@Brian the reason why its melting at the pin side is because that's the thinner part of plastic. If the whole section is heating up, the thinnest part will melt first and everything else will follow.
The outer wires are stressed MORE due to the larger bend radius (and therefore force) on them when flexing the connector assembly. One of the Golden Rules in electrical/electronic power circuits is NOT TO CONCENTRATE CURRENT IN ONE PLACE ! Best practice is to DISTRIBUTE it. E.G. four 6 or 8 pin connectors. This issue of overheating is *SO* well known that it's TRULY INSANE that Nvidia chose to go this way. Nvidia are also pushing the current rating of the contacts BEYOND the Molex ratings (8 Amps maximum per pin - as low as 5 Amps in multi-pin housings, as here).
It's likely not occurring when the pins are totally disconnected, but when the pad breaks and causes an intermittent connection. Burnt connectors like that are usually from wayyy too much current or (more likely) intermittent connections arcing.
Exactly. Plus it seems they're all testing the 12+ lines, where earth/ground is as much likely to cause excessive heat and even fire with a bad/intermittent connection.
From the automotive work I can 100% agree with this. We see this all the time
@@Brocky05 they are testing only the 12V lines, as they are more likely to fail. At least as the pictures of the burnt connectors have shown so far. Every single fail was on a 12V line, none on the ground lines. In general though you are right: In a circuit the fail can happen on both sides, be it on the positive or the negative one.
This is the way…. Completely disconnected = zero current. Failing connection = high current due to increased resistance.
@jayztwocents the issue is, that a crack or a partial break will result in resistance increase and can also cause arcs which also increase heat , best case scenario is a total break
Video like this are why I keep coming back. Thanks for the excellent explanation.
Lmao this aged even worse than I thought it would
Awesome job covering it, and thanks to Igor's Lab. I heard of this and thought it sounded ridiculous since there's no way it could stress with 4 connections, only to discover it was either 3 or 2 that were actually connected, causing high temps due to the resistance.
Take a swig of a nice aged wine every time Jay drops one of the cables onto the table in frustration.
A piece of metal that is designed to be connected to a wire to is called a "terminal". The wire could be attached in different ways depending on the type, such as crimped lug terminal or a solder terminal or a screw terminal.
I would put magsafe on everything
I feel in this application that the best way to do this would be to ultrasonic the wire to terminal buss bar.
Actually the thin strip of metal is called a Bus Bar. A terminal is specifically a device used to terminate the end of a wire (I.e. Post Terminal, Screw Terminal, Push Terminal and the like).
@@BromoDragoonFly magsafe may affect timing too much at 600 watts. It's an awesome connector, don't get me wrong
@@michaellindgren9653 exactly right. I've given up on trying to educate via KZhead comments. Hats off to those of you holding the line lol
I can't believe how many times I pause any of your videos and I get you with your eyes closed
Disregard this video and all others with the exception of Gamers Nexus. Should honestly remove this video as it is all speculation and untrue.
This takes "Nvidia GPU's are ovens" to a whole new level
Ironically enough the 4090 runs cool. Especially compared to my 3080 and 1080ti before it.
@@cosmocalypse3708 Having low temps doesn't mean it's running cool. Your 4090 has a way bigger cooler and draws a lot more power.
@@cosmocalypse3708 @Osmium you both missed the joke. Congratulations. You must both be a riot at parties.
GTX480 wants to know your location.
upgraded to an arc furnace
Hey Jay its worth mentioning that sometimes people pull and squeeze at the wires just behind the plastic housing to unplug it, I've been guilty of doing it in the past after building my system and my hands get a bit sweaty or oily and can't grip the plastic.
Igor's lab didn't figure out shit, why is it the title of the video?
Well done Jay your channel views are smashing it...!
From Igor's teardown it seems pretty clear that the most likely culplret here is either poorly and inconsistently soldered connections causing high internal resistance and then heat as a byproduct, and/or movement of the cable causing the solder joints to weaken or come away from the pads partially, which would cause even higher internal resistance and more heat. It's unlikely the cable would come away from the pad entirely while everything is being held together. Having the wire come away entirely would actually be the better scenario as a severed connection would not generate any heat, but would cause the still functioning pins to be loaded beyond their capacity provided the sense wire is still telling the card that the full 600W is available as Jay describes here. I'm betting the outcome of this is that in all cases either poorly soldered or broken internal connections are to blame.
You also need to consider the fact that manufacturers are required to use unleaded solder. The unleaded solder is harder to work with and results in a joint that is more brittle than leaded solder. In addition the unleaded solder requires higher temperatures which could result in more thermal stress to the metal contact that the wire is soldiered to.
A partial break would also explain why the outsides of the connector melts. A severed outside cable would melt the center of the connector. (Backseat drive with 20/20 hindsight!)
Having already cut back the connector probably helped it not fail. With the full connector still holding split contacts in close contact to one another likely causes arching between the two contacts likely deforming the plastic to a point it becomes a snowball effect on the connector via arching and resulting heat.
This is definitely some good information! When I do decide to upgrade to a 40xx or higher, I will certainly keep this in mind!
Thanks for the video. Very helpful for us consumers to fix the issue and keep using a good card.
Solder joins are notorious for this kind of thing in the automotive industry which is why we rely on crimp joins instead. Solder creates a hard point which can crack or snap with movement and can wick up the copper conductor as well as cause heat stress points. Crimps are purely mechanical... I think we are seeing a temporary or slight connection issue where the solder is broken just enough to create a resistance across the solder join and foil. This would cause a massive amount of heat. And geez, foil metal for this kind of power transfer?...thats pretty nuts. Thanks for sharing J.
Thank you igor's lab andJTC for this. This is why we need guys like you looking out for us.
Don't forget Igors Lab...they actually found the problem as Jay said.
You really should be thanking Igor's lab who found the exact cause. JTC just happened to be one of the first among other youtubers who has a video based on his findings.
You guys are right! Fixed
And social media for spreading the word that there's a problem.
That was very insightful, thanks for the info, J. 👍🏻
NOW it all makes sense! That’s wild!
I think a shoddy connection, disconnected and inside the harness still, arcing to the pad could generate a lot of heat
Just posted the same, melt enough plastic and you will end up with a dead short, if the plastic is melting at circa 120 degrees the wrong plastic has been specified
For sure arcing would be the biggest concern. An absent wire at worst will increase the heat by a little bit on the other three. Concerning, sure, but unlikely to be fire hazard worthy. A small enough gap to create arcing will be drastically more heat, enough to ignite the rubber and plastic housings.
Honestly the old 8 pins were fine they looked better provided better dissipation and didn’t require a stupid adapter. this was a completely unnecessary connection I wish they would just scrap it.
Plus we have cable mod if you wanted a even cleaner look
Igor's article specifically says the issue is not the connector...PSU provided cables use the new spec connector just fine. It's NVIDIA's specific adapter build quality.
Amd will still have it
Amd already said they won't use it, Nvidia is alone in this madness
The connector works perfectly fine. It's the Nvidia adapter that's defective. You folks really ought to watch the video before commenting.
This channel is so great. Filled with useful information for the average Joe.
The average Jay, from JayA to JayZ.
Great work Jay. I am wondering if the missing ingredient is arcing. If the foil breaks and it is close enough, it could potentially arc across producing a lot of heat. No way to prove it though under controlled conditions without running hundreds, maybe thousands of tests, or just getting lucky/unlucky.
Thank you, I really appreciate your efforts to locate and isolate the issue for us. I’m looking forward to that cablemod 90 degree connector for the 4090 series. Going to wait for your test on that piece before I install it.
I think you may have misspoke. If the solder joints broke causing the outside wires to no longer participate in the load, then the inside wires need to pick up the additional draw. And from what it looks like in your pictures, there is a bus bar so the load would still be spread across all the pins. It looks more like poor/intermittent connection is what would cause temps to rise.
That would still be 225W per wire (50% more than the old 8-pin) with the TDP of a 4090 or 300W per wire when two are broken.
It was definitely arcing. 👍
Please delete this video... I think you already saw GN's video
The Problem is not only to disconnect the wires, like you said, they tried it with 1400W and it only started to melt... the bigger problem is when there are tiny little remaining connections from the broken solder + the wires on the outside are only connected to 1 pin and the ones to most potentially brake off, thats why the melted ones on the outside. And thats also the most dangerous reason for cable burns in houses, when cables are not good enough fixed, but still have a little connection. Sry for my bad english
@@vladuk … what?
@@vladuk The government and your social security number.
@@vladuk Also, your comment makes no sense.
When I researched solder vs crimp for power cables last year the one take-away I got was that if you solder the cables, some amount of solder will be sucked into the cable itself and it will harden it, making it easier to break because you essentially created a kind of lever. Thats why crimping cables is the superior technique because the cable stays flexible.
yup unless you talk to the clueless aka the internet who think there's a debate. The car channels are the worst for that. Soldering wires are for hacks if it moves/bends/vibrates.
Yeah crimping allows for flexibility, I've always preferred crimping cables rather than soldering. Also makes them slightly more serviceable.
You should get ElectroBOOM to test these connectors for you.
I like the enthusiastic end to the video.
Might not only be the one problem at work here, if the contact area of the pin with the plug is smaller because its not fitted well, might increase the temps to melting point. If you bend the cable enough to force open the plug a little or not have it in all the way or there are a few plugs that are oversized or shorter maybe?
You're mostly right. What you're missing is that... A break will remain electrically connected, but because the connection is so small/weak, the resistance will cause heat. Imagine running your stove on cat5. :P It could be the side bending does not cause a break, but causes the side pin to have a weaker connection, which increases resistance and heats up. In short: actual breaks could occur, but don't need to.
As Jay pointed out, the fact that current adaptor is soldered and not crimped like the 3090's is the issue. Even when done properly, a soldered joint can't hand the same strain as a crimped connection. The danger is when the soldered connection breaks but isn't fully separated, it can still make a connection but with high resistance and will heat up and possibly arc as well.
Already had a talk with a friend about it today. Power density is about twice of the old 8 pin plugs. 600W on this connector with 6 pins vs 150W on the 8-pin on 3 pins, so the old ones have about 50W or 4.17A per pin while the new one allows for 100W or 8.33A per pin. Even worse considering that there are only 4 wires delivering to those pins. That would be 150W or 12.5A per wire. And yes I know, the 4090 isn't specced for 600W, but the adapter is.
That's exactly why automotive technicians are taught by factory training to crimp wire repairs in harnesses instead of soldering them. Solder joints don't like be jostled around and vibrated, which happens quite a bit in a moving vehicle.
your wrong a soldiered connection is much stronger than a crimp connection if done correctly
@@craigbomer8962 Technicians are taught to crimp connections because it is easier for lower skilled people and can be done in less time.
A properly soldered connection is better than a crimped connection in almost all use cases. It is easier and cheaper to properly make crimped connections. It is more expensive to train personnel to properly make soldered connections. Most likely these cables are soldered by machine in an automated process and that machine needs to be reprogrammed/adjusted for proper wicking of the solder in the wire.
one thing i found funny with this whole story is people claimed its not that big of a deal, yet if it wasnt i wouldnt have known about it cause it wouldnt have hit media
seems right, i build cables industrially and have the molex crimpers for the standard plugs, not so much computers but of all the assemblies ive built ive never see a wire split up between pins like that. my first thought was to run individual wires out of that connector and do the actual splicing somewhere else in the assembly like maybe alittle molded box in the middle of the cable
Connectors crimped onto cables are actually less reliable than a soldered connection. Less electrically sound too.. more resistance.
@@poison7512 i mean we have all the equipment necessary to test that hypothesis but industrially speaking i can crimp several thousand crimps in a day but usually the solder connections take alittle longer heating and cooling so the difference is usually outweighed by time savings in production
@@tom940 oh yeah 100% crimped is great for manufacturing and is more than reliable enough to get most jobs done. But if we're nitpicking details.. soldering is the way to go.
It might also be possible where the wire/bridge breaks but because of the housing the contact stays in proximity and could begin to arc, which might be where more heat is generated.
@@kasparjensen1822 false.
30,000 volts per centimer is generally required for arcing, with a minimum of 327 volts to arc. So yeah, no arcing from 12 volts. Familiarize yourself with Paschen's Law, or at minimum some electrical engineering.
I think this issue is also causing a form of arcing between the broken pin and the backplane. That would artificially create a larger heat issue even if the arc is relatively small.
And the fact that the electric arc temperatures are almost similar to lightning bolts (over 10000 degrees Celcius) makes it even scarier
Depends. At the time of soldering of the stranded wire onto the crimp whether or not there was a loose strand before it was all encapsulated in plastic. Should of been found during QA, before that was stage though.
Is the melting occurring at the solder joint though (basically under that molded plastic piece near the bottom? It looks like the damage is inside the female side (cable side) of the connector socket, I have to imagine some level of arcing is occurring due to pin connection, perhaps the tension/compression on the wires from bending is causing the metal to separate in some way.
@@Mike__B I think the heat source is likely at or very near the solder connection (be it from a high resistance poor connection or arcing from the connection opening and closing), but because that epoxy/rubber material is pretty thick in that area it isn't the first area to melt, the thin-walled plastic around the keying areas of the connector is much more likely to melt even if it is a few mm from the initial heat source, the pins being metal (obviously) transfer that heat to the weaker plastic area quickly.
The piece of foil it's soldered to is coming lose from the bridge and that looseness is causing arching which increase the resistance/heat. Their using only 4 wires so that 50 extra amps is alternating back and forth from the connector that's lose and 2-3 that are still connected to the pin bridge. That transfer also increases resistance further. Overall it was a cost saving measure on a device that cost someone $1600 and they cheapest out on their adapter. Corsairs cable that goes from the PSU to the block uses 14 wires. (12 for the pins and 2 sense wire to the 4 pin top bit)
I bough a Rog 4090 last week and I'm using that exact type of cable (type 4 corsair) with my HX1500i. Having heard about melting connectors I preferred to minimize the risks and not use the 4-1 adapter that came with the card.
I just want to give Jay a big thank you for crediting Igors Lab in the fricking title, and also not giving completely inn to the clickbait culture that even the most serious tech channels like LTT have given in to.
On another note I love you guys and all the content you put out. You are greatly appreciated guys. Keep it up
Noice
Thank you, the extended coverage on this has been great to raise awareness and get the info out there. I never liked the look of that cable anyway, because the 4 into 1 just doesn't sit right to me. I hope they get something better out to people have issues and refine these connectors for the next generation of cards, or at least the new PSU standards will hopefully resolve this.
Just buy a custom PSU 12 pin from cablemod. Looks way better and works 👍
@@aqulex84 yes buying overpriced cheaply made cables from that shit site is surely the answer!! No people should stop trying to cram a card in their pc that shouldn't be crammed in.
@@SPG8989 lol. Keep doing you thing. Using cablemod customs over 10 years, never had any problems qualitywise not looks or price. The new rtfm episode especially the beginning 20 minutes just describe your attitude perfectly.
@@aqulex84 everything that starts with "just..." is in most cases ignoring the core problem. sure you can circle around this issue, but it does not change the fact that nvidia has to be held responsible for this and they have also to act on it to make it up to whoever is affected if not providing a redesigned adapter for free to literally everyone who can prove via invoice that they bought a 4090. to me this last suggestion is the only right thing to do.
Yes there are a few psu's put right now and it comes with the nice clean looking cable too 12vpwr
Dude your torso is so massive and your forearms‘ dimensions are absolute intense, you have the apperance of a grizzly bear! I am flabbergasted and impressed. Unreal
I am a mechanic by trade and a computer enthusiast as a hobby. I have seen these types of failures in heavy equipment many times. I think there might be a step that is being missed in these videos. The tests done by Steve and Jay have had the strain relief removed during testing which can hide a phenomenon where a solder joint will break but the wire will still be contacting the mating surface by the force of the tape holding it down creating an insane amount of resistance. If a wire was broken but still being held on the pad out could generate the type of heat necessary to melt the connector. It might be worth trying to break or desolder the wire and then rewrap it to give it just enough contact for electricity to pass through with a very high resistance and see what happens.
I'm glad someone's figured this out
Cold solder joints could cause a poor bond and conductivity, fatigue from lateral bending would compound the issue. Great video!
I feel like these 4090 power adapter videos are gonna accompany us on this channel for the next 8-10 years
Thanks for the video. Great stuff.
what you are missing is the disconnected wires being in close proximity of the pad. This will cause arcing. It is the arcing that produces significant heat and will cause the connector to melt.
Not arcing, but a small contact, like your toaster, lots of power though a little wire makes a lot of heat.
@@TheMonthlyJack if it was due to a small wire it would melt around the remaining connected wires. as those would be the ones getting hot. however the melting is occurring on the outsides of the connecters not the plastic in contact with the remaining connect wires. its is melting around the broken wires. therefore arcing is the more likely cause of the melting.
As others have stated, a completely cut off wire/solder joint would not trigger the failure, most probable cause is the solder breaking of but slightly making contact thus raising resistance, and even maybe arcing
This is a huge Problem in some old VW Polos too, the problem is not that the cable completly brakes off, but its that it has a bad contact. If the contact is bad and has a very small surface, it will get very hot. This worses the resistance even more. So you should try to not brake the wire completly off, more like creat a small contact surface
Thank you Jayz and Igor’s Lab for keeping bringing the facts to the forefront. I’m sure Nvidia would love to keep a lid on this and deal with individual consumers making them think they got extremely unlucky. Canceling my order, I am not buying that card until Nvidia clarifies this mess.
I do not blame you most of time I get nvidia I have a 3080 but when I need to upgrade in the next year or 2 I think I will go with and graphics unless intel cards get better by then
And graphics
The next few months will definitely be interesting as more and more cards get into users’ hands.
Very interesting indeed.
Thanks Jay for making it clear that I don't have to worry about my 3070 TI Founders Card, that has the same kind of connector and it being safe to use. I was a bit worried until watching this video.
Right, I had the same GPU, but I don't ever force tight bends on heavy duty connections either so...Plus, we're pulling 1/3 of the power, probably less.
"it would break or crack the joint..." yeah, joint would be an adequate description of this connection although "solder joint" is much more accurate. You're absolutely right that creating a crack in the solder joint will increase the resistance (measured in ohms/Ω) at the compromized solder joint. The photo from Igor's Lab looks like a cold solder joint meaning that either the solder, pad or wire did not some up to temp before removing the soldering iron.
Just a speculation from an electrician on why you couldn't recreate it. Completely removing the connection doesn't always bring temps up higher when ran in parallel. A lot of times when you see the higher temps it's because of a loose connection adding more resistance. Yes the other wires will get slightly hotter, but you are loosing that extra resistance from the loose connection, thus heat being generated. Once again, just a speculation. Love your videos and keep it up.
Not sure about others but you're testing it in an air-con room. In a case where ambient is 40-50C, connector temp is probably way higher than 64C. Going over to Igor's lab to see their testing.
Jay's soldered joints also exposed to air, while it should be in tight plastic mold (pretty sure very bad for heat exchange)
Failure of those plugs reminds me of melting of old tamiya battery plugs in my younger year. One thing I learned then was items heated up and melted where the greatest resistance was... the fact your photos showed it melting around the pins themselves makes me wonder if the individual pins and sockets are spec'd high enough for the amp draw. Breaking off a wire won't increase the draw from a pin it would cause the feed lines to heat, essentially reducing the effective size of the wires feeding the plug. What I do see from the pictures is potentially routing the wires pulling on and stressing the plug which is only retained by one center clip could cause slight loss of contact on the outer pins which essentially adds resistance and heat down near the plugs. This potentially aggravated by the thin walls of the plugs not able to clamp well enough to a pin that looks more suited to data than power delivery. It should be easy enough to test for with a simple resistance test with a multimeter while flexing the plug. I am probably wrong, this is just my experience with plugs and power delivery on a hobby scale.
Mmmmm... ckay
A good video got new info on this too :D Any good reason why the wires are not crimped?
is this problem only on rtx 4090 or other models have it as well? am i safe to use the cable provided in my 4070 ti box?
It could be a combination of things like a partial crack where it archs intermittently and in a pc case with possible poor air movement can cause it to heat up to its melting point as the case start getting heat soaked increasing overall temps in case and room.
What is it’s also the fact that the cable overlaps to a different contact when bent? This would increase resistance on that contact and possibly burn the pins in there?
I am very glad you are doing this and following this Jay, this is really important and show to the face of the corporations that this can't happen
And I am very glad that I bought an ATX 3.0 PSU with native 12VHPWR support when I built my new PC. No stupid faulty adapter for me!
If anyone with access to a vented test lab, n-methylpyrrolidone can be used to dissolve the connector without damaging the metal bus bar. Ive used it to look for shorts in faulty encapsulated motor stators. Itll dissolve everything but the metal so its good when you need to remove plastic around small electronics to look for breaks or shorts.
Tip for anyone with a 40 series card already. Cablemod sells a custom cable with no reported issues.
Since the wire would break from the solder joint, try intermittently touching and removing the wire from the foil connecting pad. The loose / broken connection would definitely cause a high resistance connection which would result in more heat.
November 3rd can't come fast enough for RDNA3. 😎👍
Igor's the man. No second opinion
Jay is the man.
Hi Jay, I used to be battery tester on NiCad/Li-ion, the killer is down to resistance and timing. If the resistance on the plug gets hot enough it will build up oxide surround those pins, it stays hot over 70C and will increase in temperature as oxide build increase. It can take 100 hours more or less till the plug get so hot it eventually starts to smoke and that where the danger becomes.
If the side pins were the ones getting disconnected, that means the bulk of the resistance would be in the middle of the plug, and also where most of the heat is generated, so pins on the middle would be melting. For the sides to be melting it means that the middle pins aren't connected, however i personally think there's arcing on the sides, because Nvidia uses split barrel connectors inside, that will spread open when there's a bend sideways on the cable, causing a bad contact and arcing and then melting the plug
This is what I was initially thinking too, but isn't the voltage too low to create arcing?
You are correct arcing generates several times more heat than straight power draw. Given the bus bar’s design it will start internally arcing if damaged. This is less a failure in the new ATX spec than a failure in Nvidia’s implementation of the adapter.
Short version: Someone who didn't know WTF they were doing tried to cost down the cable. This is also why you typically want load-carrying wires in the center of a connector, and signal wires on the outside. The wire CSA is too small for the application. A higher strand count would also help mitigate broken strands. Soldering is also a very inconsistent practice in manufacturing, especially with modern lead-free solders.
I learned the hard way as a young engineer about compression pin fretting corrosion and high power interfaces. When high pulsed currents pass through a low compression interface heating causes the metal to expand and contract. This results in micro movements between the male and female parts breaking the tin oxide barrier and allowing oxygen to interact with the base metal and form microscopic corrosion. This increases resistance which in turn increases heat in turn increasing the fretting action and eventually thermal runaway melting your connection. This effect can also occur with even micro vibration between surfaces.