Compare it with 24 pin Motherboard connectors, which you cannot get removed unless you nail the motherboard to ceiling and make a swing out of power supply
I'm just gonna say "molex". Sometimes I wonder if the manufacturer put a vial of resin+hardener inside the connector that breaks when you bring them together, because they are not coming apart.
I had 3 out 4 cables with 12vhpwr, 2 from msi and 2 from thermaltake, came with PSU's. 3 cables had wires not holded in the connector and easily sliding out specially when cable is bent. I almost went of on MSI rep on a phone for that shit cable job.
I would say its a good idea (i mean run power from one cable instead of 3 or 4) but let's face it it was poorly executed and needed much more testing before being put out in the wild
I just remember being annoyed in how hard it was some times to remove the 8-pin power cables from my GPU. I don't think that's ever going to annoy me again.
@@alinzelnan haha. it is our fault for being a gamer and having no choice to buy their crapmanship. GRRR! they don"t care the costumer now, only for their own profit. not even hiring real people but all machines to do the assembly. at least if they just hire a real electronic or electrical engineer to do the calculation of what Power is, and not dumb at all to listen the advice, this would not surely happen. haha
@@dongwarrenmusic yeah sadly there'll always be people who buy that stuff. hopefully there will be some innovations (with energy efficiency as well) with Intel pushing into the market
@@WooGnar I'm sorry, but you're wrong on that. Cablemod has been replacing peoples melted GPUs that use their cables for many months silently. The cards they purchased from end users were then sent to NorthridgeFix for repair.
Wow, CableMod going way above and beyond! That's extremely generous of them, and really says a lot about how they'll back up their products. I will definitely order any cable extensions or whatever from them for my next PC.
Read between the lines and do some research. They only replace if their adapters are fault. If its user-error in future, they only offer repairs. User-error means: the connector isn't fully seated. Again, its funny, because the adapter has some design flaws because the adapter can wiggle out itself. And then its user-error.
It's not a problem...when your connector is concieved with robust specs, and then overbuilt on top of that. Did you know that extreme overclockers are able to push 1500W combined into 3 8-pin "legacy" connectors ? That 12VHPWR on the other end, is a complete joke in regards of it's conception.
This isn't the issue. It's not the size that causes the problems, nor the power the cables and connectors are able to handle. It's a flakey connection due to poor connector design.
@@kaasman78 I feel like 4090s should have a circular locking cable like some headphones and keyboards have (but scaled up obv). Maybe even a vga type design that screws in lol
If cablemod just asks you to send back the melted cable-mod connector WITH the graphics card, they can match the card and the adapter so it should be easy to filter out any false claims.
@@ivy7642 Indeed but there might also be just some very "inteligent" people that wiggle out the Adapter to force burning and then ask Cablemod for a replacement. I am sure such kind of people exist and they will do anything to harm others.
@@Trymon1980 I don't think that a defective GPU would work so nice to be able to constantly ask for the power needed to melt the connector so that someone could be able to reproduce the melting
I think with the increased power requirements, they should've increase the voltage to 24V instead of pushing more amps. Just like in USB-PD standarts. OFC this wil break backwards compability but It's been ages since last PSU standart.
This would complicate things when its not necessary if they just would make a good connector. Going smaller made absolutely no sense when the cards are already massive.
I am building with a 4090. Besides taking Steve's suggestion of making sure the cable is well seated before the video card goes in the case, I am going to make sure the that video power feeder cable is secured at two spots via wire ties with no stress so that hopefully, there will be little chance for it to unseat. But the big unknown is, of course, the effect of heating/cooling cycles on something as malleable as plastic. Thank you, Jay, for pointing out that this is still an open issue for those building with 4090s.
@@confused234milo9 I ended up building an aluminum scaffolding for the cable using load analysis (civil engineering; vector load; not electrical). No, problems. The card is more than I need for gaming, but even at $2000 USD, it was a real bargain for the AI stuff I am doing.
@@keineahnung-wv4ew I got agree with that sentiment, man. So many people complain and complain, and they still buy a 4090 at the end of the day, just to have it melt. Insanity.
do they have some kind of a contract with nvidia that they have to use this connector, and not some other connector and have this connector at the end of a cable? the temp probe option that doesn't even fucking work?
@@lasskinn474 yes. Nvidia specifies that if you want to use their silicon, you have to call your product an "RTX 40x0", you have to write "Geforce RTX" somewhere on the card, and you need to use the new pcie 5.0 spec power connector.
@@lasskinn474 yeah thats one of the reasons why evga stopped making gpus. they just went full on authoritarian and OEMs have to play withing their regulations
The more and more we hear about this connector causing issues the more I'm glad AMD just used normal connectors because this connector should've been sent back to the drawing board multiple times with how many issues it's had
"How long before they say we can't replace the card". Well, JayzTwoCents they refuse to replace mine. They offer to repair the connector but no replacement of the card. So, they went back on their word pretty quickly.
My strix can run 2985 oc all day. If it happens to me I’ll take the repair. By the way, I m very sorry it happened to you, it’s not your fault. Bad design by nvidia and pci sig. What card is it? I’m guessing you hit them up?
They should design a round plug which either screws in, or locks with a 180 degree twist, where in both cases the locking mechanism would push the plug fully into the socket.
Them replacing the card is amazing customer service. They will always have my business for the items they sell. I have used their cables on all my builds.
Them replacing the card is to prevent it being taken to court and cablemod being found liable for not designing the retention clip to the depth the schematics of the molex socket indicate. It has nothing to do with customer service.
It should be fairly easy to measure the voltage drop over the connector while moving it around. Also check with a FLIR after some time of use. Voltage drop X Amps = Watts (heat), likely in a tiny spot.
As this unfolds, I can't help but wonder if EVGA was just lucky or knew something to terminate their relationship with Nvidia when they did. Good analysis on the true issue with this one Jay, the video clearly demonstrated the probable root cause of this issue.
This gets speculated a lot, but if EVGA just wanted to avoid the 12VHPWR standard and they pulled out because Nvidia wouldn't let them use 3x 8pin instead, then they also would've either a) pulled out of the PSU market if they were being forced to include 12VHPWR on the PSUs or b) not included 12VHPWR in their new PSUs if they weren't - but they've done neither.
I think it was just EVGA saw what the 4090 FE could do at a cheaper price And they were left with no head room to overclock their more expensive version Essentially Nvidia decided to go head to head with their board partners while leaving them no real advantage in performance or way to make a meaningful profit
I think EVGA made an enormous amount of money during the mining boom and they wanted to keep all of it instead of reinvesting it into the 40 series. Now they've reduced a huge amount of their overhead/labor and can continue selling PSUs, mice, keyboards, older gpus, etc, just enough to keep the lights on until they can sell the name for big money. Basically, they're just cashing out in time for the owner to retire. Many of their ex-employees now work for Nvidia.
P.S. Maybe now the "board partners" will actually have to do their own R&D and actually find how to improve the product, not just change a few numbers in the settings and add some shiny lights.
I have always checked every connector to make sure that it has stayed seated after cable management on BOTH ends (source: power supply, fan controller, etc. and endpoint: motherboard, graphics card, fan header, SATA, etc.) before ever powering on a system whether it is necessary or not. I think it's a good rule of thumb to have and I have never had any issues whatsoever of this kind in over 20 years of computer building. Massive kudos to CableMod for doing what they're doing in this particular instance.
Good advice there! It is a good final check to ensure that everything is connected to where it should be, and a final check after cable management takes just a minute to avoid potential problems. I will be doing this with future builds.
I've always done the same as best as you can anyway - just the act of putting the case panels on could easily shift stuff around a bit, and that terrible connector would probably be in its fallen out state really easily, or so it seems. If you ever have to move that computer... Well with the rather overweight nature of the 4090 causing lots of droop and flexing potential and it needing such a large power cable - I think carefully superglueing that adaptor to the card in the most certainly fully seated position may just be required, especially if Jay is correct and the adaptors other end mates with the PSU cable reliably.
Cablemod has always done a great job with this stuff. I will gladly pay the CM premium, knowing that IF something goes wrong they'll have my back. All the cables for my last few builds have been CM, and they will continue to be for as long as they're upstanding.
@@kyoudaiken Several graphics cards SKUs were using 4 already and nobody batted an eye. Why adding one more is now an issue? These cards are already gigantic because of how inefficient and power hungry they are.
I would like to see cables for graphics power connectors with screws. Just like the old VGA monitor cables that you couldn't disconnect without unscrewing the screws. Maybe this would finally prove that the manufacturers and not the users are at fault. If we draw as many watts as the car battery when starting the car, then the cables should be of the same quality and have the same attachment.
honestly, i've considered desoldering the PCIE5 port and directly soldering my cable to the board to completely eliminate the problem... but, how practical is that for everyone to be doing? most people can barely figure out a jumper let alone how to solder correctly.
Fun fact, i recently built a 3D printer, and the wire harness that attaches to the pcb on the toolhead uses the same family of small connectors (just a 14 pin instead of 12 pin). My first thought was like "thank god i don't have to pull 600w on this one"
I don't know much about electricity, But as soon as I started hearing rumors years ago that Nvidia was planning to replace the huge 3x 8 pins with that puny little 12v connector, I knew it wasn't going to end well. I never expected to actually be right because I don't know anything about electricity and I assumed Nvidia engineers knew more about electricity than I do, but it is clear that they don't.
I undervolted my 4090 and placed a thermocouple on the connector. I don't want to take any risks. (If the temps goes above a set point the PC shuts down)
I also appreciate the coverage Jay. I have the 4090 Strix water cooled as well. Mine just got toasted after a month of use despite constant paranoia and checking on proper seating even after having it all hooked up and running. I've also vertical mounted with a perfect flat run out to the back of the case due to extreme caution. Fried on the 6 12V pin side and non-ground side. I'm also seeing new issues with RAM code failures on CPU post the crash from the melted card. Am still troubleshooting unfortunately.
i wonder what the resistance values do when you wiggle the connector. id put leads on either end, test for resistance, and wiggle the thing. i bet the numbers would be interesting
Was thinking I'd like to do that, but don't have a 4090 or a Cable Mod adapter. Would be a 4-wire resistance test and would need to know the currents involved (from the specification), but would be easy to record the change and look at the worst case power dissipation in the connector.
Hey Jay! I just bought an aorus 4090 master and I'm fairly concerned about the connector melting. i have an o11 dynamic XL and the cable is up against the side panel. i just built it and am running it with the side panel off atm and am wondering if its fine to put it back on? Thanks in advance
What they almost need is some sort of clamp that goes around both halves of the connector and keeps them from moving at all, since this is going to continue to be an issue until the connectors are redesigned.
@@coccoborg a physically bigger connector with single clip in the middle made the same way will have even bigger problem. A small gap on 1 side will be even bigger gap on other side, simple geometry. Older types od wide connectors had 2 clips on sides which render them "unwiggleable". It is as simple as that.
me too, that's really impressive if that's true. Brands like Asus will just do the general "fuck you" move when it's their fault. Like that infamous incident with their Dominus Extreme mainboard. Not just Asus, other brands too. Kudos to CableMod. But you still have to ship your card to them and wait
I immediately noticed on my adapter when I got it, that also the plastic part from the plug moves on the PCB of the adapter. You can rotate it a litte. It is not fully attached to the PCB of the adapter.
I have no idea of how much power goes through that plug but a loose wire in your house can arc and arcing causes it to heat up. (like arc welding, it will melt steel) so is there enough power to melt plastic.....I don't know . just asking.
With all that wiggle room, it could potentially wiggle itself out from vibrations too. BeQuiet has had one (that I’ve seen) incident where it melted in the PSU side also, and they are takin* it seriously too. But yeah, someone fucked up when designing this…
And the worst part: it didn't even need redesigning. And even if we really needed a new plug, why not just use one of the dozens exiting, proven power plug designs...
@@stephanweinberger Yep, fully agree. I like the larger connector. Thick, solid pins. I guess they needed to make it "smart" with the sensing stuff. But could've retrofitted that too like they did with the extra 2 ground pins.
I made it very clear on multiple occasions that cablemod as far as i can tell has nothing to do with melting connectors since the melting is happening on the graphics card side and not cablemod side.
Hey Jay, I've been saying this for a long time. Aesthetics aside, with these cards running this much power, they need to use kind of attachment that we had back in the day with 2 screws, one on either side. Think the old VGA attachment method. This locks it in. no wiggle, no movement of any kind and they could have made it look fine astatically, so it's certainly doable and will fix the problem. Nvidia needed to put the screw holes in at the factory, as I don't think making an aftermarket adaptor is possible. I realize the I have the benefit of hindsight, but Nvidia, as big as they are, should have seen this coming a mile away. (and probably did)
We have friction lock connectors capable of handling higher amperage requirements than this. This is just a poor implementation period. There are even more compact friction fit connectors that are more reliable and capable of providing this kind of current. They wouldn't look as pretty, but they would work. Take a look at the XT90 for example, ceiling for that connector is 90A which is DOUBLE the amperage requirement of the 4090. The two problems with it is that it is a solder pot termination, and it takes two very large gauge wire to handle the power delivery. That connector has NO retention mechanism, and there is NEVER a problem with a fully seated XT90 connector coming loose or being able to "walk" sideways like these connectors can and do.
@@ZeroX252 1.50 € for a power connector is probably too expensive for a 1650 € graphics card, the investors of nVidia would complain. I will not buy such poorly designed garbage. It has been shown that the RTX 4090 would run more efficiently with a lower power target, so all the "effort" with 12VHPWR could have been avoided even.
This seems like an easy fix on the male end of the connector. You just make the latch tolerance tigher so it can't back out at all or increase the thickness of the posts so they don't go in or out as easily.
Longer pins would stop it air gapping when it torques. Rather than one brand of card, I would almost guarantee that the connector manufacturer is the same. Since metal is expensive, they probably cut costs by using the shortest pins possible.
im just wondering is it stock 4090s are the only ones melting or the ones with water cooling as well. maybe just heatbuild up which lead to thermal runaway
Putting two clips in the connector would require also Nvidia or OEMs to change the card's connectors, which i believe they are not willing to do. One possible solution would be to make the clip shorter and wider or make the connectors inside longer.
I really like that bright yellow highlighter color 2 tone 12VHP plug one of the brands was showing at computex. doesn't fix the geometry of the design but visual clarity is a huge plus.
Measure the plastic part of the connector's dimensions/tolerances vs PCIE SIG spec. Either the spec is too loose, or the plastic itself is out of spec. I assume the spec would define a maximum of something like 0.6mm of play when the clip is engaged. That looks closer to 1.2mm.
I only found Intel PSU design guide - "ATX12VO (12V Only) Desktop Power Supply Design Guide" (I couldn't find PCIE SIG spec without membership) The Intel design guide doesn't specify the clip dimensions, but I can guess the official PCIE SIG spec will mention something about it, because even the Intel guide refers to it in some places. It does however mention dimension tolerances ±0.2 mm and ±0.25 mm max, and that the latch has to withstand 45+ N (4.5 kg) of force when pulled axially. What Jay shows here looks completely wrong. The connector looks so loose it would wiggle just by blowing at it, so missed tolerances? (my Gigabyte RTX 4090 + FSP PSU was *really* hard to plug, my fingers were hurt from pushing it!) And I highly doubt the "withstand 45 N of pull force" allows for this big disconnection while still latched. I would say the connector itself is not a bad idea, from the electrical standpoint it should be completely fine. But IMO it has really tight tolerances thus bigger margin for error. And we all know how manufacturers like to cheap out here and there.
Might be mixed manufacture - companies aim for mid tolerance on their tooling, but it can turn out to be slightly above or below (but still within the component tolerance). Say manufacturer 1 are at one end of the tolerance and manufacturer 2 are at the other. Both are fine with their own plug and socket interface, but when mixing the plug from one and the socket from the other, it falls outside of the intended design spec - even if both parts are "in spec".
I think the clip system on the 12 pin cable needs to be re-adjusted or realignment so there will be less play and so the cable end won’t back off slightly from movement/vibrations, also the 6/8 pin connecters didn’t have this issue before and felt more solid
I have several cable mod cable kits over the last couple of years and they have always been of the highest quality with no issues and look great. It's nice to see a company go over and above what they need to do for consumers. I really appreciate that especially now with all the company's that won't.
I agree, that is why I am throwing business their way now and in the future even though their custom stuff is pricey. If they are supporting consumers by covering their replacement cost, they are single handedly holding up the enthusiast market. They are going way above and beyond and they deserve my support.
I was on the fence about getting a 4090 for a while, but end the end I decided to save money and get a PowerColor Red Devil 7900 XT. I am now SO GLAD I did that, since I don't have to worry about all this cable adapter/melting/problems going on. Between this and nVidia's scummy practices, I'm staying with AMD.
@@fernandolezama3325 Great card nothing to complain about really. Performs as expected from benchmarks/reviews. Only 2 things that don't often come up in benchmarks/reviews. if you have mismatched displays it could lock (not always btw) the vram at max clock so you can get ~80-130watt idle draw with 2+ displays. That was the case for me but fixed it swapping out an ancient display I was still using for just random stuff on the side (I have 4 displays all different brands and refreshrates sitting at ~45-55 watt idle). Oh and NVIDIA cards are not immune to this btw just never had it on my older 10th gen and below cards and overall you hear way less about it compared to AMD cards. And if you are planning on playing VR a fair bit just Google around for a bit the optimization seems to be hit or miss on some games/headsets. Haven't looked much into the VR thing cuz I don't use it
@@GGbowz2101 "I have 4 displays all different brands and refreshrates" - You absolute heathen!!! This would probably drive me insane. The chaos...the disorder...the asymmetry 🤣🤣🤣 Just kidding, still better than producing just more e-waste and buying everything new but damn...this would took some time to get used to for me!
@@Gruftkriecher Yep it ain't the prettiest.. But they are all just my old main displays that rotate untill they are the 5th in line and get kicked out the door haha. And you know they are VESA mounted so I atleast don't have 4 different looking stands lmao. I would just much rather spend the money used for matching displays on a PC or main display upgrade you know
The clip retention sloppiness is why I plug into the GPU last. I do all of my cable management BEFORE plugging into the GPU, because I don’t move that wire at all once I know it’s fully seated.
I've seen this with other Molex style connector/sockets/pins in the industry. I think what happens and what you are dancing around is that there is too much play in the connector. This can lead to a stretching (fatigue) of the socket (female contact) which then reduces it's ability to grasp the pin (male). This leads to less surface area coming in contact between pin-socket interface. Less surface area = less conductance (or more resistance if you please) leading to the mating area heating under high current conditions. While not as relevant if you don't have vibration or regular insertion/extractions, is fretting corrosion. Fretting corrosion occurs when small amounts of the mating surfaces scrape off due to rubbing against each other. These microscopic bits of material corrode easily, and oxidized metals are poor conductors (that's why gold is the "gold" standard for contacts due to it being nearly impossible under normal circumstances to corrode). These particles are trapped between the mating surfaces of the contacts and decrease the conductance over time. Another issue is the corroding of the surface of the contacts themselves. There are many things in a consumer environment that contribute to corrosion, including the consumer. What I see as the problem with these connectors, and what you've demonstrated is the amount of play in the connector. The clip should be much wider than it is, and the latch on the card connector needs the width to match. This should help minimize the amount it can move side to side. The other issue is the latch on the card needs to be further down so that once the two connectors are mated, you can not pull them apart as far as you've demonstrated. The last point will also help with the side to side movement. There are a couple other solutions. The first would be a completely redesign of the connector. That solution would be to increase the mating surface area. So either make the diameter of the pins larger, and the sockets to match, or increase the area by increasing the length of the pins and sockets. If corrosion is a contributing factor in this failure, then change the metal used to make the contacts, something like brass.
I just built a system with the new plug. When I first looked at the pins on the connector I was shocked how small they are to carry 600 watts. I believe that's the whole problem.
The terminals are specced to carry more current than the microfit jr. terminals of the 24-pin atx, pcie etc.. But they're pushing so much current through there that the safety margins have become way smaller.
@@joeyverliesharennot to mention... If either end of the connector is out of spec, you can have poor pin-to-sockets connections. If the sockets don't grab tightly onto the pins, you'll get excess heat
Yup it is, if there was a larger margin of error, say the cable was designed for 800 watts, small issues wouldn't be a big deal because it would run at lower temps. Instead they have a cable running near it's limit so anything goes wrong and out comes the magic smoke, which never goes back in.
@@joeyverliesharenLess metal and smaller are specced for more power? They have to be making it up in design or tolerances. Seems like a better designed standard atx pin of the same size would be foolproof then.
@@jon4715smaller gauge wire and less material goes against physics. They are making it up because they want a small sexy adapter to make amd cards look like pigs because they use a “old” many connectors. Form over function, and people are defending this is unreal.
Perhaps a more longterm solution would be for cable mod to work out something with NRF And other places to get customer cards in and repaired at a discount, sure going without a card for awhile would suck but it could also build a potential case for a future lawsuit that could force companies to get together and fix this connector issue
I just got my 4090 Suprim Liquid X.. what is the best recommendation to powering it safely to prevent something from melting?? I currently have Corsair's 1200x Shift that came with their one cable for their gen 5 cables.. It's firmly plugged in and the cable isnt bent much so not sure how likely it is to happen to my card.
The problem is that the new connectors are so small and dinky what even the slightest tension can unseat it. Imagine if every plug coming off your PSU was like this.
All i know is that they made a problem that wasn't a problem before that connector So that's the evolution nowadays. Don't make anything better, screw something up and try to fix it just to have something to say you do
I have a foolproof suggestion. Maybe the solution is to go back to the old-school approach. The cable connecting the monitor to the GPU used two thumb screws attached to a metal plate to secure the cable to the GPU. This method prevented ANY connection play. What do you think?
They won't do it, it's a profit scheme at this point. They want customers to buy a 2nd and 3rd card when they can void warranty because "it's not their fault" in the eyes of the law.
Jay, I'm glad I took your advice and went with a 3090ti instead of waiting for the 4090. Looks like I saved myself a bunch of worrying and cash. Thank you.
Wow, with all the 4090 problems surrounding both the connector burn and cracked pcb's. I highly recommend vertical mounting that card and plugging that connector in directly in the top with no 90deg adapter. That's what I did. The vertical bracket from Lian Li is solid and handles all the weight of the card without any stress on the pcb, and with the cable / connector clicked directly in to the top with it's weight pushing it in to the connector without any additional play or wiggle, I don't have to worry about any of that stuff.
I think the point is: The 12VHPWR connector is defective by design. That's it. And that was obvious to me since the beginning. The moment I saw the connector I knew that's not capable of handle all the power that is specified to handle. Or at least that the margins where too short. Everyone that tries to make products around that connector will have problems. I think that CableMod and the entire community should ask for a better solution to power those graphics cards. A more robust connector. I know that this would be a lot to ask from the manufacturers but I think that's the best possible solution. And the right one.
It's the connector retention design that's poor. Not the power delivery design. When operating normally, there is no issues. It's not a size issue or a gauge issue....but a 'solid connection' issue. Like Jay said...if the retention clip was a bit shorter, there would have been no play. Maybe a bit wider would also help.
I'd say it's possible that due to the vibration of any moving parts, it could shake itself loose to the point the clip is the only thing keeping it attatched at all.
Why the good guy? As can be seen in the video, their adapter is unreliable. They offer warranty, but it is not about money. You want to use such a product reliably, without RMA process. An RMA process makes you lose time, because you have to send in the card and wait for a replacement. Furthermore, the logistics involved leads to unnecessary global warming. nVidia has to do proper design, as well as PCI-SIG and Intel - period.
I can confirm the great Cablemod support as i've had first hand experience. I had the new CM right angle adapter on my GB 4090 and it melted, both the CM adapter and the plug on the GPU. I reported it to CM. I sent them all the pics and proof of everything, shipped the GPU to Northridge fix etc. and they fully supported me, reimbursing me for the GPU, the original adapters I purchased from them and provided me with new ones which i'm using on the new 4090. Great support from CM. Although as Jay mentioned I'm concerned how long they can keep doing this. This is not even their fault, Nvidia needs to take accountability of their design flaw!! Who knows if this will happen with my new GPU, at that point maybe Cablemod won't be so generous, and I can't blame them.
Same actually. My Asus 4090 and CM right angle adapter fused together this past weekend. It doesn't visually appear it melted, but I literally smell burnt plastic long after i killed the power. Were you able to separate your adapter and GPU?
@@Angelic257 yep, the first indication was instability, crashing, blue screening. Then started smelling burnt plastic. It did seem to be more difficult to remove the plug but after I did, noticed the burnt pins on the CM adapter and the gpu plug. I had started an RMA with Gigabyte but cancelled it after I realized cablemod was offering to cover the cost of everything, so was really impressed by that. If i had gone through GB, they'd probably give me the runaround and blame me or CM. Just hope this doesn't happen with the new msi suprim.
I'm not using one of CableMod's 90 or 180 adapters, but I am using their 12VHPWR cable. It was more readily available than the official Seasonic one, but the way they've been standing behind their product just made it an easy decision.
Would it not be an idea to glue the adapter clip into place on the GFX card side and simply disconnect and connect the cable to the adapter if we need to ever remove the card from the machine? Surely that would remove any chance of the interface coming loose at the card side connection?
I think the most important thing here to think about is the failure rate. Most of us will NEVER have an issue regardless of what connector we use. Yes they are failing and it's good to know about but the internet has a way of blowing things out of proportion. Also, it's great that Cablemod has such a good warranty but it's too bad it's even necessary and I hope that the failure rate remains low enough for them to be able to continue on. Personally I have a 4080 and a 4070 so I don't have to worry about it lol.
I think the base design of the connector is overlooking user interaction. They should put a metal retain hook to have a more clicky noise when it engage to prevent user error of not pushing the plug enough and maybe shorten it to prevent that much play on a plug that have such small tolerance. I mean, the plug have smaller pin, it mean the resistance to put it increase and make people think that it is seated when it is not fully seated. The other option would be to put a plastic pin that would raise when you reach a sufficient depth so you have a visual indicator of it being set correctly.
MSI made connector's end in bright yellow. Meaning, that if you see yellow the plug is not inserted enough. Edit: it's kind of visual indicator you mentioned.
@@Mervinion Nice move from them, i mean, having a visual indicator that the position is not good should be done. maybe cable mode could put some colour on their adaptor, but as jay shown, you have to really watch to see it, but another colour would help.
@@Mervinion The problem is the play the connector has, the wiggle room. This problem cannot be resolved with color, the connector has to be redesigned. In many industries reliable power connectors are used. PCI-SIG and Intel just have to do some research, and be serious about the matter.
Adding anything will add more resistance thus more heat will be the result. It seems to be a design flaw in the undersized connector. (Contact surface area)
Yep. They should have used modified XT-90 with additional contacts for the power designation and also a locking mechanism that involves METAL and not plastic crap.
As someone who's just bought myself a new prebuilt with a 4090 that'll be here soon, and worrying about this issue, I appreciate the videos from you and Gamers Nexus displaying this so I can be more confident that mine won't have that issue, as long as I make sure the thing's ACTUALLY seated properly, flush with the GPU, no gaps, not crooked.
Lets be real, the entire design of that plug is a disaster waiting to happen, why using 3 or even 4 8pins, when we can force 450-600Watts thru one small plug? Yeah great idea, lets do that, what could possibly go wrong? 😉 IIRC Igor from Igor's Lab was involved in the Design of those adapters and he surely knows his stuff. And when you are called CableMod and are selling alot of cable related stuff, you sure do as well. So no wonder they stand by their product, as it is a solid and cool design, while also stayin cool as this is by design. But replacing cards that used the adapter, holy moly that really goes above and beyond, RESPECT! 😇
The whole problem is that many 4090 cards are too wide to fit in many midtower cases...gigabyte recently came out with a version of the card that had the port in a different location on the card probably for this very reason XD. No one really would need a cable mod adapter if there was enough room for the one provided by Nvidia/board partners.
I route cables, then plug. :P BTW, does CPU fan sucking air straight near upper connection of 180 degree adapter help in any way to keep connector temperature lower?
The question in my mind is, are these connectors and their sloppiness an actual manufacturing defect, or is it just the sloppiness of the standard? It looks to me like Jay said in the end of the video, if the latch had been just a little bit forward that there would be no issue. I seem to remember the molex connectors and other snap-in fittings on my older computers were really kind of a bitch to connect, and were even tougher to get loose again afterwards.
I understand the concept behind the 12VHPWR adapter, that it was supposed to make it easier to get 600W to the graphics card. But if it being not *100% absolutely unequivocally perfectly seated* is enough to frikin melt it then *the resistance is too damn high*
Yup and it's crazy that people aren't calling this out as just straight up unsuitable for the open market. If something has to be installed 100% perfect it's not ready to be sold to general consumers.
@@bobbygetsbanned6049 People still fuck up installment of their graphics card, CPU, and RAM. Should they all be recalled since they have to be installed 100% too.
@@hirotakasugi4891 The difference between that and this 12VHPWR connector is that once the RAM, GPU and CPU are seated properly, they generally stay that way. Whereas with the 12VHPWR connector, it only takes some a small bit of shaking on the case / pulling on cables (i.e. during cable management or cleaning) for the connector to no longer make 100% contact. This is a fundamental design flaw.
I think a link to Cable Mod should have been placed in the description. I wanna buy from them now. Any brand that stands by their product deserves my money.
The material is sufficient for the circuit to flow unimpeded, the failures have nothing to do with the surface area as a whole, but lack of it when the user has not got a flush connection or is using an improperly made adapter like these cablemod clowns have produced.
@@Squall4Rinoa Yes, and no, if it was 90% copper used, or high-grade stuff - then sure (not even mentioning that we know that there is almost none resistance through nano-wires) but there is also 99% copper - just as soon as you consider that you know why this is an issue - 'c o s t c u t t i n g'.
@@michalwiktorow2188 Then I will cut costs and not buy a graphics card with 12VHPWR connector. These cards are the most expensive ones and cheap out at the power connector. Ridiculous!
I tried to wiggle the old style 8 pin cables in my old 1070 and it wiggles just as much , if not more then what is shown in this video with the 12 pin connector. The clip on that one was also not securing it tightly, but somehow those cable didn't have such issue, probably because they were rated for lower power and the tolerance on them was much higher to these sort of things then it is for the 12 pin connector.
I kinda like the idea of a slotted power delivery standard. I remember when I had a PCI-e X800 card that did not require external power, slot it and done.
Compare it with 24 pin Motherboard connectors, which you cannot get removed unless you nail the motherboard to ceiling and make a swing out of power supply
😆🤣
I'm just gonna say "molex". Sometimes I wonder if the manufacturer put a vial of resin+hardener inside the connector that breaks when you bring them together, because they are not coming apart.
out of all comments, yours is the most accurate experience many of us share haha
Oh, so your 24-pin cables plug in normally?? I swear it feels like I'm breaking the board every time I plug them in
I'll have to try that trick next time, the neighbors give you funny looks when you are taking PCs apart with a rope and the bumper on your truck.
I'm beginning to think the 12VHPWR connector wasn't such a great idea. Can't put my finger on it though...
I had 3 out 4 cables with 12vhpwr, 2 from msi and 2 from thermaltake, came with PSU's. 3 cables had wires not holded in the connector and easily sliding out specially when cable is bent. I almost went of on MSI rep on a phone for that shit cable job.
because it's melting 🤣
The pins are too small in my opinion.
I would say its a good idea (i mean run power from one cable instead of 3 or 4) but let's face it it was poorly executed and needed much more testing before being put out in the wild
You need to apply even pressure across your finger to get it ;)
I just remember being annoyed in how hard it was some times to remove the 8-pin power cables from my GPU. I don't think that's ever going to annoy me again.
It's 100% a cost cutting thing from the manufacturer
that was their intent of saying it is our fault. because we always want it smaller. lol
@@dongwarrenmusic and meanwhile the cards are so massive they fall through the floor if you accidentily drop them lol
@@alinzelnan haha. it is our fault for being a gamer and having no choice to buy their crapmanship. GRRR! they don"t care the costumer now, only for their own profit. not even hiring real people but all machines to do the assembly. at least if they just hire a real electronic or electrical engineer to do the calculation of what Power is, and not dumb at all to listen the advice, this would not surely happen. haha
@@dongwarrenmusic yeah sadly there'll always be people who buy that stuff. hopefully there will be some innovations (with energy efficiency as well) with Intel pushing into the market
Huge, huge respect to cable mod for going WAY above and beyond to help consumers. I've never seen a company do that
If the adapter costs *$50*, what they are actually selling is an aftermarket warranty for 12VHPWR melting.
only doing that cause Jay and Steve are publicly calling them out.
@@WooGnar I'm sorry, but you're wrong on that. Cablemod has been replacing peoples melted GPUs that use their cables for many months silently. The cards they purchased from end users were then sent to NorthridgeFix for repair.
@@kulilin3104 100% agree. It has nothing to do with Jay or Steve
what are you talking about? NZXT did way way way more to help comsumers when the H1 case started setting fire
Massive respect to cable mod for going above and beyond on a problem that isn't exactly theirs to deal with.
Just because of this, they've definitely gained a customer for the future.
It's not sustainable.
Its an industry problem tbh. A failing cable standard like this shouldnt exist.
@@jasonwatson9011 Oh I agree. This problem should not exist.
@@Allacks and a customer now
Engineering 101, an increase in failure even when caused by user error is still considered a design flaw.
This community knows what to do, keep supporting CableMod as often as possible. They've absolutely earned more business and respect.
I'll wait to hear the first public rejected warranty excuse
@@rockyninja5089 found anything yet?
The community would be blatant morons then.
happy "voluntary" recall day!
Wow, CableMod going way above and beyond! That's extremely generous of them, and really says a lot about how they'll back up their products. I will definitely order any cable extensions or whatever from them for my next PC.
Read between the lines and do some research. They only replace if their adapters are fault. If its user-error in future, they only offer repairs. User-error means: the connector isn't fully seated. Again, its funny, because the adapter has some design flaws because the adapter can wiggle out itself. And then its user-error.
Lets increase the power for our cards and make a smaller connector to concentrate all the heat to a small area. Sounds like a great idea...
✨Aesthetics✨
It's not a problem...when your connector is concieved with robust specs, and then overbuilt on top of that. Did you know that extreme overclockers are able to push 1500W combined into 3 8-pin "legacy" connectors ?
That 12VHPWR on the other end, is a complete joke in regards of it's conception.
This isn't the issue. It's not the size that causes the problems, nor the power the cables and connectors are able to handle. It's a flakey connection due to poor connector design.
@@kaasman78 I feel like 4090s should have a circular locking cable like some headphones and keyboards have (but scaled up obv). Maybe even a vga type design that screws in lol
@@kaasman78because 1500/2000w through a few 11ga wires is always a bad idea
I hope this warranty won't be abused! We need this to exist.
^THIS^
If cablemod just asks you to send back the melted cable-mod connector WITH the graphics card, they can match the card and the adapter so it should be easy to filter out any false claims.
@@ivy7642 Indeed but there might also be just some very "inteligent" people that wiggle out the Adapter to force burning and then ask Cablemod for a replacement. I am sure such kind of people exist and they will do anything to harm others.
@@Trymon1980 I don't think that a defective GPU would work so nice to be able to constantly ask for the power needed to melt the connector so that someone could be able to reproduce the melting
I think with the increased power requirements, they should've increase the voltage to 24V instead of pushing more amps. Just like in USB-PD standarts. OFC this wil break backwards compability but It's been ages since last PSU standart.
and less connector points..
This would complicate things when its not necessary if they just would make a good connector. Going smaller made absolutely no sense when the cards are already massive.
I am building with a 4090. Besides taking Steve's suggestion of making sure the cable is well seated before the video card goes in the case, I am going to make sure the that video power feeder cable is secured at two spots via wire ties with no stress so that hopefully, there will be little chance for it to unseat. But the big unknown is, of course, the effect of heating/cooling cycles on something as malleable as plastic.
Thank you, Jay, for pointing out that this is still an open issue for those building with 4090s.
Hows the 4090 been these months later?
@@confused234milo9 I ended up building an aluminum scaffolding for the cable using load analysis (civil engineering; vector load; not electrical). No, problems. The card is more than I need for gaming, but even at $2000 USD, it was a real bargain for the AI stuff I am doing.
There was a time when manufacturers knew how to select the size and type of connector that was appropriate for the load it was meant to carry.
@@keineahnung-wv4ew I got agree with that sentiment, man. So many people complain and complain, and they still buy a 4090 at the end of the day, just to have it melt. Insanity.
You wrote load 😅
do they have some kind of a contract with nvidia that they have to use this connector, and not some other connector and have this connector at the end of a cable? the temp probe option that doesn't even fucking work?
@@lasskinn474 yes. Nvidia specifies that if you want to use their silicon, you have to call your product an "RTX 40x0", you have to write "Geforce RTX" somewhere on the card, and you need to use the new pcie 5.0 spec power connector.
@@lasskinn474 yeah thats one of the reasons why evga stopped making gpus. they just went full on authoritarian and OEMs have to play withing their regulations
The more and more we hear about this connector causing issues the more I'm glad AMD just used normal connectors because this connector should've been sent back to the drawing board multiple times with how many issues it's had
inb4 AMD adapts the 12VHPWR in their next generation. That would be the foot-in-mouth-way AMD works often.
This burning issue would’ve easily been solved if they increased pin size by like 15-25%
No more buggy than AMDs AM5 platform and burning CPUs.
There's always problems with new designs.
Can't wait to go AM5 though, ugh 🤤
@@andrewvirtue5048 bro that cpu burning issue is asus’s fault
@@andrewvirtue5048 That wasnt AMD fault it was ASUS motherboard bios giving the CPU to much voltage.
"How long before they say we can't replace the card". Well, JayzTwoCents they refuse to replace mine. They offer to repair the connector but no replacement of the card. So, they went back on their word pretty quickly.
My strix can run 2985 oc all day. If it happens to me I’ll take the repair. By the way, I m very sorry it happened to you, it’s not your fault. Bad design by nvidia and pci sig. What card is it? I’m guessing you hit them up?
the problem with that 40:000 number is that it is not disclosed how many of those are sat on shelves waiting to go to consumers.
They should design a round plug which either screws in, or locks with a 180 degree twist, where in both cases the locking mechanism would push the plug fully into the socket.
for real they need professional grade ports at that point
Reminds me old VGA cables with screws.
Twist and lock.
a round plug isn't smart cause it takes away to much space
They should just make better plugs. It's not that hard.
This makes too much sense for it to actually get implemented.
Them replacing the card is amazing customer service. They will always have my business for the items they sell. I have used their cables on all my builds.
Them replacing the card is to prevent it being taken to court and cablemod being found liable for not designing the retention clip to the depth the schematics of the molex socket indicate.
It has nothing to do with customer service.
It should be fairly easy to measure the voltage drop over the connector while moving it around. Also check with a FLIR after some time of use.
Voltage drop X Amps = Watts (heat), likely in a tiny spot.
I'm just curious, are manufacture of that plastic housing of plug & socket fallow the spec, if we see then being loose in vid ?
As this unfolds, I can't help but wonder if EVGA was just lucky or knew something to terminate their relationship with Nvidia when they did. Good analysis on the true issue with this one Jay, the video clearly demonstrated the probable root cause of this issue.
This gets speculated a lot, but if EVGA just wanted to avoid the 12VHPWR standard and they pulled out because Nvidia wouldn't let them use 3x 8pin instead, then they also would've either a) pulled out of the PSU market if they were being forced to include 12VHPWR on the PSUs or b) not included 12VHPWR in their new PSUs if they weren't - but they've done neither.
I think it was just EVGA saw what the 4090 FE could do at a cheaper price
And they were left with no head room to overclock their more expensive version
Essentially Nvidia decided to go head to head with their board partners while leaving them no real advantage in performance or way to make a meaningful profit
$10 they started testing 40XX cards, saw this melting and went "nah fuck this shit, We out." /s
i really think EVGA just had enough of Nvidia's shit.
I think EVGA made an enormous amount of money during the mining boom and they wanted to keep all of it instead of reinvesting it into the 40 series. Now they've reduced a huge amount of their overhead/labor and can continue selling PSUs, mice, keyboards, older gpus, etc, just enough to keep the lights on until they can sell the name for big money. Basically, they're just cashing out in time for the owner to retire. Many of their ex-employees now work for Nvidia.
P.S. Maybe now the "board partners" will actually have to do their own R&D and actually find how to improve the product, not just change a few numbers in the settings and add some shiny lights.
I have always checked every connector to make sure that it has stayed seated after cable management on BOTH ends (source: power supply, fan controller, etc. and endpoint: motherboard, graphics card, fan header, SATA, etc.) before ever powering on a system whether it is necessary or not. I think it's a good rule of thumb to have and I have never had any issues whatsoever of this kind in over 20 years of computer building.
Massive kudos to CableMod for doing what they're doing in this particular instance.
Good advice there! It is a good final check to ensure that everything is connected to where it should be, and a final check after cable management takes just a minute to avoid potential problems. I will be doing this with future builds.
I've always done the same as best as you can anyway - just the act of putting the case panels on could easily shift stuff around a bit, and that terrible connector would probably be in its fallen out state really easily, or so it seems. If you ever have to move that computer... Well with the rather overweight nature of the 4090 causing lots of droop and flexing potential and it needing such a large power cable - I think carefully superglueing that adaptor to the card in the most certainly fully seated position may just be required, especially if Jay is correct and the adaptors other end mates with the PSU cable reliably.
Actually the 6 pin or 8 pin still much better, won't break your gpu if not put it in properly...
Sounds like they need a metal 'spring clip' system. Many auto wiring harnesses have them. Absolutely positive hold down and engagement.
Cablemod has always done a great job with this stuff. I will gladly pay the CM premium, knowing that IF something goes wrong they'll have my back. All the cables for my last few builds have been CM, and they will continue to be for as long as they're upstanding.
The solution: Start using those 8 pin power connectors. Those might be great! You can use like 2, or even 3!
Issue is for 600W you need 5. LOL. But yeah, maybe make reasonable GPUs instead.
@kyoudaiken No 600w you need 4. 150 watts per connector. But obviously 4 is still a large amount of connections coming off psu and going into GPU.
This appears to be a case where they fixed something that wasn't broken. The older connectors worked just fine, all the time.
@@kyoudaiken Several graphics cards SKUs were using 4 already and nobody batted an eye. Why adding one more is now an issue? These cards are already gigantic because of how inefficient and power hungry they are.
@@BrunodeSouzaLino Several? Got any examples beyond XOC cards?
The heat pad is likely there more for electrical isolation from the metal cover, than actual thermal reasons.
on the website there's a recall of those adapters, so what the hell do I use so I can put my glass back on my case?
I would like to see cables for graphics power connectors with screws. Just like the old VGA monitor cables that you couldn't disconnect without unscrewing the screws. Maybe this would finally prove that the manufacturers and not the users are at fault. If we draw as many watts as the car battery when starting the car, then the cables should be of the same quality and have the same attachment.
Car starters draw 400-600 Amps at 12ish Volts which is about 6000 Watts, but I understand your point, the design should have been far more robust.
honestly, i've considered desoldering the PCIE5 port and directly soldering my cable to the board to completely eliminate the problem... but, how practical is that for everyone to be doing? most people can barely figure out a jumper let alone how to solder correctly.
Fun fact, i recently built a 3D printer, and the wire harness that attaches to the pcb on the toolhead uses the same family of small connectors (just a 14 pin instead of 12 pin). My first thought was like "thank god i don't have to pull 600w on this one"
I honestly think it's over speced 600 watt connector should be 450-460watts max current
Well if you consider its rated at sustained 600W and „safe“ up until 720W
@@1NIGHTMAREGAMER Well is usually only uses 450W or less, since many 4090s have biosses that don't even support 500w or higher...
@@gucky4717 i defs would not be pulling 600W thru that connector. too small. even 450W is pushing it IMO.
I don't know much about electricity, But as soon as I started hearing rumors years ago that Nvidia was planning to replace the huge 3x 8 pins with that puny little 12v connector, I knew it wasn't going to end well. I never expected to actually be right because I don't know anything about electricity and I assumed Nvidia engineers knew more about electricity than I do, but it is clear that they don't.
I undervolted my 4090 and placed a thermocouple on the connector. I don't want to take any risks. (If the temps goes above a set point the PC shuts down)
That’s sad spent so much to do this
NOBODY should have to do that to stop a GPU from melting.
Both ends?
@@SiAnon I always undervolt my GPUs but yes nobody should be worried about a connector melting..
😂😂😂. Dam that extreme
whats brand is 70% of the ones that are melting?
Last time I checked, northridge fix said it was NVIDIA issue instead of connector issue.
Yup but Nvidia fanbois not gonna like that.
I also appreciate the coverage Jay. I have the 4090 Strix water cooled as well. Mine just got toasted after a month of use despite constant paranoia and checking on proper seating even after having it all hooked up and running. I've also vertical mounted with a perfect flat run out to the back of the case due to extreme caution. Fried on the 6 12V pin side and non-ground side. I'm also seeing new issues with RAM code failures on CPU post the crash from the melted card. Am still troubleshooting unfortunately.
i wonder what the resistance values do when you wiggle the connector. id put leads on either end, test for resistance, and wiggle the thing. i bet the numbers would be interesting
Was thinking I'd like to do that, but don't have a 4090 or a Cable Mod adapter. Would be a 4-wire resistance test and would need to know the currents involved (from the specification), but would be easy to record the change and look at the worst case power dissipation in the connector.
it would probably look fine on a normal multimeter if there weren't high current.
If you connect a dummy load to it you could easily measure Vdrop when wiggling at any in/out pair of cables
Glad I found this video. My connector just melted.
Hey Jay! I just bought an aorus 4090 master and I'm fairly concerned about the connector melting. i have an o11 dynamic XL and the cable is up against the side panel. i just built it and am running it with the side panel off atm and am wondering if its fine to put it back on? Thanks in advance
What they almost need is some sort of clamp that goes around both halves of the connector and keeps them from moving at all, since this is going to continue to be an issue until the connectors are redesigned.
Or, hear me out, a physically bigger connector.
@@coccoborg a physically bigger connector with single clip in the middle made the same way will have even bigger problem. A small gap on 1 side will be even bigger gap on other side, simple geometry.
Older types od wide connectors had 2 clips on sides which render them "unwiggleable". It is as simple as that.
Respect to cable mod for honoring the warranty. Will continue to buy cablemod cables
me too, that's really impressive if that's true. Brands like Asus will just do the general "fuck you" move when it's their fault. Like that infamous incident with their Dominus Extreme mainboard. Not just Asus, other brands too. Kudos to CableMod. But you still have to ship your card to them and wait
Big shout out to CableMod. if they handle customers this good, i have no problems paying the premium price they have on their products. Well done.
I haven't seen and direct soldered cards. Are there any? How have they faired if there are?
Thank you for the work you do!
I immediately noticed on my adapter when I got it, that also the plastic part from the plug moves on the PCB of the adapter. You can rotate it a litte. It is not fully attached to the PCB of the adapter.
I have no idea of how much power goes through that plug but a loose wire in your house can arc and arcing causes it to heat up. (like arc welding, it will melt steel) so is there enough power to melt plastic.....I don't know . just asking.
What was the reason to change connector on videocard?
With all that wiggle room, it could potentially wiggle itself out from vibrations too. BeQuiet has had one (that I’ve seen) incident where it melted in the PSU side also, and they are takin* it seriously too. But yeah, someone fucked up when designing this…
lets not forget Nvidia fucked up using it
And the worst part: it didn't even need redesigning. And even if we really needed a new plug, why not just use one of the dozens exiting, proven power plug designs...
@@stephanweinberger Yep, fully agree. I like the larger connector. Thick, solid pins. I guess they needed to make it "smart" with the sensing stuff. But could've retrofitted that too like they did with the extra 2 ground pins.
I made it very clear on multiple occasions that cablemod as far as i can tell has nothing to do with melting connectors since the melting is happening on the graphics card side and not cablemod side.
@JayzTwoCents
how many heat--cycles before it unseats itself from expansion/contraction even if you are very careful? or every time you move your case?
u mean everytime the connector is heated, the heat pushes the connector out? wouldnt the clip on the connector prevent this to happen?
What monitor you using in the world of warships ad?
Hey Jay,
I've been saying this for a long time. Aesthetics aside, with these cards running this much power, they need to use kind of attachment that we had back in the day with 2 screws, one on either side. Think the old VGA attachment method. This locks it in. no wiggle, no movement of any kind and they could have made it look fine astatically, so it's certainly doable and will fix the problem. Nvidia needed to put the screw holes in at the factory, as I don't think making an aftermarket adaptor is possible. I realize the I have the benefit of hindsight, but Nvidia, as big as they are, should have seen this coming a mile away. (and probably did)
We have friction lock connectors capable of handling higher amperage requirements than this. This is just a poor implementation period. There are even more compact friction fit connectors that are more reliable and capable of providing this kind of current. They wouldn't look as pretty, but they would work. Take a look at the XT90 for example, ceiling for that connector is 90A which is DOUBLE the amperage requirement of the 4090. The two problems with it is that it is a solder pot termination, and it takes two very large gauge wire to handle the power delivery. That connector has NO retention mechanism, and there is NEVER a problem with a fully seated XT90 connector coming loose or being able to "walk" sideways like these connectors can and do.
@@ZeroX252 1.50 € for a power connector is probably too expensive for a 1650 € graphics card, the investors of nVidia would complain. I will not buy such poorly designed garbage. It has been shown that the RTX 4090 would run more efficiently with a lower power target, so all the "effort" with 12VHPWR could have been avoided even.
This seems like an easy fix on the male end of the connector. You just make the latch tolerance tigher so it can't back out at all or increase the thickness of the posts so they don't go in or out as easily.
Longer pins would stop it air gapping when it torques. Rather than one brand of card, I would almost guarantee that the connector manufacturer is the same. Since metal is expensive, they probably cut costs by using the shortest pins possible.
@@piscikeeper No, there are multiple vendors providing the microfit female socket, asus specifically uses an inverted design.
im just wondering is it stock 4090s are the only ones melting or the ones with water cooling as well. maybe just heatbuild up which lead to thermal runaway
Putting two clips in the connector would require also Nvidia or OEMs to change the card's connectors, which i believe they are not willing to do. One possible solution would be to make the clip shorter and wider or make the connectors inside longer.
I really like that bright yellow highlighter color 2 tone 12VHP plug one of the brands was showing at computex. doesn't fix the geometry of the design but visual clarity is a huge plus.
Cablemod would do well to copy that idea.
Measure the plastic part of the connector's dimensions/tolerances vs PCIE SIG spec.
Either the spec is too loose, or the plastic itself is out of spec.
I assume the spec would define a maximum of something like 0.6mm of play when the clip is engaged. That looks closer to 1.2mm.
I only found Intel PSU design guide - "ATX12VO (12V Only) Desktop Power Supply Design Guide" (I couldn't find PCIE SIG spec without membership)
The Intel design guide doesn't specify the clip dimensions, but I can guess the official PCIE SIG spec will mention something about it, because even the Intel guide refers to it in some places.
It does however mention dimension tolerances ±0.2 mm and ±0.25 mm max, and that the latch has to withstand 45+ N (4.5 kg) of force when pulled axially.
What Jay shows here looks completely wrong.
The connector looks so loose it would wiggle just by blowing at it, so missed tolerances? (my Gigabyte RTX 4090 + FSP PSU was *really* hard to plug, my fingers were hurt from pushing it!)
And I highly doubt the "withstand 45 N of pull force" allows for this big disconnection while still latched.
I would say the connector itself is not a bad idea, from the electrical standpoint it should be completely fine. But IMO it has really tight tolerances thus bigger margin for error. And we all know how manufacturers like to cheap out here and there.
Maybe the PCIE SIG spec, is what is at fault. Maybe it was a poor design for a power adapter, from the start, and should never have been used.
@@LaXi0rCZ Tolerances shouldn't be so fine, that they are no longer idiot proof. Shit happens, and when it does, it can start a fire.
@@gamewizard1760 I'm more implying that the plastic connector shroud is made out of spec. It's a 10c part.
Might be mixed manufacture - companies aim for mid tolerance on their tooling, but it can turn out to be slightly above or below (but still within the component tolerance). Say manufacturer 1 are at one end of the tolerance and manufacturer 2 are at the other. Both are fine with their own plug and socket interface, but when mixing the plug from one and the socket from the other, it falls outside of the intended design spec - even if both parts are "in spec".
So what do we have to start taking pictures/video of the cables now clicked into our GPUs when installed?
I think the clip system on the 12 pin cable needs to be re-adjusted or realignment so there will be less play and so the cable end won’t back off slightly from movement/vibrations, also the 6/8 pin connecters didn’t have this issue before and felt more solid
I have several cable mod cable kits over the last couple of years and they have always been of the highest quality with no issues and look great. It's nice to see a company go over and above what they need to do for consumers. I really appreciate that especially now with all the company's that won't.
I agree, that is why I am throwing business their way now and in the future even though their custom stuff is pricey. If they are supporting consumers by covering their replacement cost, they are single handedly holding up the enthusiast market. They are going way above and beyond and they deserve my support.
I was on the fence about getting a 4090 for a while, but end the end I decided to save money and get a PowerColor Red Devil 7900 XT. I am now SO GLAD I did that, since I don't have to worry about all this cable adapter/melting/problems going on. Between this and nVidia's scummy practices, I'm staying with AMD.
I'm planning on getting a 7900xtx, how have you felt it overall?
i wouldve gone with the xtx lol
@@fernandolezama3325 Great card nothing to complain about really. Performs as expected from benchmarks/reviews.
Only 2 things that don't often come up in benchmarks/reviews. if you have mismatched displays it could lock (not always btw) the vram at max clock so you can get ~80-130watt idle draw with 2+ displays. That was the case for me but fixed it swapping out an ancient display I was still using for just random stuff on the side (I have 4 displays all different brands and refreshrates sitting at ~45-55 watt idle). Oh and NVIDIA cards are not immune to this btw just never had it on my older 10th gen and below cards and overall you hear way less about it compared to AMD cards.
And if you are planning on playing VR a fair bit just Google around for a bit the optimization seems to be hit or miss on some games/headsets. Haven't looked much into the VR thing cuz I don't use it
@@GGbowz2101 "I have 4 displays all different brands and refreshrates" - You absolute heathen!!! This would probably drive me insane. The chaos...the disorder...the asymmetry 🤣🤣🤣 Just kidding, still better than producing just more e-waste and buying everything new but damn...this would took some time to get used to for me!
@@Gruftkriecher Yep it ain't the prettiest.. But they are all just my old main displays that rotate untill they are the 5th in line and get kicked out the door haha. And you know they are VESA mounted so I atleast don't have 4 different looking stands lmao.
I would just much rather spend the money used for matching displays on a PC or main display upgrade you know
The clip retention sloppiness is why I plug into the GPU last. I do all of my cable management BEFORE plugging into the GPU, because I don’t move that wire at all once I know it’s fully seated.
I've seen this with other Molex style connector/sockets/pins in the industry. I think what happens and what you are dancing around is that there is too much play in the connector. This can lead to a stretching (fatigue) of the socket (female contact) which then reduces it's ability to grasp the pin (male). This leads to less surface area coming in contact between pin-socket interface. Less surface area = less conductance (or more resistance if you please) leading to the mating area heating under high current conditions. While not as relevant if you don't have vibration or regular insertion/extractions, is fretting corrosion. Fretting corrosion occurs when small amounts of the mating surfaces scrape off due to rubbing against each other. These microscopic bits of material corrode easily, and oxidized metals are poor conductors (that's why gold is the "gold" standard for contacts due to it being nearly impossible under normal circumstances to corrode). These particles are trapped between the mating surfaces of the contacts and decrease the conductance over time. Another issue is the corroding of the surface of the contacts themselves. There are many things in a consumer environment that contribute to corrosion, including the consumer.
What I see as the problem with these connectors, and what you've demonstrated is the amount of play in the connector. The clip should be much wider than it is, and the latch on the card connector needs the width to match. This should help minimize the amount it can move side to side. The other issue is the latch on the card needs to be further down so that once the two connectors are mated, you can not pull them apart as far as you've demonstrated. The last point will also help with the side to side movement.
There are a couple other solutions. The first would be a completely redesign of the connector. That solution would be to increase the mating surface area. So either make the diameter of the pins larger, and the sockets to match, or increase the area by increasing the length of the pins and sockets.
If corrosion is a contributing factor in this failure, then change the metal used to make the contacts, something like brass.
I just built a system with the new plug. When I first looked at the pins on the connector I was shocked how small they are to carry 600 watts. I believe that's the whole problem.
The terminals are specced to carry more current than the microfit jr. terminals of the 24-pin atx, pcie etc.. But they're pushing so much current through there that the safety margins have become way smaller.
@@joeyverliesharennot to mention... If either end of the connector is out of spec, you can have poor pin-to-sockets connections. If the sockets don't grab tightly onto the pins, you'll get excess heat
Yup it is, if there was a larger margin of error, say the cable was designed for 800 watts, small issues wouldn't be a big deal because it would run at lower temps. Instead they have a cable running near it's limit so anything goes wrong and out comes the magic smoke, which never goes back in.
@@joeyverliesharenLess metal and smaller are specced for more power? They have to be making it up in design or tolerances. Seems like a better designed standard atx pin of the same size would be foolproof then.
@@jon4715smaller gauge wire and less material goes against physics. They are making it up because they want a small sexy adapter to make amd cards look like pigs because they use a “old” many connectors. Form over function, and people are defending this is unreal.
Perhaps a more longterm solution would be for cable mod to work out something with NRF And other places to get customer cards in and repaired at a discount, sure going without a card for awhile would suck but it could also build a potential case for a future lawsuit that could force companies to get together and fix this connector issue
Really good explanation fra Jay again.
I agree with the wondering of: How many times does this needd to be explained...
I just got my 4090 Suprim Liquid X.. what is the best recommendation to powering it safely to prevent something from melting?? I currently have Corsair's 1200x Shift that came with their one cable for their gen 5 cables.. It's firmly plugged in and the cable isnt bent much so not sure how likely it is to happen to my card.
The problem is that the new connectors are so small and dinky what even the slightest tension can unseat it.
Imagine if every plug coming off your PSU was like this.
Doesn't matter. As long as it's latched, the individual metal pins and sockets should still be seated deep enough
@@CM-xr9oq Bro just say you didn't watch the video
Exactly, everything would be burning. This cable is just straight up no suitable to be sold to the open market, at least not rated for 600 watts.
It's already happening. One person on Reddit got a PSU side melted connector. But the question is, why it's happening only with 4090?
@@nareshfree because only the 4090 has that much power draw when using all 4 of the 8-pin connectors
All i know is that they made a problem that wasn't a problem before that connector
So that's the evolution nowadays.
Don't make anything better, screw something up and try to fix it just to have something to say you do
I have a foolproof suggestion. Maybe the solution is to go back to the old-school approach. The cable connecting the monitor to the GPU used two thumb screws attached to a metal plate to secure the cable to the GPU. This method prevented ANY connection play. What do you think?
Or they could actually design with the fact that cables move in mind.
that would work, no more loose connection issues
They won't do it, it's a profit scheme at this point. They want customers to buy a 2nd and 3rd card when they can void warranty because "it's not their fault" in the eyes of the law.
the issue with that was tripping on a cord or pulling it to hard, you trip on one of those you destroy what its plugged into
@@big_bird8597 I mean if you trip on a GPU cable, I'd classify that as user error.
Jay, I'm glad I took your advice and went with a 3090ti instead of waiting for the 4090. Looks like I saved myself a bunch of worrying and cash. Thank you.
Doesn't the 3090 Ti use the same connector?
They aren't burning the connectors like the 4090s
You overspent on a lesser product, congratulations.
@@larkan511 You don't know how much he spent and 2k for a 4090 definitely isn't not overspending lol.
Wow, with all the 4090 problems surrounding both the connector burn and cracked pcb's. I highly recommend vertical mounting that card and plugging that connector in directly in the top with no 90deg adapter. That's what I did. The vertical bracket from Lian Li is solid and handles all the weight of the card without any stress on the pcb, and with the cable / connector clicked directly in to the top with it's weight pushing it in to the connector without any additional play or wiggle, I don't have to worry about any of that stuff.
I think the point is: The 12VHPWR connector is defective by design. That's it. And that was obvious to me since the beginning. The moment I saw the connector I knew that's not capable of handle all the power that is specified to handle. Or at least that the margins where too short. Everyone that tries to make products around that connector will have problems. I think that CableMod and the entire community should ask for a better solution to power those graphics cards. A more robust connector. I know that this would be a lot to ask from the manufacturers but I think that's the best possible solution. And the right one.
It's the connector retention design that's poor. Not the power delivery design. When operating normally, there is no issues. It's not a size issue or a gauge issue....but a 'solid connection' issue.
Like Jay said...if the retention clip was a bit shorter, there would have been no play. Maybe a bit wider would also help.
Idk, they have sold over 150k 4090/4080 cards alone. You would think if it was the plug there would be a lot more than a few hundred cards melting.
Ahhh yes you're such a genius man.
I'd say it's possible that due to the vibration of any moving parts, it could shake itself loose to the point the clip is the only thing keeping it attatched at all.
I cant seem to find them on the link web site.
Dont see a search bar.
Any help would be appreciated
Man id actually love to see Chris fix build a custom water loop with blinker fluid in it,
or at least see him make a tutorial on fixing a gpu lmao
i literally lol’d when i heard him say chris fix. 😂
Watch that be his next April Fool's special, lmao!
Eh, he's good at fixing vehicles... I don't know how good he'd be in this avenue...
cablemod being the good guy is awesome
Why the good guy? As can be seen in the video, their adapter is unreliable. They offer warranty, but it is not about money. You want to use such a product reliably, without RMA process. An RMA process makes you lose time, because you have to send in the card and wait for a replacement. Furthermore, the logistics involved leads to unnecessary global warming. nVidia has to do proper design, as well as PCI-SIG and Intel - period.
Props to cablemod, that is very above and beyond. I don't use them anymore due to the number of issues I've had with them.
I can confirm the great Cablemod support as i've had first hand experience. I had the new CM right angle adapter on my GB 4090 and it melted, both the CM adapter and the plug on the GPU. I reported it to CM. I sent them all the pics and proof of everything, shipped the GPU to Northridge fix etc. and they fully supported me, reimbursing me for the GPU, the original adapters I purchased from them and provided me with new ones which i'm using on the new 4090. Great support from CM. Although as Jay mentioned I'm concerned how long they can keep doing this. This is not even their fault, Nvidia needs to take accountability of their design flaw!! Who knows if this will happen with my new GPU, at that point maybe Cablemod won't be so generous, and I can't blame them.
My fear is a good company is going to held hostage by dishonest people just looking to get new hardware.
Same actually. My Asus 4090 and CM right angle adapter fused together this past weekend. It doesn't visually appear it melted, but I literally smell burnt plastic long after i killed the power. Were you able to separate your adapter and GPU?
@@Angelic257 yep, the first indication was instability, crashing, blue screening. Then started smelling burnt plastic. It did seem to be more difficult to remove the plug but after I did, noticed the burnt pins on the CM adapter and the gpu plug. I had started an RMA with Gigabyte but cancelled it after I realized cablemod was offering to cover the cost of everything, so was really impressed by that. If i had gone through GB, they'd probably give me the runaround and blame me or CM. Just hope this doesn't happen with the new msi suprim.
I'm not using one of CableMod's 90 or 180 adapters, but I am using their 12VHPWR cable. It was more readily available than the official Seasonic one, but the way they've been standing behind their product just made it an easy decision.
Would it not be an idea to glue the adapter clip into place on the GFX card side and simply disconnect and connect the cable to the adapter if we need to ever remove the card from the machine? Surely that would remove any chance of the interface coming loose at the card side connection?
gets my upvote for the funny world of warships ad... comedy gold
I agree with Jay. More brands need to step up to the plate, the way Cable Mod has, and truly support their products!
I think the most important thing here to think about is the failure rate. Most of us will NEVER have an issue regardless of what connector we use. Yes they are failing and it's good to know about but the internet has a way of blowing things out of proportion. Also, it's great that Cablemod has such a good warranty but it's too bad it's even necessary and I hope that the failure rate remains low enough for them to be able to continue on. Personally I have a 4080 and a 4070 so I don't have to worry about it lol.
"Here's how mine looks as of a couple of days ago" proceeds to not show a picture 😂
Is there a better way to secure it? Do I need heat resistant epoxy?
I think the base design of the connector is overlooking user interaction.
They should put a metal retain hook to have a more clicky noise when it engage to prevent user error of not pushing the plug enough and maybe shorten it to prevent that much play on a plug that have such small tolerance.
I mean, the plug have smaller pin, it mean the resistance to put it increase and make people think that it is seated when it is not fully seated.
The other option would be to put a plastic pin that would raise when you reach a sufficient depth so you have a visual indicator of it being set correctly.
MSI made connector's end in bright yellow. Meaning, that if you see yellow the plug is not inserted enough. Edit: it's kind of visual indicator you mentioned.
@@Mervinion Nice move from them, i mean, having a visual indicator that the position is not good should be done.
maybe cable mode could put some colour on their adaptor, but as jay shown, you have to really watch to see it, but another colour would help.
an easier solution would be this connector have two side locks instead of a single frontal one.
@@BrunodeSouzaLino Yes, or a larger one to help it prevent moving
@@Mervinion The problem is the play the connector has, the wiggle room. This problem cannot be resolved with color, the connector has to be redesigned. In many industries reliable power connectors are used. PCI-SIG and Intel just have to do some research, and be serious about the matter.
Adding anything will add more resistance thus more heat will be the result. It seems to be a design flaw in the undersized connector. (Contact surface area)
Yep. They should have used modified XT-90 with additional contacts for the power designation and also a locking mechanism that involves METAL and not plastic crap.
Hi J, is the GPU replacement in US only? I am in UK and curious if Cablemod will replace the 4090 if the connector melts?
As someone who's just bought myself a new prebuilt with a 4090 that'll be here soon, and worrying about this issue, I appreciate the videos from you and Gamers Nexus displaying this so I can be more confident that mine won't have that issue, as long as I make sure the thing's ACTUALLY seated properly, flush with the GPU, no gaps, not crooked.
Lets be real, the entire design of that plug is a disaster waiting to happen, why using 3 or even 4 8pins, when we can force 450-600Watts thru one small plug?
Yeah great idea, lets do that, what could possibly go wrong? 😉
IIRC Igor from Igor's Lab was involved in the Design of those adapters and he surely knows his stuff.
And when you are called CableMod and are selling alot of cable related stuff, you sure do as well.
So no wonder they stand by their product, as it is a solid and cool design, while also stayin cool as this is by design.
But replacing cards that used the adapter, holy moly that really goes above and beyond, RESPECT! 😇
The whole problem is that many 4090 cards are too wide to fit in many midtower cases...gigabyte recently came out with a version of the card that had the port in a different location on the card probably for this very reason XD. No one really would need a cable mod adapter if there was enough room for the one provided by Nvidia/board partners.
Such an arrogant & bad design decision to release a product that does not fit into the majority of cases.
i mean majority of cards already exceed ATX/PCI specs...
I route cables, then plug. :P
BTW, does CPU fan sucking air straight near upper connection of 180 degree adapter help in any way to keep connector temperature lower?
Can they not make the clip tighter so it does not un seat or a screw in lock that can´t unseat?
The question in my mind is, are these connectors and their sloppiness an actual manufacturing defect, or is it just the sloppiness of the standard? It looks to me like Jay said in the end of the video, if the latch had been just a little bit forward that there would be no issue. I seem to remember the molex connectors and other snap-in fittings on my older computers were really kind of a bitch to connect, and were even tougher to get loose again afterwards.
I understand the concept behind the 12VHPWR adapter, that it was supposed to make it easier to get 600W to the graphics card. But if it being not *100% absolutely unequivocally perfectly seated* is enough to frikin melt it then *the resistance is too damn high*
I mean, if it's melting, it's not making anything easier either, now is it?
Yup and it's crazy that people aren't calling this out as just straight up unsuitable for the open market. If something has to be installed 100% perfect it's not ready to be sold to general consumers.
@@bobbygetsbanned6049 People still fuck up installment of their graphics card, CPU, and RAM. Should they all be recalled since they have to be installed 100% too.
@@hirotakasugi4891 The difference between that and this 12VHPWR connector is that once the RAM, GPU and CPU are seated properly, they generally stay that way. Whereas with the 12VHPWR connector, it only takes some a small bit of shaking on the case / pulling on cables (i.e. during cable management or cleaning) for the connector to no longer make 100% contact. This is a fundamental design flaw.
@@hirotakasugi4891 Installing them wrong isn't a fire hazard.
I think a link to Cable Mod should have been placed in the description. I wanna buy from them now. Any brand that stands by their product deserves my money.
Is there the same problem with 4070ti or 4080 or because of the big amount of watts that passes throught the cable is only in 4090's?
Any idea what brand of 4090 is failing the most. Getting a good look at their card in a tear down would yield a lot of info hopefully
ASUS
Bullshit
When I saw that tiny connector, I was like "You have to draw 50A of current through that little thing? Really?"
Then this happens.
The material is sufficient for the circuit to flow unimpeded, the failures have nothing to do with the surface area as a whole, but lack of it when the user has not got a flush connection or is using an improperly made adapter like these cablemod clowns have produced.
@@Squall4Rinoa Yes, and no, if it was 90% copper used, or high-grade stuff - then sure (not even mentioning that we know that there is almost none resistance through nano-wires) but there is also 99% copper - just as soon as you consider that you know why this is an issue - 'c o s t c u t t i n g'.
@@michalwiktorow2188 Then I will cut costs and not buy a graphics card with 12VHPWR connector. These cards are the most expensive ones and cheap out at the power connector. Ridiculous!
I tried to wiggle the old style 8 pin cables in my old 1070 and it wiggles just as much , if not more then what is shown in this video with the 12 pin connector. The clip on that one was also not securing it tightly, but somehow those cable didn't have such issue, probably because they were rated for lower power and the tolerance on them was much higher to these sort of things then it is for the 12 pin connector.
The accidental mention of ChrisFix was hilarious haha.
I kinda like the idea of a slotted power delivery standard. I remember when I had a PCI-e X800 card that did not require external power, slot it and done.
PCIE can only deliver 75 watts.
How do you expect to push 600W trough the motherboard? Use your brain, please...
So… 75 watts?
Not with those thin traces on motherboards these days. They wouldn't even last half a second with that much power going through them.