UPDATE: In light of the excellent investigative video posted by Steve / GamersNexus (ruclips.net/video/EIKjZ1djp8c/видео.html), it would appear that this problem is still unresolved and there's more to it than the 14AWG wire solder points in the plug. Steve is asking for more people with these adapters to get in touch with him, and of course NVIDIA is investigating although they haven't made any statements beyond that. Locating more adapters with the 150V rated wires that IgorsLab saw (mine used 300V wires which can be seen at 8:26) would definitely help. Most are still leaning towards this being a manufacturing issue with the adapters, but what specifically the defect is remains to be seen.
Seems there are different versions of the adapter. @10:14 yours looks exactly like Igor's version and the outer cables soldered separately and not bridging with the adjacent cables. While GN's ones are different and seem more sturdy.
@@kamui004 yeah nvidia definitely spotted the issue, redid the cable design, but the 1st ones were already shipped (with 150v wires) and they should have recalled them
Paul, honestly speaking, has it ever occurred to anyone that the damages could be produced by shills with a flame and posted as "it happened naturally without me doing anything wrong!"? Has either you, Steve or Jay ever seen the problem firsthand? Ever? And no the story of first ones being faulty doesn`t stand. Both GamerNexus and JayZ have review samples sent in advance and definitely belonging to the first batch. Have they (Jay or Steeve) been able to reproduce the problem even after torturing the connector in the worst possible ways? And if not, is anyone starting to smell the much ado-about-nothing stink? (shills melting the connector with a flame anyone). P.S. I have been torturing my 4090 (MSI Suprim X purchased day 1) for days now, overclocking, 6 hours of consecutive gameplay at Cyberpunk and Dying light 2 (Full RT on) you name it. Checking the connector several times, It looks pristine as if just out of the box brand new.
@@winnieid2727 is there an issue with toms hardware? After jonny went to corsair, they're the only ones posting in depth reviews of psus. So it'd suck if they were shady.
This is one thing I love about youtube tech creators, for the most part they tend to support each other. You rarely see national media outlets ever collaborating on a investigative story like this.
10:18 Those are some of the jankiest solder joints I've ever seen on any 14g wire. I build 6S quadcopters for fun; I wouldn't even plug in and arm with solder joints that bad -- i'd be worried about letting out the magic smoke. And Nvidia thinks this is OK for sustained 450W+...
Gamers Nexus carried report on validation test failures of adapter cables with 12VHPWR connector from the PCIE5 SIG on the RTX 40 announcement day. Very few of the tech press/reviewers covered it, or advised users about the fragility
I don't. Bunch of know it alls pretending they know better than actual electrical engineers over a couple of freak accidents, likely more to do with the manufacturing process in some Chinese plant.
G'day Paule & Joe, While not being someone who will buy a 4090 it is good to know that there are multiple reviewers looking into what is going on & trying to help solve the problem before 12VHPWR becomes the norm for all GPUs & PSUs for PCIe Power in the future if this is where we are heading.
Can't wait for AMD's marketing of RX 7000. "RX 7900 XT, stronger in rasterise than competition, lower power usage, protected by our special AMD It-Does-Not-Melt technology."
I think you hit the nail right on the head. The broken outer cables are causing intermittent contact between the cable and the pin/pad. This is causing the higher resistance that you describe and depending on how bad the break is and outside conditions could also be causing arcing (There's a literal taser inside the housing) This is why its the OUTSIDE cables that are shown melted in all of the pictures, not the inner ones.
I don't think it's causing arcing, the 12 volts is too low for current to jump through air. It's probably just a lot more heat from more energy throughout through the fewer pins.
@@tristanweide Yep, this video does not explain either why plastic around the terminals are melted, not around the soldering. Buildzoid has also made a video on this subject and so far he was the only one giving a reasonable explanation: it is the Nvidia terminals that must be the cause. I recommend his video.
Nice! JTC also noted that the 30 series connectors were similar, but were individually wired to the pin, as the Corsair cable. Nvidia stepping backward.
14:01 - Thank you. It has been obnoxious seeing people act as if AMD just made this decision right before launch. No, of course AMD didn't just make this decision and force themselves and all their partners to redesign their PCBs. This decision was made well before adapters started melting.
@@johnc8327 As shown by the fact AMD is using PCIe 5.0 and DP 2.1 when Nvidia is using PCIe 4.0 and DP 1.4a. 🙄 I swear, it is incredible how little credit Team Red gets for having common sense.
@@benjaminoechsli1941 So they're using ports and tech that their hardware can't even utilize because they can't even max out the bandwidth of the previous gen tech and are wasting money on new standards and ports and transferring the cost to the customer for no good reason.
This failure is most likely caused by a pin fitment issue. A slightly too loose female pin connected to a male pin will have a slight air gap. At high current draw, this air gap will cause excess resistance which will generate excess heat and you will end up with melted connector. I have seen it before. This failure is most likely attributed to cheaply made connectors based on images shown. The outer pin gets hot and melts first because that is where the slight air gap will be and likely the highest point of resistance of the circuit. This is the simplest and most logical explanation (occam's razor).
Igors adapter looks like I soldered it. Jay's a lot nicer. Yous is somewhere in between. You can tell from all 3, the soldering is inconsistent in quality as well as wire placement. Some have more contact area than others. I think think that 8.3a is too much load for that pin. It is beyond what molex specs for it.
When I saw that, it was clear that these are almost certainly all soldered by hand. That means that soldering wasn't a decision they made to save cost.
Some of the same principles that dictate the use of crimp connections over solder in automotive or aviation industries apply here. If wiring is allowed to flex or move freely then it cannot be terminated by a solid joint (solder) as solder is intended to form an electrical connection only (in most cases), not a physical AND electrical connection.
You are the only one of the youtubers who understands how electricity works. It's precisely because of loose connection, either loose connection in the plug or broken but still rubbing together connection that causes the increase in amperage and the one contact being heated. It's not about totally broken off contact. It is resolved. people just being slow af.
Nice video Paul, I just put in my Corsair cable in this morning for the 4090. The Corsair cable feels nice and robust, and its working great. My old plug I took out and have been using for a week showed no signs of melting or any problems. Keep up the great work.
@@williamrestrepo98 you are sort of right. The one available at best buy is for the 30 series. The pcie5 600w version is offered by Corsair but out of stock at the moment. So yes and no.
absolutely ridiculous! NVIDIA recall the bloody adapters, fix all damaged cards and offer refund, to your customers. The cards are 2.5k in Ireland/UK. So you need a mortgage for feck sake! Normal person would be insane to buy one....
Timely video Paul. I feel for the ones out of a card/setup due to the issues seen here. AMD had the right idea going with ole faithful. 3 8pins for 450 watt is a safe bet. Cheers.
Amazing. Nvidia is going to f around and get the NEC involved in PC internals at this rate. $300+ Billion global company saved 20 cents by including a POS adapter with a $1600 card. Brilliant. But hey, the more video cards you melt, the more you save!
bottom line no built in safety just cost cutting as much as they could... plain and simple too much current and not heavy enough gauge/or robustness of terminals to support it.... its simple physics not rocket science.
I'm not really worried about the four honkin wires supplying the 600W of juice. That's not the failure point. The failure point is the pins themselves being pulled back in the socket and not making good contact. That's where they're heating up and melting the plug.
Great analysis but I don't think a single loose connection with multiple wires in parallel would cause that much heat because the power will go the path of least resistance. It is more likely something is causing a short or multiple pins are becoming lose and causing the additional resistance across multiple pins. You could put an amp clamp on each of the four wires and see if there is an imbalance after multiple plugging and unplugging or various configurations.
exactly ,the net resistance of a parallel circuit is always less than any of the individual resistance values. and base on P=I^2 R , the total heat generated on the cable should still lower then 2 line situation like Jayz experimented , Isn't this high school physics? and if only 2 lines are fine in Jayz test, it's more likely causing by loose pin
In 12 volt systems a 3 foot (1 meter) 14 gauge wire is good for a maximum of 30 amps or 360 watts (2 would handle 720 watts, 3 = 1080 watts). So, this has to be an issue of bad solder connections between the wires and the "foil" or metal plate to which the wires are soldered plus bad connections on the male pin and female side of the connectors. If it was a PSU issue, the PSU should be shutting down due to over current protection. But, it seems the issue is not the PSU shutting down due to the current on individual rails. There has to be more than an issue with the 14 gauge wires and their connection to the foil. It is most likely bad quality control, combined with someone deciding to cut cost and cheaping out on the design - i.e. too thin of a foil plate, 4 wires instead of 6, poor solder connections, tiny easily damage female pin connectors on tiny male pins...etc Seems Nvidia created a bad solution to what was more of a specification problem, rather than an actual power load issue. Again, one 14 gauge wire can handle 360 watts on a meter long wire, 2 can handle over 700 watts and 3 can handle over 1000 watts at 12 volts. Seems someone made some really bad design choices and engineered a bad connector...
Think your analysis about loose contact is the answer to the issue. I troubleshooted my flickering lights once and the root of the issue was loose contact on one of the wires, and that wire/surrounding plastic was a little burnt. In many cases, loose contact is a lot worse than no contact
You are correct in your thoughts of a small/ poor connect causes the heat. Just like a space heater I had. The plug would get so hot that I could not touch it. I replaced the plug and resolved the trouble. Bad connection in plug
We are Teclab and made the tests (yoi called us Galax). We run more than than 25, considering more than 12 minutes overload (rising 300W per step), and reached 127 Celsius. Neither connector or cable melted, and it's still working. Then we installed bad mounting (bad assembly, both horizontal and vertical, then temperature increasing over the time, what can cause damage and melt. About 150 min testing alive. Thanks
I actually sent a request for be quiet! to send me a 12VHPWR adapter as I bought a Straight Power 11 1200W platinum power supply on the same day I bought my 4090. Maybe they'll send one, as I can't find them in stock anywhere.
Thanks for the thorough breakdown. Frankly, it would be great to have some focus on energy efficiency for a change. We can't keep scaling up power to improve performance.
Not sure if anybody else has mentioned this. But a broken connection may also cause arcing to occur. As the current jumps the short gap. This actually can cause an extremely high temperature jump. This is actually how many house fires are caused on mains power connections where there is a poor connection. So Paul's conclusion on the poor connection is probably correct but, it might also be worsened by internal arcing adding extra heat.
I ran the numbers just as a sanity check for the possibility of a bad connection causing the problem. I believe power loss will be "high" when the impedance of the poor connection matches the impedance of the wires coming from the PSU to the connector. Doing some back of the envelope calculations, the expected power loss for 4 - 14 AWG Aluminum, 0.5 meter wires carrying 50A total is about 4W. (using 13.47 mOhm/meter at 70C). That's higher than I expected, but distributed over that much wire, it's fine. I ran through the current divider circuit, and found that the weakened connection could dissipate a little over 1 watt in the worst case scenario. This is obviously in a much more concentrated and insulated area than before. My intuition tells me that could be enough power to start a bad thermal situation. Resistors rated for 1 watt are pretty chonky as is, and they would probably be significantly derated being incased in plastic. Maybe the Sense #2 and #3 should have been used for a thermocouple in the connector ¯\_(ツ)_/¯
I think you have some terms backwards, where do you find large resistors rated for 1 watt? I think you meant 1 ohm, because at 12v that would be 12 amps through a resistor. A 1 watt resistor at 12v would be like 0.08 amps.
Never had a warm fuzzy about that adaptor design. When the cards had three 8 pin connectors it was always recommended to use three separate 8pin wires from the PSU.
The issue seem boil down to that the wires are simply soldered into place (instead of being crimped down). This can cause arcing/sparks inside the plastic housing (and in turn, melt the plastic) if they come lose. Users simply bending the cable from side to side a little can be enough to cause the wires to break lose (which can easily happen when people try and do some cable management). nVidia will need to recall these cheaply made adapters, as they are not safe to use. That native Corsair-cable is complete safe to use as the wires are crimped down, and not soldered.
If it's ABS plastic then the glass temp is 105c, but it will have started to soften a bit before that. Based on my own experience of 3D printing with ABS.
check out "rambling about the 12VHPWR failures" by Actually Hardcore Overclocking. he makes a very good case for why the cable is failing. They used double split terminal pins instead of single sided like corsair and cablemods. So any bending or side by side strain can make poor contact in the pins which then increases resistance in the pins way higher which then increases heat then thermal run away.
The fact that these cables failed in the way they did does not suggest that they were cheap. Far from it. It would have been much cheaper to use something closer to what the 3090 used, even adding the sense pins. The adapter is too clever by half, and too clever for its own good.
People could also be connecting extra 8-pin power cables from a different power supply than the one in their PC. Not everyone knows that mixing power cables is the fastest way to fry ports/cards. But if the person uses the wrong power cable, the failure point will be the 12-pin connector even though it'd actually be blameless. I know when I got my 4090, initially I only had two 8-pin cables coming from my PSU to my 3080. So I had to go find more 8-pin cables. I bet others did the same thing and some grabbed the wrong cables.
burning issues happened when they used a 2x8pin to 12+4 pin adapter. to avoid this, use a real PCIe gen5 PSUs with dedicated 12vhpwr pins. once you smell the smoke, it means that's the smell of a lawsuit. gpus are not the problem, rather it's on the psus. Reply
I bet if you asked Dave Jones of the EEVBlog, he'd have a lot of problems with this adapter. He'd probably tell you that this connection should be crimped, not soldered. In fact, a lot of electronics coming out of China have a soldered third pin ground connection. He'll call that out because for code reasons, that should be crimped. A crimped connection provides an interface from solid to stranded metal that will not fatigue and crack with stress or age. A solder connection might. For low-current signal wires it's all right, but at the point you're talking about 50 F**KING AMPS OF CURRENT, you want a heavy cable and a more solid connector. If you've got a gasoline-powered car, think about the battery cables for a moment. They're meant for six to eight times the current capacity (for a few seconds) that this ATX 3.0 12+4 connector is meant to carry. But think about the thickness of the cable and the fact that there's a positive and negative and they're both that massive. And they're "flexible" stranded cables that are crimped down to big solid slugs of metal that are heavily bolted to similar slugs of metal in the battery. We need to not screw around with the quality of this connector, it's providing up to 50 MOTHERF**KING AMPS to your very delicate $1600 rebadged cryptomining device.
For gods sake, please give me just two fat, simple power wires (+/-12V) I can screw down on the backside of GPU. No cable mess, no issue with connectors. 19th century tech that's miles ahead of this 21st century failure of electric engineering and cable management. For f. 2000 EUR can I have as much!? Or a solid copper connector attached to the end of such cable. You can feed the whole PC with 2 dead simple sticks put into your power outlet, but GPU needs all this cable and connector crap?!
was it only the 4090 that had an issue? or all the cards that used the 12vhpr? I've been looking at a 4070 ti super and I'm concerned because I've been news that it's STILL an issue.
How it can be that #igorsLAB have only 70k subscribers!? 🤷♂️🤦♂️ Come on guys, hit the button! 🙏😘 This is one of the most competent people when it comes to electronics and testing them. Believe me.
Nvidia Engineers: "This new $1600.00 card is going to draw unprecedented amounts of power." Nvidia Bean Counters: "Then let's cheap out on the power connectors."
Igor's conclusion was already debunked by Buildzoid, LUL ruclips.net/video/yvSetyi9vj8/видео.html Just think about it logically - the solder joints are on opposite part of connector where there is no melted plastic. That doesn't mean Igor is wrong about that part of connector being shit quality - I'm sure it is - but those posted melted connectors are not melting anywhere close to solder joints. It's weak receptacle contact on pins, very likely because nvidia is not using OEM receptacle terminals.
As it states int he datasheet of that connector, it's durability is 30 mating cycles.. .after that you're supposed to toss it. And reading the datasheet, the connector is used at it's limits by the 4090, meaning that one has to be absolutely sure that the connector is properly seated, which means very clean contact surfaces (no dust, no grease) and that it's been mated straight and pushed all the way in and is not subjected to ANY strain what so ever. No pulling, no torque to the side (by bending the cabling) nor anything as such. That connector is not suitable for the application and for consumers grubby hands and people building their own computers. It's meant for special applications where you need large amount of current over very small wires and in tight space PLUS when you need control signals on top of that. Using that type on consumer GPU and pushing it's limits in current handling is really a recipe for destruction. nVidia is either stupid, or didn't ever mean a consumer to buy the 4090 card and it was meant for servers and high end system integrator where in highly clean environments, with very strict quality control, professionals install them ONCE to a system that's never touched again. that's not a consumer environment.
I think it's the factory mix up on NVIDIA 12V PWR Adapter. There are some NVIDIA adapters out (Like Igor's one) that have each wire rated 150V but majority of us including me have each wire rated 300V. You can easily find out if you unpeel bending tape and expose little of the wire. Again I don't think you can recreate the melting if you have each wire rated 300V.
O look. They couldn't even finish their R&D for $1600 graphics cards. Will the people buying these things ever feel stupid? I doubt it. But we can dream
Nice to see you doing a bit more research than most of the "influencers" who jumped on the doom and gloom bandwagon and overblew the situation with fear mongering. To me this looks a lot more like some faulty adapters just like every generation of power connector has had forever but was especially more common in the early days of each plug. While 600W through a tiny plug like that seems like a lot to people even a 16 awg wire in a x4 config should be able to handle more than 600W this is almost certainly a faulty batch(or partial batch) because of poor soldering rather than an actual issue with the plugs and wires.
It all comes down to money and Nvidia being cheap. I want to know who made these cables, like the company Nvidia contracted to actually manufacture them in a factory made them. If Nvidia wasn't trying to produce the cable as cheap as possible and actually cared about quality this would not be happening. Edited to add that Nvidia wants all "defective"/damaged cards sent directly to HQ so they can prove/show that nothing was actually wrong with the card and sue the pants off whoever they contracted to manufacture these adapters.
Hardware Busters International's video, "16-Pin (12VHPWR) Cable Bended vs Straight. Is it dangerous on an RTX 4090?", is well worth a watch too. His power monitoring device is top notch.
Why do we still use all these screwball multi wire connectors? 99.9% of cases out there are made of conductive material. Thus go with a chassis ground. Instead of multiple smaller gauge wires, how about one large gage wire for the positive side? Cars have been successfully running their electrical systems in this fashion for over a century. Proven concept for 12 volt systems.
There are at least two different variations of this adaptor, 150v and 300v. the 300v like you have is a lot more heavy duty. Id like to see the burnt adapters, I bet they are all 150v.
This is all Goobidy Gop to me.. I would never buy that card anyway.. If it needs more that 2/3 8 pins I don't even want to use it no matter how fast it is.. Fuc that..
I think it's safe to say for me even if the new AMD cards have 10% less performance and I will buy day 1. I cannot believe Nvidia released these cards with this garbage cable
You don't watch jays vids for testing, he don't have much technical knowledge and more of entertainment side of things. Free to disagree, that's what I got from following him for years.
In residential wiring, 12 gauge wire is used for 20 amp circuits and 14 gauge wire (what the connector uses) is limited to a15 amp load (@120v). So the first problem is that if one of the four wires comes loose you are carrying 50 amps divided by 3 or 16.67 amps on a wire rated for 15 amps. However, what makes it worse is in residential wiring, the wire should only draw 80% of its maximum wattage in a continuous use situation. So in a 12v scenario, one of the four wires in a continuous use situation should only be supplying (15amps x 12 volts x .8) 144 watts. Even if all four wires are intact, at 600 watts, each wire would be supplying 150 watts. That is 4% over the 80% rule but probably okay. However, reduce that to three wires and you get 200 watts. That is 39% more than the wires recommended continuous loading. As Igor might say, "Nicht gut."
So there's different standards for different uses. I did some looking at automotive 12v standards, and it looks like with 6 wires, they'd need to be 16 or 18ga. So 4x 14ga is probably fine. The connector also has a rating of course, and it looks like the connector is the problem, not the wires themselves.
@@tippyc2 My analysis was based on solid copper wiring (i.e., Romex). Typically, stranded aluminum wiring needs to one size larger for the same rating. So those 14 gauge wires are really equivalent to 16 gauge copper. Put another way, for the ratings I mentioned the four wires should have been 12 gauge to meet a 14 gauge copper spec. It could be that electronics have a different spec, but since it's used in a residential setting I'd think safety standard would be similar. I don't have enough information to definitively state the root cause; just wanted to point out that there wasn't much margin for error in what they used. I will say that the weak link in most electric components that I've seen is the wire and connectors used going into an assembly and not the supply wires. Just look at a typical light fixture. I hope the industrial engineering in the interconnects improves moving forward to keep up with this trend in power usage.
I have to agree with Buildzoid on this one. Where Nvidia's adapter has female connectors that have 2 splits in them, other manufacturers are using female plugs with only 1 split (which is mechanically and functionally much more rigid), So Nvidia's female adapters have much less rigidity especially when bending horizontally. Once bent this would cause the same amount of amperage to transfer through a smaller connection point (from the adapter to the GPU) and in turn create more heat. am not a electrical engineer, but to me this would make more sense as to why the connectors seem to be melting at the pin connections rather than melting closer to where the solder point is on the Nvidia adapter.
I suspect the melting is due to heating of the pin caused by high current arcs, itself due to embrittlement (phase changing) of the solder due to heat cycling. The embrittlement increases resistance at the connection point, lowering the voltage and therefore increasing the current (for a power controlled system) increasing heat. Eventually, the connection is so poor we see arcing which results is large temperature spikes. In one of the failure images you can see arcing tracers on the female end. I suspect that the specification requires these connections to be crimped and not soldered (like the Corsair cable) so I'm surprised this went through to manufacturing and passed QA. Bad day for the Manufacturing Engineer who signed off on this.
I believe it is just a manufacturing error. Or a combination of several Issues happening at once. Looking at Igor's adapter and at yours, you can see a big difference in soldering quality and how much of the cables actually touches the pins. I think adapters that were produced first were good quality, but adapters made at a later time have poorer quality. It may also be a case of cutting corners of the supplier, because they can't make enough in time. A possible reason why that is happening is missing workers due to covid... Im just guessing of course, but it is not like it hasn't happened before elsewhere. And 4090 FE's are not "burnung up" yet. It might be, that Nvidia simply uses those good "early bird" adapters and AIBs recieve the later bad "revisions" unintentionally.
That doesn't seem like sufficient isolation between +12V and ground inside the connector. I bet those two skinny little plates are somehow shorting a little bit inside the connector and going nuclear from there.
Does no one else think it is crazy that the gaming industry just keeps increasing its power draw when the rest of the world is trying to save energy? 600W is nuts. That's the same as a small room heater.
Paul, this launch is the dumbest launch I've ever seen in technology. There is no stock, it doesn't fit in anyone's computer, and it literally bursts into flames when you use it.
Not for people with experience... I would never use an adapter on a 2000 dollar GPU as its an extra point of possible failure.. Get a 20 dollar dedicated cable and problem solved. It has to do with transitioning to new PSU and connector designs while trying to make current PSU designs work with the new.
Nvidia heard that new gen tech consumers are nicknamed "1st wave Beta testers" then went the extra mile and added the "final QA" title just for laughs.
when I first saw the pics, I thought it would be a bad solder joint in the connector. I've ran into a similar issue once with one of the old molex connectors back in the day where it wasn't crimped correctly and it melted the connector.
I think they populated the two unused connections to prevent people from sending angry social media posts about "missing wires" in their connector 😂 Great series of videos from the community, hopefully it helps with the resolution coming faster and better!
UPDATE: In light of the excellent investigative video posted by Steve / GamersNexus (ruclips.net/video/EIKjZ1djp8c/видео.html), it would appear that this problem is still unresolved and there's more to it than the 14AWG wire solder points in the plug. Steve is asking for more people with these adapters to get in touch with him, and of course NVIDIA is investigating although they haven't made any statements beyond that. Locating more adapters with the 150V rated wires that IgorsLab saw (mine used 300V wires which can be seen at 8:26) would definitely help. Most are still leaning towards this being a manufacturing issue with the adapters, but what specifically the defect is remains to be seen.
You are just pathetic.
Seems there are different versions of the adapter. @10:14 yours looks exactly like Igor's version and the outer cables soldered separately and not bridging with the adjacent cables. While GN's ones are different and seem more sturdy.
@@kamui004 yeah nvidia definitely spotted the issue, redid the cable design, but the 1st ones were already shipped (with 150v wires) and they should have recalled them
Paul, honestly speaking, has it ever occurred to anyone that the damages could be produced by shills with a flame and posted as "it happened naturally without me doing anything wrong!"? Has either you, Steve or Jay ever seen the problem firsthand? Ever?
And no the story of first ones being faulty doesn`t stand. Both GamerNexus and JayZ have review samples sent in advance and definitely belonging to the first batch.
Have they (Jay or Steeve) been able to reproduce the problem even after torturing the connector in the worst possible ways? And if not, is anyone starting to smell the much ado-about-nothing stink? (shills melting the connector with a flame anyone).
P.S.
I have been torturing my 4090 (MSI Suprim X purchased day 1) for days now, overclocking, 6 hours of consecutive gameplay at Cyberpunk and Dying light 2 (Full RT on) you name it. Checking the connector several times, It looks pristine as if just out of the box brand new.
send Steve your cables today.
The melting plastic is a feature. It fuses you to nvidia's ecosystem indefinitely..the smell is just a notification of successful synergy.
Thanks for the best laugh of my day. :-)
I thought that was gsync...
@@edwardallenthree LOL!
Nvidia doesn't have an ecosystem
Hilarious truth
I don't think it was a contest, but if it was, you definitely win the "Cleanest adapter deconstruction award"
Very nicely done. That's not easy.
Agreed! Paul's looked like something that was intentionally made for educational purposes or something.
Igor really is a gem, I always watch his videos.
The entire industry has attacked him, basically _always wrongly,_ on _several_ occasions now. That's how you _know_ he's an actual threat to them.
RUclips's auto translation leaves me vexed, so I follow the link to his articles. I like the way the man thinks.
He is the original TH when it was the best out there, glad he break away early and keep the high standards even now.
@@winnieid2727 is there an issue with toms hardware? After jonny went to corsair, they're the only ones posting in depth reviews of psus. So it'd suck if they were shady.
@@Tagiau Here on RUclips, look for Hardware Busters International. The man does very thorough PSU reviews.
This is one thing I love about youtube tech creators, for the most part they tend to support each other. You rarely see national media outlets ever collaborating on a investigative story like this.
Thank you for investigating problems like these. It really helps push new tech forward. Watch out here comes sherlock Paul!
10:18 Those are some of the jankiest solder joints I've ever seen on any 14g wire. I build 6S quadcopters for fun; I wouldn't even plug in and arm with solder joints that bad -- i'd be worried about letting out the magic smoke. And Nvidia thinks this is OK for sustained 450W+...
I appreciate the entire community working together to bring us better data and information regarding this.
Gamers Nexus carried report on validation test failures of adapter cables with 12VHPWR connector from the PCIE5 SIG on the RTX 40 announcement day.
Very few of the tech press/reviewers covered it, or advised users about the fragility
I don't. Bunch of know it alls pretending they know better than actual electrical engineers over a couple of freak accidents, likely more to do with the manufacturing process in some Chinese plant.
RUclips you work together here ???
who does that ??? you share likes and commands ? subscribers ???
G'day Paule & Joe,
While not being someone who will buy a 4090 it is good to know that there are multiple reviewers looking into what is going on & trying to help solve the problem before 12VHPWR becomes the norm for all GPUs & PSUs for PCIe Power in the future if this is where we are heading.
How many times are you going to reiterate not being interested in the 4090 to "Paule & Joe" every time there's a 4090 video upload?
Thanks Nvidia for using the cheapest connections you can get while charging us 1700 bucks for the card.
Funny when it is likely just customers not plugging in the cable enough. KEKW
Na, cable works fine. User error
Can't wait for AMD's marketing of RX 7000. "RX 7900 XT, stronger in rasterise than competition, lower power usage, protected by our special AMD It-Does-Not-Melt technology."
I think you hit the nail right on the head. The broken outer cables are causing intermittent contact between the cable and the pin/pad. This is causing the higher resistance that you describe and depending on how bad the break is and outside conditions could also be causing arcing (There's a literal taser inside the housing) This is why its the OUTSIDE cables that are shown melted in all of the pictures, not the inner ones.
I don't think it's causing arcing, the 12 volts is too low for current to jump through air. It's probably just a lot more heat from more energy throughout through the fewer pins.
@@tristanweide I mean it could… a car battery connector arcs when you connect it, and that’s ~13 volts
@@tristanweide
Yep, this video does not explain either why plastic around the terminals are melted, not around the soldering. Buildzoid has also made a video on this subject and so far he was the only one giving a reasonable explanation: it is the Nvidia terminals that must be the cause. I recommend his video.
Na, cable works fine. User error
The best thing is... there really was no need for this new connector at all. The old connectors would work fine.
Nice PSA, Nice to see Jay, Steve and yourself collaborate on such issues and support each other doing it.
Nice! JTC also noted that the 30 series connectors were similar, but were individually wired to the pin, as the Corsair cable.
Nvidia stepping backward.
Money over everything. Why pay for 12 wires when you could pay for 8?
Ah yes, the Apple manouver
14:01 - Thank you. It has been obnoxious seeing people act as if AMD just made this decision right before launch. No, of course AMD didn't just make this decision and force themselves and all their partners to redesign their PCBs. This decision was made well before adapters started melting.
AMD doesn't do anything until Nvidia has already been doing it for 5+ years
@@johnc8327 As shown by the fact AMD is using PCIe 5.0 and DP 2.1 when Nvidia is using PCIe 4.0 and DP 1.4a. 🙄
I swear, it is incredible how little credit Team Red gets for having common sense.
@@benjaminoechsli1941 So they're using ports and tech that their hardware can't even utilize because they can't even max out the bandwidth of the previous gen tech and are wasting money on new standards and ports and transferring the cost to the customer for no good reason.
They will be able use the bandwidth of display 2.0. You are free to give your money to whoever you wish.
@@Talishar Sad....
Took out my 4090. Not using it until Nvidia sends me a new cable. And no, I'm not paying 60 EUR to Cablemod just because NVIDIA are idiots.
This failure is most likely caused by a pin fitment issue. A slightly too loose female pin connected to a male pin will have a slight air gap. At high current draw, this air gap will cause excess resistance which will generate excess heat and you will end up with melted connector. I have seen it before. This failure is most likely attributed to cheaply made connectors based on images shown. The outer pin gets hot and melts first because that is where the slight air gap will be and likely the highest point of resistance of the circuit. This is the simplest and most logical explanation (occam's razor).
Yes, you are right. I recommend Buildzoid's video, it confirms your opinion.
Igors adapter looks like I soldered it. Jay's a lot nicer. Yous is somewhere in between. You can tell from all 3, the soldering is inconsistent in quality as well as wire placement. Some have more contact area than others.
I think think that 8.3a is too much load for that pin. It is beyond what molex specs for it.
When I saw that, it was clear that these are almost certainly all soldered by hand. That means that soldering wasn't a decision they made to save cost.
Paul: I can confirm Igor's findings.
That moment when you realize that Paul And Igor in fact does NOT have the same adapter design. 😂
Some of the same principles that dictate the use of crimp connections over solder in automotive or aviation industries apply here. If wiring is allowed to flex or move freely then it cannot be terminated by a solid joint (solder) as solder is intended to form an electrical connection only (in most cases), not a physical AND electrical connection.
You are the only one of the youtubers who understands how electricity works. It's precisely because of loose connection, either loose connection in the plug or broken but still rubbing together connection that causes the increase in amperage and the one contact being heated. It's not about totally broken off contact. It is resolved. people just being slow af.
Wrong resistance does not increase current, it decreases it, but we all know ur real smart
Nice video Paul, I just put in my Corsair cable in this morning for the 4090. The Corsair cable feels nice and robust, and its working great. My old plug I took out and have been using for a week showed no signs of melting or any problems. Keep up the great work.
Where do you find it? I am having no luck in finding one, about as much trouble as I had for the dang 4090 card
@@killer01ws6 best buy has the Corsair 12 pin cables for $20.
Have you noticed if you can set the maximum power over 93% on the geforce experience performance tab?
@@Stormborn29 The cable you are thinking of is only for 3090FE being a 12 pin not a 12+6 which is the one needed for 3090ti/4090 cards.
@@williamrestrepo98 you are sort of right. The one available at best buy is for the 30 series. The pcie5 600w version is offered by Corsair but out of stock at the moment. So yes and no.
Great video. Personally I'd love to get the Corsair VHPWR cable, but those things are more scarce than actual 4090s at this point.
For real... its wild. I would love if nvidia would just send out these cords. Would be great...
CableMod has their version in stock for about $20USD. I think they're much better equipped to crank these cables out.
I'm very glad you mentioned the possibility of broken but still (poorly) contacting outer connections, thank you.
absolutely ridiculous! NVIDIA recall the bloody adapters, fix all damaged cards and offer refund, to your customers. The cards are 2.5k in Ireland/UK. So you need a mortgage for feck sake! Normal person would be insane to buy one....
Let's be real... Anyone who bought a 4090 deserves a fire.
Timely video Paul. I feel for the ones out of a card/setup due to the issues seen here. AMD had the right idea going with ole faithful. 3 8pins for 450 watt is a safe bet. Cheers.
Amazing. Nvidia is going to f around and get the NEC involved in PC internals at this rate. $300+ Billion global company saved 20 cents by including a POS adapter with a $1600 card. Brilliant.
But hey, the more video cards you melt, the more you save!
bottom line no built in safety just cost cutting as much as they could... plain and simple too much current and not heavy enough gauge/or robustness of terminals to support it.... its simple physics not rocket science.
I'm not really worried about the four honkin wires supplying the 600W of juice. That's not the failure point. The failure point is the pins themselves being pulled back in the socket and not making good contact. That's where they're heating up and melting the plug.
Great analysis but I don't think a single loose connection with multiple wires in parallel would cause that much heat because the power will go the path of least resistance. It is more likely something is causing a short or multiple pins are becoming lose and causing the additional resistance across multiple pins. You could put an amp clamp on each of the four wires and see if there is an imbalance after multiple plugging and unplugging or various configurations.
exactly ,the net resistance of a parallel circuit is always less than any of the individual resistance values. and base on P=I^2 R , the total heat generated on the cable should still lower then 2 line situation like Jayz experimented , Isn't this high school physics?
and if only 2 lines are fine in Jayz test, it's more likely causing by loose pin
In 12 volt systems a 3 foot (1 meter) 14 gauge wire is good for a maximum of 30 amps or 360 watts (2 would handle 720 watts, 3 = 1080 watts). So, this has to be an issue of bad solder connections between the wires and the "foil" or metal plate to which the wires are soldered plus bad connections on the male pin and female side of the connectors. If it was a PSU issue, the PSU should be shutting down due to over current protection. But, it seems the issue is not the PSU shutting down due to the current on individual rails. There has to be more than an issue with the 14 gauge wires and their connection to the foil. It is most likely bad quality control, combined with someone deciding to cut cost and cheaping out on the design - i.e. too thin of a foil plate, 4 wires instead of 6, poor solder connections, tiny easily damage female pin connectors on tiny male pins...etc Seems Nvidia created a bad solution to what was more of a specification problem, rather than an actual power load issue. Again, one 14 gauge wire can handle 360 watts on a meter long wire, 2 can handle over 700 watts and 3 can handle over 1000 watts at 12 volts. Seems someone made some really bad design choices and engineered a bad connector...
Think your analysis about loose contact is the answer to the issue. I troubleshooted my flickering lights once and the root of the issue was loose contact on one of the wires, and that wire/surrounding plastic was a little burnt. In many cases, loose contact is a lot worse than no contact
"My connector is horribly worn mother, I need a new one"
Yes dear. Why does it seem to happen every two years?
The WAY Better question would be "Why hundreds of thousands are NOT melting while those 12 on Reddit did?"...
those users probably got defective adapters cables or are using low quality 3rd party cables
@@ssaini5028 Na, they just didn't know what they were doing and never fully seated their power plugs. Defective users
@@kenmasters916 agreed lol
You are correct in your thoughts of a small/ poor connect causes the heat. Just like a space heater I had. The plug would get so hot that I could not touch it. I replaced the plug and resolved the trouble. Bad connection in plug
We are Teclab and made the tests (yoi called us Galax).
We run more than than 25, considering more than 12 minutes overload (rising 300W per step), and reached 127 Celsius.
Neither connector or cable melted, and it's still working.
Then we installed bad mounting (bad assembly, both horizontal and vertical, then temperature increasing over the time, what can cause damage and melt.
About 150 min testing alive.
Thanks
I actually sent a request for be quiet! to send me a 12VHPWR adapter as I bought a Straight Power 11 1200W platinum power supply on the same day I bought my 4090. Maybe they'll send one, as I can't find them in stock anywhere.
So to summarize... The new 12VHPWR connection is great but the bundled Nvidia adapters are trash! Who would have thought...
Thanks for the thorough breakdown. Frankly, it would be great to have some focus on energy efficiency for a change. We can't keep scaling up power to improve performance.
Not sure if anybody else has mentioned this. But a broken connection may also cause arcing to occur. As the current jumps the short gap. This actually can cause an extremely high temperature jump. This is actually how many house fires are caused on mains power connections where there is a poor connection. So Paul's conclusion on the poor connection is probably correct but, it might also be worsened by internal arcing adding extra heat.
I ran the numbers just as a sanity check for the possibility of a bad connection causing the problem.
I believe power loss will be "high" when the impedance of the poor connection matches the impedance of the wires coming from the PSU to the connector. Doing some back of the envelope calculations, the expected power loss for 4 - 14 AWG Aluminum, 0.5 meter wires carrying 50A total is about 4W. (using 13.47 mOhm/meter at 70C). That's higher than I expected, but distributed over that much wire, it's fine.
I ran through the current divider circuit, and found that the weakened connection could dissipate a little over 1 watt in the worst case scenario. This is obviously in a much more concentrated and insulated area than before. My intuition tells me that could be enough power to start a bad thermal situation. Resistors rated for 1 watt are pretty chonky as is, and they would probably be significantly derated being incased in plastic.
Maybe the Sense #2 and #3 should have been used for a thermocouple in the connector ¯\_(ツ)_/¯
I think you have some terms backwards, where do you find large resistors rated for 1 watt? I think you meant 1 ohm, because at 12v that would be 12 amps through a resistor. A 1 watt resistor at 12v would be like 0.08 amps.
Never had a warm fuzzy about that adaptor design. When the cards had three 8 pin connectors it was always recommended to use three separate 8pin wires from the PSU.
the problem is the mechanical design, not the amount of wires ffs
The issue seem boil down to that the wires are simply soldered into place (instead of being crimped down).
This can cause arcing/sparks inside the plastic housing (and in turn, melt the plastic) if they come lose.
Users simply bending the cable from side to side a little can be enough to cause the wires to break lose
(which can easily happen when people try and do some cable management).
nVidia will need to recall these cheaply made adapters, as they are not safe to use.
That native Corsair-cable is complete safe to use as the wires are crimped down, and not soldered.
nVidia can't ignore the deficiencies in the cable design. Possible FIRE hazards are too real. THX Paul.
(sarcasm) Melted wires and plastic, what every computer owner loves to smell after spending nearly $2,000.00 for their new GPU.
I'm glad you pointed out it was sarcasm
We need a sarcastica font.
@@Sorrocasharked Me Too!
This is such a non-issue. Some faulty things and everyone is blowing it way out of proportion.
Imagine spending that much money to receive such a janky and potentially life threatening product.
If it's ABS plastic then the glass temp is 105c, but it will have started to soften a bit before that. Based on my own experience of 3D printing with ABS.
It's a relief for the 5.0 connector, but a damn shame for Nvidia. They should be sending out a revised version of the adapter.
check out "rambling about the 12VHPWR failures" by Actually Hardcore Overclocking. he makes a very good case for why the cable is failing. They used double split terminal pins instead of single sided like corsair and cablemods. So any bending or side by side strain can make poor contact in the pins which then increases resistance in the pins way higher which then increases heat then thermal run away.
If they skimped on a simple thing like a cable, you have to wonder, what else did they skimp on?
The fact that these cables failed in the way they did does not suggest that they were cheap. Far from it. It would have been much cheaper to use something closer to what the 3090 used, even adding the sense pins. The adapter is too clever by half, and too clever for its own good.
This drama is the perfect thing for these youtube personalities to fluff their view counts with in between the Nvidia and AMD GPU launches.
I am no electrician but my car aux lights that draw 100W have much thicker copper wire than 500W GPU power.
People could also be connecting extra 8-pin power cables from a different power supply than the one in their PC. Not everyone knows that mixing power cables is the fastest way to fry ports/cards. But if the person uses the wrong power cable, the failure point will be the 12-pin connector even though it'd actually be blameless.
I know when I got my 4090, initially I only had two 8-pin cables coming from my PSU to my 3080. So I had to go find more 8-pin cables. I bet others did the same thing and some grabbed the wrong cables.
burning issues happened when they used a 2x8pin to 12+4 pin adapter. to avoid this, use a real PCIe gen5 PSUs with dedicated 12vhpwr pins. once you smell the smoke, it means that's the smell of a lawsuit. gpus are not the problem, rather it's on the psus.
Reply
I bet if you asked Dave Jones of the EEVBlog, he'd have a lot of problems with this adapter. He'd probably tell you that this connection should be crimped, not soldered. In fact, a lot of electronics coming out of China have a soldered third pin ground connection. He'll call that out because for code reasons, that should be crimped. A crimped connection provides an interface from solid to stranded metal that will not fatigue and crack with stress or age. A solder connection might. For low-current signal wires it's all right, but at the point you're talking about 50 F**KING AMPS OF CURRENT, you want a heavy cable and a more solid connector.
If you've got a gasoline-powered car, think about the battery cables for a moment. They're meant for six to eight times the current capacity (for a few seconds) that this ATX 3.0 12+4 connector is meant to carry. But think about the thickness of the cable and the fact that there's a positive and negative and they're both that massive. And they're "flexible" stranded cables that are crimped down to big solid slugs of metal that are heavily bolted to similar slugs of metal in the battery.
We need to not screw around with the quality of this connector, it's providing up to 50 MOTHERF**KING AMPS to your very delicate $1600 rebadged cryptomining device.
I think Nvidia should be employing Igor as they don't seem to have anyone with that level of knowledge, if they did this wouldn't be happening
For gods sake, please give me just two fat, simple power wires (+/-12V) I can screw down on the backside of GPU. No cable mess, no issue with connectors. 19th century tech that's miles ahead of this 21st century failure of electric engineering and cable management. For f. 2000 EUR can I have as much!? Or a solid copper connector attached to the end of such cable. You can feed the whole PC with 2 dead simple sticks put into your power outlet, but GPU needs all this cable and connector crap?!
was it only the 4090 that had an issue? or all the cards that used the 12vhpr? I've been looking at a 4070 ti super and I'm concerned because I've been news that it's STILL an issue.
How it can be that #igorsLAB have only 70k subscribers!? 🤷♂️🤦♂️
Come on guys, hit the button! 🙏😘
This is one of the most competent people when it comes to electronics and testing them. Believe me.
Nvidia Engineers: "This new $1600.00 card is going to draw unprecedented amounts of power."
Nvidia Bean Counters: "Then let's cheap out on the power connectors."
Igor's conclusion was already debunked by Buildzoid, LUL ruclips.net/video/yvSetyi9vj8/видео.html
Just think about it logically - the solder joints are on opposite part of connector where there is no melted plastic. That doesn't mean Igor is wrong about that part of connector being shit quality - I'm sure it is - but those posted melted connectors are not melting anywhere close to solder joints. It's weak receptacle contact on pins, very likely because nvidia is not using OEM receptacle terminals.
As it states int he datasheet of that connector, it's durability is 30 mating cycles.. .after that you're supposed to toss it. And reading the datasheet, the connector is used at it's limits by the 4090, meaning that one has to be absolutely sure that the connector is properly seated, which means very clean contact surfaces (no dust, no grease) and that it's been mated straight and pushed all the way in and is not subjected to ANY strain what so ever. No pulling, no torque to the side (by bending the cabling) nor anything as such.
That connector is not suitable for the application and for consumers grubby hands and people building their own computers. It's meant for special applications where you need large amount of current over very small wires and in tight space PLUS when you need control signals on top of that.
Using that type on consumer GPU and pushing it's limits in current handling is really a recipe for destruction.
nVidia is either stupid, or didn't ever mean a consumer to buy the 4090 card and it was meant for servers and high end system integrator where in highly clean environments, with very strict quality control, professionals install them ONCE to a system that's never touched again.
that's not a consumer environment.
I think it's the factory mix up on NVIDIA 12V PWR Adapter. There are some NVIDIA adapters out (Like Igor's one) that have each wire rated 150V but majority of us including me have each wire rated 300V. You can easily find out if you unpeel bending tape and expose little of the wire. Again I don't think you can recreate the melting if you have each wire rated 300V.
O look. They couldn't even finish their R&D for $1600 graphics cards. Will the people buying these things ever feel stupid? I doubt it. But we can dream
Well I’m a lifelong gamer and I definitely won’t ever be wasting $1600 USD on a card. Fuck that, I’m happy with my 1660Ti atm.
Nice to see you doing a bit more research than most of the "influencers" who jumped on the doom and gloom bandwagon and overblew the situation with fear mongering.
To me this looks a lot more like some faulty adapters just like every generation of power connector has had forever but was especially more common in the early days of each plug.
While 600W through a tiny plug like that seems like a lot to people even a 16 awg wire in a x4 config should be able to handle more than 600W this is almost certainly a faulty batch(or partial batch) because of poor soldering rather than an actual issue with the plugs and wires.
It all comes down to money and Nvidia being cheap. I want to know who made these cables, like the company Nvidia contracted to actually manufacture them in a factory made them. If Nvidia wasn't trying to produce the cable as cheap as possible and actually cared about quality this would not be happening.
Edited to add that Nvidia wants all "defective"/damaged cards sent directly to HQ so they can prove/show that nothing was actually wrong with the card and sue the pants off whoever they contracted to manufacture these adapters.
Hardware Busters International's video, "16-Pin (12VHPWR) Cable Bended vs Straight. Is it dangerous on an RTX 4090?", is well worth a watch too. His power monitoring device is top notch.
Why do we still use all these screwball multi wire connectors? 99.9% of cases out there are made of conductive material. Thus go with a chassis ground. Instead of multiple smaller gauge wires, how about one large gage wire for the positive side? Cars have been successfully running their electrical systems in this fashion for over a century. Proven concept for 12 volt systems.
Jay is a schill. I'm so upset at his "you are dumb if you don't go out and buy a 3900 right now" video. I've unsubscribed from him twice now.
There are at least two different variations of this adaptor, 150v and 300v. the 300v like you have is a lot more heavy duty. Id like to see the burnt adapters, I bet they are all 150v.
This is all Goobidy Gop to me.. I would never buy that card anyway.. If it needs more that 2/3 8 pins I don't even want to use it no matter how fast it is.. Fuc that..
I think it's safe to say for me even if the new AMD cards have 10% less performance and I will buy day 1. I cannot believe Nvidia released these cards with this garbage cable
You don't watch jays vids for testing, he don't have much technical knowledge and more of entertainment side of things. Free to disagree, that's what I got from following him for years.
Thanks Paul. You made quick work of figuring this out and presenting it in a way that makes sense.
I mean, Igor figured it out, not Paul
In residential wiring, 12 gauge wire is used for 20 amp circuits and 14 gauge wire (what the connector uses) is limited to a15 amp load (@120v). So the first problem is that if one of the four wires comes loose you are carrying 50 amps divided by 3 or 16.67 amps on a wire rated for 15 amps. However, what makes it worse is in residential wiring, the wire should only draw 80% of its maximum wattage in a continuous use situation. So in a 12v scenario, one of the four wires in a continuous use situation should only be supplying (15amps x 12 volts x .8) 144 watts. Even if all four wires are intact, at 600 watts, each wire would be supplying 150 watts. That is 4% over the 80% rule but probably okay. However, reduce that to three wires and you get 200 watts. That is 39% more than the wires recommended continuous loading. As Igor might say, "Nicht gut."
So there's different standards for different uses. I did some looking at automotive 12v standards, and it looks like with 6 wires, they'd need to be 16 or 18ga. So 4x 14ga is probably fine. The connector also has a rating of course, and it looks like the connector is the problem, not the wires themselves.
@@tippyc2 My analysis was based on solid copper wiring (i.e., Romex). Typically, stranded aluminum wiring needs to one size larger for the same rating. So those 14 gauge wires are really equivalent to 16 gauge copper. Put another way, for the ratings I mentioned the four wires should have been 12 gauge to meet a 14 gauge copper spec. It could be that electronics have a different spec, but since it's used in a residential setting I'd think safety standard would be similar. I don't have enough information to definitively state the root cause; just wanted to point out that there wasn't much margin for error in what they used. I will say that the weak link in most electric components that I've seen is the wire and connectors used going into an assembly and not the supply wires. Just look at a typical light fixture. I hope the industrial engineering in the interconnects improves moving forward to keep up with this trend in power usage.
The design seems marginal.
If the adaptor is defective it becomes a risk too quickly
@@chillysourdough8924 sounds like you think those are stranded aluminum wires?
@@potatopobobot4231 it looked like it in the close ups Jay showed; if not then disregard.
I have to agree with Buildzoid on this one. Where Nvidia's adapter has female connectors that have 2 splits in them, other manufacturers are using female plugs with only 1 split (which is mechanically and functionally much more rigid), So Nvidia's female adapters have much less rigidity especially when bending horizontally. Once bent this would cause the same amount of amperage to transfer through a smaller connection point (from the adapter to the GPU) and in turn create more heat. am not a electrical engineer, but to me this would make more sense as to why the connectors seem to be melting at the pin connections rather than melting closer to where the solder point is on the Nvidia adapter.
Thanks for the tips and reassurance, Paul. Much appreciated.
nVIDIA has to recall these cables and issue fixed cables.
I suspect the melting is due to heating of the pin caused by high current arcs, itself due to embrittlement (phase changing) of the solder due to heat cycling. The embrittlement increases resistance at the connection point, lowering the voltage and therefore increasing the current (for a power controlled system) increasing heat. Eventually, the connection is so poor we see arcing which results is large temperature spikes. In one of the failure images you can see arcing tracers on the female end. I suspect that the specification requires these connections to be crimped and not soldered (like the Corsair cable) so I'm surprised this went through to manufacturing and passed QA. Bad day for the Manufacturing Engineer who signed off on this.
Safety aside, just water-cool the thing. And if that's not enough go sub-zero. Should be right up Jay's alley.
I believe it is just a manufacturing error. Or a combination of several Issues happening at once. Looking at Igor's adapter and at yours, you can see a big difference in soldering quality and how much of the cables actually touches the pins.
I think adapters that were produced first were good quality, but adapters made at a later time have poorer quality. It may also be a case of cutting corners of the supplier, because they can't make enough in time. A possible reason why that is happening is missing workers due to covid... Im just guessing of course, but it is not like it hasn't happened before elsewhere.
And 4090 FE's are not "burnung up" yet. It might be, that Nvidia simply uses those good "early bird" adapters and AIBs recieve the later bad "revisions" unintentionally.
That doesn't seem like sufficient isolation between +12V and ground inside the connector. I bet those two skinny little plates are somehow shorting a little bit inside the connector and going nuclear from there.
Does no one else think it is crazy that the gaming industry just keeps increasing its power draw when the rest of the world is trying to save energy? 600W is nuts. That's the same as a small room heater.
Good add Paul! A real value add piece. Your Dremel skillz are legendary, respect.
I was convinced this was a widespread problem with the new connector. I was very wrong. Just wanted to admit it.
Yet.
@@darreno1450 now you jinxed it.
Paul, this launch is the dumbest launch I've ever seen in technology. There is no stock, it doesn't fit in anyone's computer, and it literally bursts into flames when you use it.
Well thats Nvidia for you
It does not "literally burst into flames" but okay bro
Not for people with experience... I would never use an adapter on a 2000 dollar GPU as its an extra point of possible failure.. Get a 20 dollar dedicated cable and problem solved. It has to do with transitioning to new PSU and connector designs while trying to make current PSU designs work with the new.
Nvidia heard that new gen tech consumers are nicknamed "1st wave Beta testers" then went the extra mile and added the "final QA" title just for laughs.
In short, Nvidia shipped sub-standard power adapter cable for a 1600USD GPU.
Nvidia are completely idiots to develop this cable-joke. Now they should come up with a real solution and fast!
when I first saw the pics, I thought it would be a bad solder joint in the connector. I've ran into a similar issue once with one of the old molex connectors back in the day where it wasn't crimped correctly and it melted the connector.
problem with reviewer tests... they're all on open air test beds. But hey. Thanks for the cable tear down. That was fun.
It's possibly time that GPUs run on 24VDC instead of 12 thus reducing current.
I think they populated the two unused connections to prevent people from sending angry social media posts about "missing wires" in their connector 😂 Great series of videos from the community, hopefully it helps with the resolution coming faster and better!
Someone needs to make water blocks for power connectors 😂
As a former technician, seeing 4 power inputs shoved into 1 plug was the first red flag...
It's not melting, it's a feature called Nvidia Fusion technology as it fuses your gpu to your pc.