Let's see if we can get this NVIDIA 4090 Cable to MELT!
Вставка
- Опубліковано 27 вер 2024
- Get your JayzTwoCents Merch Here! - www.jayztwocen...
○○○○○○ Items featured in this video available at Amazon ○○○○○○
► Amazon US - bit.ly/1meybOF
► Amazon UK - amzn.to/Zx813L
► Amazon Canada - amzn.to/1tl6vc6
••• Follow me on your favorite Social Media! •••
Facebook: / jayztwocents
Twitter: / jayztwocents
Instagram: / jayztwocents
SUBSCRIBE! bit.ly/sub2Jayz...
I have a hard time believing AMD decided to not use it now a month from release(at most). It must have been a design choice early on.
It 100% was but people are just looking for excuses to call this a dunk on Nvidia when it really has nothing to do with them. People really think AMD is last minute making AIBs completely change their power delivery a month out
It may be both. Suspicions that turned out to be truth later. And preparation made behind scenes "just in case"
They could have seen the plug and had a feeling it was going to be a problem later on, hence the reason they didn't use it. Heck Intel doesn't even use it on their new GPU's either.
Absolutely. If launch is next month, cards are already assembled and ready to ship. They didn't just decide right now to go "lol whoops that connector sucks, recall all the finished product waiting to go out, and redesign and remanufacture all the PCBs to remove it!"
They never planned to use it in the first place. This is only coincidentally a dunk on Nvidia because the connector turned out to be rubbish. If the connector had been fine, AMD might have been dunking on themselves by using the old connectors that some would see as outdated.
But of course, the AMD fanboys are being exhausting, because as far as exhausting fanbases go, AMD's is definitely one of the worst.
Yea it'd take a while to replace if they were already in production. Alternative they could announce a delayed date during the "launch" without us knowing any better x3
Hi Jay, you also should test it within the build. The adapter will touch the side panel in most cases and the side panel will apply quite some force on the terminal and contact points increasing contact resistance due to misalignment.
Put it in an SFF PC and watch it burn.
The 4090 is the size of a SFF
Exactly, while it's not needed in a build he should have tried to zip tie it down extremely bent to create constant pressure rather than just bending it for a second or 2
@@MrAllthatgoodstuff You can't put a graphics card in a PC that's smaller than the graphics card is.....
@@aaronthomas6155 they never said put it in a case where the gpu doesnt fit...
409p fe definitely fits in sff in some sff cases.
Few things to note:
-You're using an open bench, so there's less buildup of heat as would be found inside a case.
-You were only bending the wires at the end of the connector and NOT putting a side load on those connectors, which includes the pins. If you had a side panel pushing on the connector, that itself would apply some load to the connector, increasing the chances that it could start to run away.
The main issue you get melting in these connectors isn't due to the wire itself, but due to the resistance within the male and female pin not having enough surface contact to transfer that amount of current. Resistance is a function of the gauge of the wire and just as a thicker gauge wire can handle more load, less surface area connecting two pins together would mean it would reduce the load it can take before getting too hot due to resistance.
Once the resistance and heat builds up, it runs away as the heat increases the resistance therefore making it hotter, creating more resistance, etc. It effectively becomes a runaway fire hazard once it goes over a certain threshold.
The weakest part of all of these setups is usually the pins connectors. The wires generally have enough current capacity that this would almost never happen to the wire itself before it happened to the pins where it connects.
The pins are simply the weakest point in these situations due to their small size and also the ability for them to be bent/angled therefore reducing the surface contact between them.
A bit rambly, but just my thoughts on the situation.
Exactly, the problem is at the pins which can be deformed when mated and submitted to lateral stress, thus reducing contact and making the resistance higher, which causes heat 👍
Nailed it. Well put.
The issue with the connectors melting is most likely 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. This failure is most likely attributed to cheap connectors.
I'd like to see this post pinned to the top of the comments!! Great explanation, and I'm glad you did it because I was looking to see so I don't have to. Great point!
Does anyone remember how badly Micro USB did this in the first gen?
Lucky they werent pushing power or everyone would've been screwed 😂
I think it might be a good idea to recreate this in a case. as a case would have a warmer internal temperature.
you also have more constant stress or just less space in the case
i agreed
It's not the card that's melting. It's the power draw, and surface area of the contacts that's causing the melting, not ambient temperature. You need thicker contacts for more amperage. If the cable or contact delivering current is too small, it gets hot. This is why it's so important to plug in connectors firmly and correclty. If it's not making proper contact, it will get hot.
Steve had an issue with Linus the last time they tried to use a thermal camera through tempered glass lol
@@itnearu You can't just ignore the fact that a higher ambient temperature is going to have a factor. And for these tests to be accurate you need to recreate as close to as possible the same conditions as a normal user would have.
The glass transition temp of the nylon used in the connector is probably 80C. Once it reaches that temp, the connector loses its strength and the pins have the potential to move out of place causing further temp increases until the point where it catches fire.
Also keep in mind that the connector will get much hotter in a warm case than an open testbench
The spec for the connector is supposedly rated for up to 105'C - its entirely possible that materials being used in some cables are not up to the standard.
It's also glass fibre nylon, PA12. Same stuff I use in my 3d printer, it's glass transition temp is more like 130. I think all this heat is just related to the card conducting heat into the wires. The 1st test the card was entirely cold
@@SquishyThing it's actually slightly better than the stuff used in 3d printers because the fibre grain is much longer but I doubt it would deform at all until it's about to catch fire
@@infernaldaedra It depends on the filament, each one you buy has listed the length of the fibres. Some of them are even longer. Rn I have a carbon fibre one with 1 continus strand of CF that runs through the whole roll. Markforged do one similar with glass fibre. Most consumer grade use a very low % of CF or glass fibre as the nozzles wear out fast, mine has a hardened steel one so I usually buy the industrial stuff that's nearly 30% glass fibre
this, and more, like, if ppl bend, it streses the pins, it moves them, makes the connection surface less,
is the design of that small plug
At this rate 5 years from now we will be plugging our GPUs into the 220 line that we use in our basements to run our clothes dryers. "Hey man! I can't game right now. It's laundry day." "I gotta wait until my clothes are dry."
dry ur clothes using the sun, quick and simple, well, not quick, just simple
@@energetic0oak329 Cloudy days be like:
@@StickyKeys187 xD
nah, the cables are gonna come with fans pre-installed
@@StickyKeys187 ever hear of gravity?
Best way to figure out your side panel compatibility with any cable is to find out what your maximum air cooler height is for your case. Find out your GPU height and minus the maximum cooler height and you'll see how much cable bend room you'll have or to see if the CableMod 90 degree connector fits.
The problem is there aren't many cases wide enough...
@@StoneAgedGaming True all the O11 cases even struggle with the side panel closing and other cases don't have the space for GPU length. Like for example I can only go with under 345mm max length for my card.
Ok, using this method, for the O11 AIR MINI, the max GPU width while using the CableMod 90 degree adapter is 146.8mm.
You should try to bend the cable after you plug it and not before, that way the pins inside the connector may stick out of socket a little on one side and recreate the issue.
This
I was going to say the same thing.
Don't you know it's all for show?
Pretty sure it is a way to go.
1. Install GPU
2. Plug power cable
3. Press it with side panel
Either side force deforming socket loosens connection between pins or they are pulled away by cables.
@@iBolitN But you forget its Jay the gorilla, might aswell get a monkey to test it!!!!!!!!
This thing is absolutely unacceptable. How many 4090s are out there? We aren't talking millions, it's a $1600+ GPU released only two weeks ago - we're talking thousands units, tens of thousands at best. For such a small number of units and the fact it's not a professional equipment, 5-10 cases of burned connectors it already way too much. It's a fire hazard and it must be recalled and solved.
The heat problem is going to be tied to the surface area of the connection between the connector pins (and the conductor size themselves). As wire size goes down, resistance goes up and a video card trying to pull high amps through a connection that isn't making good contact will heat up fast.
It is bad connection at the connecting surface that causes overheating and fire hasard... In my van that also uses 12v and same 50A you normally use big lugs that is tighened down HARD with big solid bolts and nuts + thick vires. DC 12 and 24v is notorious for creating hot spots and fires if the contact is just a little bit bad. . .
@@a64738 ha! Good point, in my old caravan from 1983 it's exactly the same. If the voltage is down, amps go up for same to archive the same stuff. Cause power has to keep the same more or less.
At a first look i thought man, this is overkill in terms of wire thickness 1/4 and bolted down with screws to the fuseboxes and with very Bick conductors)
Learned alot about power and currents working with 12 V in the caravan and this stuff applies also to our beloved gaming PCs
that's an absurdly common criticism i've seen of the Ultra High Voltage Bullshit 69420 connector they're trying to push.
It's like everything you just said is grade school level electrical engineering.
Funny that.
Yup, one time we tried to jump start a car and the wire we were using was a bit thinner than it should've been, and god did that stuff get hot..lol, life of working at a salvage yard though, we were always doing random stuff.
How are those tiny pins ever gonna be sufficient, even with a good connection?
Like I mentioned in your last video, I don't think it's the bending of the cable but rather the rounding/ovaling/enlarging of the female side of the adapters and cables which can be done when plugging it in or bending the cables after it's been plugged in.
Yes, I agree. Especially if there is constant stress from being smashed by a side panel with heating and cooling cycles causing further deformation and travel of the contacts.
Jepp the wires are big enough in cross section to handle the current. The problem is the lead on female to male connection. There seems to be not enough material so the crossection becomes to bottleneck e.g. resistor which then heats up and or if the connection is only partial even sparks will happen. And the later seem to be corresponding with the photos of the plebbit post. A melting or partial melting would just reform the material.
Those CableMod cables look SOOOOOOOOO good. Thank you for all the work you do!
using an open test bench for this may be difficult to reproduce results. need to find out which cases are being used along with the other variables like jay was saying.
Ltt posted a video showing a 4090 pulling more wattage in cyberpunk 2077 than in stress tests most likely due to the added ray tracing. Could this make enough difference?
Was also seeing draws pretty much exceeding the ratings on some those psu's used, but are very top end, so I'm wondering if the ones that are failing are a combination of psu makes/watts, cable seating/bend, load type and, maybe even more importantly, just user error.
The Founders on an open bench is probably the least realistic test, but at least Jay tried, and asked for more info on what setups are having the issue.
According to igorsLAB the problem is inside the 12vhpwr connector itself, especially the solderpoints and far too thin bridges. You would have seen it after cutting that side open! Other than that the standard itself isn't the problem so most native 12vhpwr cables from power supplies seem to be safe (haven't looked into cablemod). So it lies inside the adapter only
I haven't seen cable mod use bridges before, so it looks as though there is only one problem manufacturer for now; nVidia.
Even cheap ass adaptor cables (third party cables like cable mod, but at "unbelievably" low prices, and with dodgy descriptions) could be more robust and safer than nvidia's.
I bet AMD will have a field day with this.
According to recent reports, this appears to be a narrative for damage control
@@LiveBenchmarks It's looking a lot more like the socket on the cards. I'm just glad l can't justify that much cash on a graphics card, so I can just watch the story slowly unfold.
This is likely also more of a long term issue: if that connector stays at 60 degrees for hours, it's going to become a little more flexible and eventually bend more and more and contact in the connector will worsen up to a point it really heats up and melts.
Honestly they should switch to higher voltages like 24V maybe even 36-48V. The amps we push through that 12V rail is getting a bit crazy. 500W at 12V is a whopping 40A, that's a lot for any connector to handle especially ones that just clips on like that.
Either that, or return to the good ol' 8pins. 150W/plug, need 4 for 600, ok yes it's massive, but it has been already done, so no big deal.
I'm guessing the GPU would need a step-down transformer though. and the psu would need to step up. could work in the future though, neat idea.
@@Born_Stellar would require slight changes to the VRM but it shouldn't be too crazy. Step-up/down transformers are in the AC realm, but we're in DC land so the exact same tech that's already on the card but with a higher input voltage. It already steps down the 12V to ~1V for the GPU core and memory and doesn't really use 12V for much if anything. Phones and laptops already use higher voltages: USB-C PD supports up to 20V/5A and we still have super thin devices. Higher voltages are generally easier to handle than higher currents so it makes more sense to convert a high voltage to a lower one closer to where it's being used to minimize losses and high current cabling. Which is why houses use 100-240V outlets at 15A.
The main reason we ended up with this connector is attempting to remain compatible with existing PSUs, which I imagine the new ATX12VO standard weights in as well.
@@maxpoulin64 USB C goes even to 120W, so more then 20/5. More like 24/5.
thank you for this video. I ordered the cablemod cable that ends with 4x 8 pin, the safest option before buying the actual card: 600w/4
For what I have seen, it is not the bent of the cable but the strain in the connector proper. Since the connector has sleeves of metal inside that have a seam, and open cut, if you put strain in the connector (not the cable) sideways (towards any of the shorter sides of the connector) the internal pins seem to open the sockets and then get hot. In sum: the problem here is not the cable but the connector, somewhat like the USB 3.0 internal header that tends to break, these tend to burn.
Try to make this test with the cable managed in a very tight way, so the cable is pulling from the connector sideways.
I think the most common "bend config" will be putting it in a regular atx case and trying to force the side panel back on over the adapter cable. If they don't know that much they also might not know how many fans to use in the case if they did manage to close it.
My opinion is that instead of aggressive bend but ultimately no tension on the connection from the wire, add some kind of tension from pulling the wire like some people might do when cable managing inside a case
Hi Jayz, there are one possible situation you might want to test. When build the system, it make sense for people not to pre-bend the adaptor cable, and let the side panel / cable management to form the bend, thinking it will give the biggest radius. However, doing that will put stress constantly at the connector, result as a potentially worse situation. In addition, when the bend is formed by pressing the side panel / stretching the cable behind the motherboard tray, the connector is not guarded by hand compare to the pre-bend method as you show in your video.
cablemod genuinely sounds like its going to be a lifesaver
Other than people who don't know better will think the cablemod cable is designed wrong because it won't allow a bend close enough to the card to allow a side panel, if the person wants to put a side panel on...ie,like normal.
@Shionne I think the easiest will be for them to make a proper stronger cable and them them out for free
I think the corsair one is fine too
NVidia is not thorough enough to check this, other companies have to do it for them.
Nvidia said if you use a cable mod cable it voids your warranty
One big oversight here.... Open test bench vs closed computer case.... Ambient temps will rise higher over time inside a sealed case. If your open test bench ambient is low 20's in celsius, and the ambient inside a case running a 4090 at full tilt for a few hours is say in the 40's celsius (or higher), then that will cause the connector to be way hotter. I dont think you will get it to fail with it on an open test bench.
I say try again, but next time put it in a mid or full tower tempered glass sides with the usual computer's fan configuration (not the perfect fan configuration that you would use because you have been building PCs for years).
or better still, everyone that reached out to you, get thier system spec, including fcooling system (number of fans and push or pull config) and then document each system and create a system in a similar case and repeat the test... Whats the bet your results couldnt be more polar different.
20:19 if that had happened on an open air test bench, then its likely this video wouldnt have been made as the recall would already be happening and your video would be about the recall instead.
The CableMod 12VHPWR cable (like Jay showed in this video) come in two types - terminating into 3 x PCIE or 4 x PCIE at the PSU end. I got the 4 x PCIE for my AX1200i.
RIP AX1200i - went through 4 of them
@@Don.Lamaack I had 3 of the AX1200’s pop in three months, eventually I asked Corsair to give me an upgrade to the AX1600 which they did, been solid ever since
Where'd you buy the cable from?
@@joophommie Cablemod. I ordered a 4 plug version yesterday for my 4090 and signed up for the 90° adapter that is coming out in 3 days time. A must have to eliminate any harsh bending of the cable and help reduce any potential damage. Cablemod cables aren't cheap. It cost me £65 to get the 4 plug 4090 custom coloured cable shipped to the UK. Worth every penny though if it saves your £2000+ GPU setting on fire.
Unless you don't plan to overclock I would get the four plug one for 600 watt I guess the only reason why you might not do that is if your PSU doesn't have enough plugs but then you really should probably invest in a new BSU
jay: wants to test under usual user conditions
also jay: tests on a 200$ test bench
Your right a user of a 1600$ card just hasn’t 200$ boards
@@domi06021988 More like no one uses test benches.
@@domi06021988 They put it in cases, not test benches, smartass
@@domi06021988 how many people do you know who use an open air setup?
@@Null_Experis if the comment make a point on open air I go with you but furgfury points the price out not the thermal difference to a case
Hey Jay, thanks for this, but: OPEN the 12PIN plug and see how the cables are soldered and how thin the metall of the plug is compered to the soldered cable, Igors Lab has made a video in german about it, search for it, please stay on that, thanks again :)
Thank you for showing us an attempt to recreate the failures that have been cropping up. Glad to hear AMD is not going this route with the connector, but I’m concerned since they probably already have the product standing by ready to ship. I think they have the product announcement for the RX7000 series set for next week?
They could have tested connector before this event and decided to go with previous one (because of some issue or another). So they are just promoting their use of previous connectors. (which they had planned all along)
Unlike what Jay said, they didn't decide last minute to switch connectors. Before the 4090 launch it was already known that the new cables/connectors were problematic and getting hot. Nvidia used it anyway. AMD decided not to go with it probably because it didn't satisfy their testing/standards.
There nowhere was information that AMD will use this pin. Jay says it because he assumed "new standard, so AMD for sure uses it"
This was fast
The Problem is NOT the plug itself. IgorsLAB made a great Video about that.
The cable that's provided with the cards is just really bad.
These really thick cables a soldered to very very thin soldering pads inside.
The cables in the middle are each soldered to two pads, but the outer ones are only soldered to one pad each. The pads literally just snap off by bending the thick cables.
That's why a dedicated cable doesnt have the same fault.
14:50 _Do not do this._ Different PSUs often have different pinouts. It isn't even safe to assume that PSUs from the same manufacturer will have the same pinout. Plugging a set of cables with a different pinout to the PSU can fry whatever you plug it into.
I can vouch for this, I had a 850w evga power supply then upgraded to a used rmx1000 Corsair ps but had no cables, tried the cables from the evga and nothing happened, came across a video about mixing cables so I bought the original cables from Corsair and everything worked fine, I was lucky this time, can’t say the same for everyone
@@bluberrialpha You can vouch for this but you've never had it happen to you, gotcha
You are correct about not swapping between PSU's; Jay should have mentioned this. Jay likely knows Corsair has been using their Type 4 cables for a long time now. The RM1000X I got 6-7 years ago uses Type 4 cables that are the same as the cables I received with an RM 850X I just got a couple weeks ago.
"We gotta try it" - same thing Nvdia said as they released cards with dangerous connectors
“I wonder what can go wrong”
Also when they tried to sell the 4080 12GB.
@Muscleman8562 *looking for anyone who cares*
Noone found!
its not the connector thats dangerous, its the bending of the cable that is dangerous...
cables and connectors have been melting due to overly aggressive bending for over 20 years.
You call it dangerous but nobody proofed that those cables are actually dangerous, yet. Just because some people have a burning problem we shouldn't immediately assume that this is actually a hardware error. Most of the time it is actually a user error. I myself work in a service team in a tech company and there are way more people doing stupid shit with their hardware than the hardware actually failing themselves.
I think a more accurate and probably higher failure rate would be to bend the cable WHILE it is plugged in, putting more pressure on the pins themselves which seems to be the point of failure. I understand you may not want to do that while plugged into a 4090 as you don’t want to damage the 4090’s connector but maybe plug it into something else. I think stress being put on the connector while plugged in is the main weakness of this connector here.
Try it in a case, the pressure from the case would slightly unseat the connector (ever so slight in theory) and that bit of space could cause arcing which in turn could cause the failure to happen more consistently. Working on airplanes and how they're wired I've learned that if something isn't FULLY seated it could melt.
Exactly, when just bending the cable and then connect it, the inner connectors "go back into place", however with the side panel against the bend, the connectors inside can't go back into place
I love how everybody is trying to rationalize this as being okay at this point. Like I said in another comment, I've already blown my computer budget on other stuff for the fall. I don't even play games much these days, I'm busy, and I had other autumn/winter things I can spend it on than a 5700XT that doesn't need to be replaced or replacing my monitor. A part of me thinks I probably shouldhave looked at ultrawides and 7800XTs first but let's face it I'm not gonna wait and I don't know AMD isn't going to do stupid shit because simps let nVidia normalize it like the connectors, prices, and absurd TDPs.
You should try a stabilty test for hours on a intensive amp draw like 600 wats for a few hours see if the connectors get hot or burn like others .
AMD moving away from this is such a smart move. Can't wait to see what they release and am hopefully they deliver as I'm itching for almost any excuse to move away from nVidia and Intel at this point.
AMD marketing has had it's great moments in the past, only their silicone and drivers held them back most of the time.
Before someone says it: Nope, I'm not hating AMD, I'm currently running a R7 3800X and if the market weren't that F'd up and the cards here basically unobtainable I'd have a RX 6900 in my PC.
AMD is making money purely by being a little bit smarter and a little less shitty than their competition
AMD has previously released a 500w Tdp GPU with only 2 8 pins. So I guess they are quite confident in good old PCIe 8 pin..
FSR and ray tracing on AMD are the main things keeping me on Nvidia because they aren’t even close to matching dlss 2 or nvidia’s ray tracing so it doesn’t matter what they show
my last two systems are AMD, I use one as my work PC and one as my home PC/simracing PC, no issues. My GPUs in both are nvidia though, purely due to best performance available at the time. I previously used an rx480 though and worked even more flawlessly honestly, drivers were an issue with the first gen ryzen stuff but got sorted. Memory stability is still a bit worse on AMD I believe, takes a bit more intelligence/knowledge to get your rams rated speed, or just run it a couple hundred megahertz lower and you'll have 0 issues. It's been fun learning the new BIOS and setting wicked CPU testing scores.
Tbh they should just use 3 normal PCIE connectors. If it would look the same as the motherboard connector, it would be just fine, maybe put it on the right side so it's out of sight
That's the whole thing, they're trying to solve a non-existing issue
yup, if it aint broke, why fix it (and potentially make it worse)
Exactly, making it smaller is fixing no issue.
Jay I'm a miner i came across this a few times, with a poor connection/poor contact causing sparks that melts the plastic, you need to fine a defective plug to simulate the melt down and the power supply won't shutdown till a short happens. As always it comes down to quality.
Ever thought about bad crimped contacts? Or a too lose connection between the metal parts, could be caused by bending, too? There are a lot of amps going through those thin cables and connectors, combined with small arcs or higher resistance, this is nuts!
@@tacticalcenter8658 wow, just wow! Check igor's Lab's latest video, they opened one of those connectors. They are soldered! Way too less metal as pins, just crazy
Yes there are "rail design" differences between PSUs but I would imagine every PSU's 8 pin socket (PSU side) would have at least 6 power connections, and two of these resulting in 6 connections perfectly sufficient for a... 12 pin plug....
Would like to see cable mod add 2 more 90° designs.
Going out lengthwise to the 12 pin with both latch orientations.
Having all 4 would cover almost any build with any card, and if the 12 pin does eventually get full integration with all the card manufacturers, will be damned near required of the psu and cable manufacturers.
They're working on. They're very active in Reddit rn posting blueprints and progress
@@cadenzproductions nice.
Open bench is a weird format to test for thermal failure in a typical desktop build.
Jay, the type 4 plug on corsair power supplies is designed to handle 300 watts each.
Also, the corsair PCIE cables are running 16AWG wire, and are designed to deliver 150 watts per connector.
I normally like Jay's episodes. But these are definitely NOT public service announcements or journalism. They're deliberate attempts to drive views, with an inherent willingness to talk crap about the electronics side of things.
He claims to be trying to help drive up consumer information. But he's seemingly very willing to Not inform himself on how power cabling Actually works and is Actually specced.
As if his ignorance is worth other people's expertise.
"i don't understand why" is NOT a rational for something being done "poorly".
Ignore ALL of his electronic "engineering" knowledge. It's spurious confirmation bias to drive a preffered narrative.
Might be variance in the adapter, e.g. uneven wire gauge or crimping issue. Also you should try to run it inside a not-completely-well ventilated (but reasonable) case. These things can add up.
ambient heat is not the cause of why the wire fries and melts. so that would not affect the experiment whatsoever. the experiment is about trying to make the pin get in contact with each others through bending and handling, lowering that pin resistance and let a surplus of ampere through it, causing the burn. it's not ambient heat making the cable and connection fry, it's the unusual amount of ampere that gets into the damaged connectors. It's ohm's first law. Study more, kiddos.
@@davidepannone6021 You must be a troll, because that made no sense.
@@karmakh you must be an idiot, because you don't know ohm's first law.
@@karmakh You need an education....
Maybe it’s a combination of bending the cable AND pressing the bent section against the edge of the case, causing it to pick up more heat, not dissipate that heat, and melt. The two people who had melted cables may not admit it was squashed against the case in order to try to get a refund from Nvidia
Even if this were the case it isn’t an excuse to pin the blame on the consumer. If a cable does melt from bending then there’s a problem.
I think you missing a situation of not fully plugged connector, if there is tiny bit of space between connector and pin it will arc at this amonu of power. It might be bad fit that causes it to melt, since arc will produce incredibly high temperature.
he did this in the video
@@morganlimes but he didn't do it for 3 days like the reddit poster said.
60c is still 40c over ambient, in a closed case, with a bunch of cables crumped together and basically no airflow, it could get much hotter. It would be interesting from a purely theoretical point of view to know what temperature the plastic melts at, to know if it's even achievable with just heat or if it must involve arcing or fakes.
Testing the plastic melting temperature should be easy with a heat gun. But still, 60c is pretty far from any plastic melting temp.
@@InfoDav Exactly, you can put plastic in a pot of boiling water and it won't even melt.
If this is ABS plastic which it likely is, the melting point is around 104°C
@@thunderbolt10031 it's not 104c at all, it'll be closer to 200c for abs, it'll get weak at 104c but would never melt.
Well ambient for me is 32°C to 35°C, so the connector would run in excess of 70°C give or take and even hotter inside a case ☠️
"I didn't want to damage any 4090s, but dammit I gotta try!"
That's our Jay!
I'm sure the RMA process will be a headache. Nvidia is going to blame your PSU, case config, or user error. Too many outs for them
They can eat shitt
I’d want to see this after 3 months being used in a case and see the state of it.
Ok let me get the time machine out
@@DR19X why is the need for that? One can create a video about it in three months time, rightv
There is major flaw in this test i noticed. Its on open air test bench, while users have them in closed cases that can have bad airflow. Connector is then surronded by heat produced by CPU and GPU. That may cause it to fail.
I think you're going to have to try it in a case. I think that if the cable was to touch the side panel, then that would add pressure to the connection that could cause higher temps. I think you're onto the right track with this cable and the issues it has though.
as he stated. the event he was trying to recreate was with a vertically mounted GPU so there was no case pressure on the connector at all.
I think 2 plugs is fine, like most PSUs have that connector where it terminates into 2 6+2 Pci-e connectors, each of which should handle 150W so that means the cable could, in theory, support 300W. So the Corsair cable which ends on 2 300W plugs for the PSU, should provide the 600W on the new plug needs to power the card.
Correct the pins on the PSU end are mini-fit jr and can support 9.2a, so that is in excess of 660W across two of them. The issue with using a pigtail/daisychained/dual output cable is the wire itself has to be able to support 300W at least along the portion between from the PSU to the first connector on the GPU. Even 18AWG wire will support that amount of power over the short 2' or so lengths used by PSUs. You just have to be careful with either custom cables or cheap power supplies that come with under gauge wires, I've never seen a reputable PSU come with cables that wouldn't support 300W over a pigtail 2 output cable.
@@jmarynicz Best to go ATX 3.0 PSU forget these cables the power delivery is probably the issue here u need smooth power one plug from ATX 3.0 PSU to the card don't trust these adapters or they would not be making ATX 3.0 PSU they know this was going to start melting not worth risking with new technology don't cut corners
@@fortnite360HZ the reason for the ATX 3.0 PSUs are for communication between the GPU and PSU and future products. A good quality Current gen PSU is fully capable of powering a 4090 no problem assuming it's connected with a good quality cable/adapter.
@@jmarynicz Its better To just go ATX 3.0 PSU Thats just me piece of mind Its one cable vs splitters or pig tails i dont trust adapters u really dont and i own 1600 watts corsair i wont do it
@@jmarynicz i think Nvidia Is gonna get rich if people start damaging 4090s This is not worth The headache
I have the win 101 case and the glass was right up against the bent cable and it was fine for over a week using the adapter from the box, Corsair 1000w psu with no Corsair cable. After seeing the burnt cable, I took the glass off and straightened the cable to make sure it does not happen. I been using my pny 4090 since day one as well
You should try it in a case with standard airflow, and with the glass side panel holding it bent.
Hey Jay, Galax says if you pull the connector slightly out of the GPU to where it's still powering the GPU but the cable isn't fully in the socket - then the connector will go up to 100c
Exactly. All the pictures I have seen show a melted plug that was not properly inserted into the socket. Maybe the socket was too loose or too tight, or maybe the retaining clip didn't engage. It's even possible that installing the side panel knocked it loose. Regardless, a poor connection between the cable and the socket caused the connection to heat up and melt.
BTW, this really can't be blamed on the new connector type. I've seen it happen with the 12v CPU plug as well. Typically this is user error.
Yea, but I say making that condition is a condition nobody should make when building their computers and or adding a 4090 to the existing build.. since its already being talked about and everyone is hearing about it, they should make the bend in the cable before plugging it into the graphics card and ensuring the connections to the graphics card are very very secure and a solid seat into the socket of the gpu.. that was a very aggressive bend Jay just did here.. and it was fine.. I myself think that the 90* angle coming directly out of the card, should force NVIDIA to create a 90* angle cable for their card, eliminating all the stress one could possibly put on the cable, clearly the cable company has already been looking at it and making them.. maybe Nvidia should buy a bunch of them cables and ship them with their graphics cards..
the unplugged part could even be arcing, explaining the melting.
@@tommyrottn The smaller size of the socket doesn't help, though. It makes it more fragile and reduces the margin of error for that to occur in the first place.
Actually power supplies from the SLI days (like my EVGA G2 850) has 4 dedicated 8 pin VGA connectors... of course those were also the days where a single GPU wasn't 600W
that cable-mod 90 should have an RGB little 50mm fan on it 😁
give the plug 60 connection cycles and then test it again
Jay, in the first (control) test, you had the AIO positioned so that it gave the connector active air cooling. In the following tests the AIO were moved. Might explain the temp differences. Next time I suggest a bit more controlled environment :)
Preferably inside the 3 of the most purchased cases in the last 5 years.
@@1950sAmericanFather agreed
Jay: "We're not going to try and set our card on fire because, honestly... they're too expensive..."
Also Jay: "Watch us try to set our card on fire!"
Surely the easiest fix for NVidia (& also for AMD) would be to stick to the ATX 3.0 power cable, BUT move its location to the back of the GPU (facing the front of the PC case) rather than having it on the side, which it has to be bent to fit in without pushing against the case side panel? It would also LOOK better there, as the power cable would not have to travel as far into the visible area of the case, especially if the power cable terminated in a hard 90 degree turn plug, so no bending is necessary. The smart communication of ATX 3.0 should not be thrown away by AMD - it is a step forward that should be kept as a standard, despite dodgy cheap cables, and poorly thought-out socket placement.
That's a bit of a tall order.
On most of the cards, the PCB is only two thirds of the length of the heatsink. You're proposing putting the connector on the end of the heatsink.
The last time I recall anyone adding a connector termination to the heatsink rather than pcb, it was glued and soldered and widely panned for disassembly shenanigans.
Well I have a Silverstone Raven. So that means the "back" (opposite the video connectors) faces the case fans. Any card with a power connector on that end could have a bend problem if it was too long. The NVIDIA 4090 [13.22 inches x 5.5 inches x 2.4 inches] power connector would fit with space for a bend. Same with the current "poorly thought out socket placement" AMD Vega card in the system just smaller.
its also in open air, that helps cool the cable.
Jay can you do a inside case test since it will run a lot hotter and a more real case problems ?
I love the frequent uploads so god damn much!!
@Muscleman8562 it's funny cause no one asked
its certainly not a heat issue .... its due to a poor fit/connection .. which then leads to a heat issue.
the old connectors being bigger give a more secure fit/ plus a larger surface area allows for better load distribution .
The power supply comes with PCI-e cables which terminate in *two* 8 pin connectors (as you've shown), each of which can handle up to 150W, so the PSU connector must be able to handle 300W. So terminating 600W cable in two of the PSU connectors should be fine.
This. There seems to be some failure to realize that the spec is wattage per connector, not per cable. each connector is rated for AT LEAST 150 watts, meaning the plug on the PSU for a pigtail cable is good for at least 300 watts. Likely far more, considering that the new plug has like half the contact surface and is rated for 100 watts per pin minimum.
no, it's not just the connector that has to handle the power... each strand too. each PCI-E cable is rated to handle 250W for standard gauge (that can handle between 6 to 9A). if you go over that, some strands will heat up so much that their sheath/insulation sleeves will melt.
if you apply too much pressure on the cable and pins, some could get disconnected, that putt more load on the other pins, creating spikes.
electricity is no joke. you have to have enough margin to take spikes into accounts too. it seems nvidia went too close to the specs this time.
@@HiPnautique is that a spec for pcie power cable that’s listed somewhere? The 16 AWG the connector is made for can handle an easy 10 amps over this sort of distance. Both the wire and the connectors of the old style can handle close to 400 watts from an 8 pin connector on the psu side.
@@johngaltline9933 the connector might accept 16 AWG, but all cables are not 16 AWG, unfortunately. At 16, they would be fine with 400W. But most are at 18. Which should carry 286 W exactly, but I always subtract 20% for safe margin (and something even 20% is not enough). Manufacturers trying to save a few cents/pennies on cables, and the consumers don't know about it until they find the burnt cables in their PC.
@@HiPnautique 18 AWG is rated for 16 amps max current, that’s 192 watts at 12 volts. On an 8 pin that’s 768 wats per cable, well more than double the 300 watts required by the spec to provide for two 150 watt 8 pin connectors.
Even using your numbers, the issue once again is not what the PSUs factory cables can take but rather if it is safe to use an aftermarket extension from two 8 pin connectors on the psu side. It is perfectly safe to do so, as long as the aftermarket cable is using at least 18 AWG as that configuration provides more than twice the current capacity that is needed.
You should test plug with forced bend like when it pushes itself against side panel.
Yes what's missing here is a force being exerted upon the cables. Just bending them is only half of the equation. The cable will be pulling away from the connector in many setups, either from the side panel or cable management.
I want to see the two communication pins that had the free wire bridged, see how that effects it. If they're exposed ends in the same sleeve there's a possibility that they were touching in the connectors and causing issues.
yeah soon as he revealed those and pointed out that they had exposed wire that was right where my mind went too, did they create a circuit that ended up pulling even more power and thus generating more heat, was there a potential short or exposed section that started melting the plastic, lots of possibilities for from a pair of wires that should not even exist
@@amythistxue1 They were data pins so they wouldn't carry much power, my guess is that if they bridged it maybe had the card think it was plugged into more rails and had more power available than it actually does.
Definitely try recreating the melting in a case next time, i think inside a mid tower case where there's a lot of heat in case is when the cable will burn up, too much heat in and out of the cable
I think the into the case is totally negligible because you need hundreds of °C in order to have the connector to burn.
ambient heat is not the cause of why the wire fries and melts. so that would not affect the experiment whatsoever. the experiment is about trying to make the pin get in contact with each others through bending and handling, lowering that pin resistance and let a surplus of ampere through it, causing the burn. it's not ambient heat making the cable and connection fry, it's the unusual amount of ampere that gets into the damaged connectors. It's ohm's first law. Study more, kiddos.
@@davidepannone6021 current.... Not Ampere Ampere is the amount. Schoolboy
The ambient heat would not change much. While it could maybe increase the temperature by maybe 5-10C it could also have reduced temps because of faster airflow
I'm liking the fix Cable Mods is working on. Looks good too.
I appreciate CableMod's effort, but the fact of the matter is you shouldn't need to buy 3rd party connectors to use a ~$1,800 video card in a standard PC case. NVIDIA really screwed up on the plug design for this thing.
@@LapsedMemory you don’t. The vast majority of people have had zero issues with their cables
@@LapsedMemory Keep in mind that your reply is based on, like, 3 confirmed melted cables in the world.
@@Wrigglevision 3 out of how many in total?
@@LapsedMemory Nvidia didn't design the plug. Somebody else did. Nvidia's screw up was using this instead of 3 or 4 of the old style 8 pin connectors.
I'm wondering if it may also be a manufacturing defect too, with only some cables failing, like a bad batch or something
Other techtubers have investigated and apparently they're using very cheap contact pins, which are not robust/reliable at all
Here's a small thing that I'd like you to know: Thinner wires and pigtails make it less likely for the connector to fail, not more likely. That's because of their higher resistance, which forces current balance between the individual pins of the connector and prevents the setup from violating the connector specs. It's got the highest chance of failure when all other wires and connectors in the system are extremely low resistance (thick wires, good connections) and the 12VHPWR connector is the only weak point in the power delivery. This will prevent "accidental" current balancing due to the wire resistance, which means that even small problems with the 12VHPWR's connection resistance will result in large current imbalances and therefore connector melting. (If one pin has a dodgy connection and therefore high resistance, the current will necessarily flow "around" it through the other good pins, overloading them in the process.)
You've got the best chance of seeing it actually fail by using the adapter together with the shortest cablemod PCIe cables you have (no pigtails), together with a PSU that only has a single 12V rail. That combination will turn any resistance variation of the connector's pins into a huge current imbalance.
Something to note as well, if you put the system into a not so well ventilated case, you would see temps rise well above 60C, especially if you were to run the card for for hours at a time. I do not have a 4090 adapter at hand but previous 6 to 8 pin PCI connector cables were rated for 80 to 95C max (not continuous though). Cables that plug directly into PSUs often are rated much higher like 135C but that does not mean that the connector on the other side can handle that as well. AFAIK most brand name PSUs are rated at or up to 50C. If the connector on the card is made to the same spec as on the PCB of the PSU, that could be a problem.
It's ok Jay, it was fun watching you try. You have the same card as me, you have the potential to get top 10 in Port Royal! Just gotta try!
Thanks for trying this! Mine never came close to melting, despite extreme overclocking.
I wonder if they didn't clip it in all the way.
I looked at some pics on Google regarding these accidents. Just a thought. High temps could temper the metal to dilate and shrink in a way that can facilitate electric arcs, therefore melting.
Glad to see something I suggested in a previous video that cablemod were already beginning the manufacture!
your cablemod plug caused alot of people grief lol
The problem is not the tighter bending radius of the cable, but rather the puny size of the connecting pins/metal contacts within the 12VHPWR plug & socket. Any misalignment due to the cable pressuring against the side cover or other components will reduce conductivity. This, in turn, creates additional heat. The situation worsens when the GPU pulls a lot of current from the PSU.
Remember that the temps within a case is also warmer than an open test bench so that might also have an effect
Exactly
doubt it
nah 5-10c isn't doing anything. if the connector is gonna melt, its gonna do it. we're talking somewhere around 250+ c at least for those melted pins.
Would testing it in a closed case be more likely to have it fail?
yup
switching to a smaller power supply doesn't stress the connector more. It only stresses the PSU more. Also, feeding fewer wires into the connector doesn't add stress to the connector at the video card, it increases the load the feed wires carry and as a result increases the loss through the wire. SO, switching to the smaller PSU and using the pig tailed wires actually reduced the load going through the connector at the video card.
For the best chance to melt the connector, you need to minimize your losses between the PSU and the pig tails. This means a beefy PSU that won't be stressed, and beefy power wires feeding into the pigtails. This transfers the bottle neck to the connector at the video card.
Any loss due to barely large enough PSU or under sized feed wires reduces the load transferred to the pigtail and there for reduces the load at the video card connector and as a result reduces your chances of replicating the melted connector....
You seemed to focus primarily on the 90 degree bend. The worse bend was supposed to be the sideways/horizontal/diagonal bend though. Wish you had done it that way because it seems more likely to shorten the length of connections on one side.
Look at Igors newest Video and english article there he tells you why this nvidia adapter is a piece of crap but the 12VHPWR done right is ok. Spoiler it isnt the Heat its the construction of this piece of crap.
give the cable a viking funeral.😁
the problem with this test is it was not a realistic scenario. what you need to do is to put it in a tight case and then bend the wires with a glass panel, that way the heat is trapped inside the wires
Ah according to Igor´s lab (see "The horror has a face") the bending seems to be the trick, since the Nvidia connector has a build-in breakpoint 😂So the question only seems to be if you are lucky or not when you bend it.
nope. He also saying that pins arent strong enough thus whole design is flawed
RTX 3000 owners must be extremely happy with their purchasing decision. The 4000 series are setting up to be a disaster. All AMD has to do at this point is release solid cards at a fair price that don’t catch on fire
16:40 and there we might have it?
Really hate it, that people have to be beta testers for such an expansive product.
Also solid tips, thanks man.
Thanks for doing this, people need testing and someone willing to take the risk to find out if it’s a common problem.
honestly i find his tests as PR by nvidia, he didn't test inside a case where there is more heat and the cable is constantly pressured against the side panel, not only that but he didn't test for long enough LIKE 2 days for example.
Has to be a defect or partially connected pins that cause the issue. A very easy thing to have happen
Good video. Would have been good to test out what if those 2 wires touched that were snipped. Maybe that has something to do with it?
The psu can overclock too by forcing a single rail as opposed to four rail with an overclock button included that fits in the io.
I use to manage over-heating melting connections, most of them are caused by dirty (oxydation) or used connection that cause micro electric-arc, that fact there is a LOT of power didn't help
Imagine a 3d file for protecting the cable, would be nice for those who have a 3d printer imo
You'd want a high-temp material. PLA (which is the most common 3d printing filament) would just melt
@@VelvetSanity yup ik, petg or abs (maybe even those carbon fiber infused ones which I can’t remember the name of) would be recommended but still a good thing to have
I bet carbon fiber would make it look really sharp
@@VelvetSanity yea, or some white pearl-ish colour, depends on the GPU manufacturer but it allows for more customization
@@playmoth I wonder if the carbon fiber filament comes in various colors
Hey Jay, IgorsLab put up a video about it shortly after yours (Edit: Well, if 8 hours can be counted as "shortly", it's basically been up for an hour) . TLDR, it basically shows that the wire in the adapter is so thick (14 AVG, I think, he said), it breaks off this super thin solder contact on both end sides of the plug when bent. This leads to the card sucking around 600W through only the middle cable and thus higher temps...
When you put it inside a case the avarage ambient temp helps to increase more.
I've been interested in seeing what AMD has to offer this gen even though I usually don't buy AMD. But now because of the connector thing (not gonna say connector-gate), and the crazy awesome leaks about performance, I'm really stoked to see what they have.
My experience through the years with AMD GPUs is that they always hype A LOT and never ever deliver anything that even matches Nvidia. I'm not even hopeful this year.
I wouldn’t get too excited if you are hoping the 7000 series will overtake the 40 series in performance. Highly unlikely. However, if AMD can offer a good alternative at a great price, it’s still gonna be a good option anyway.
As long as it’s somewhere between 30 series and 40 series in performance and not marked up in value (grossly at least) then I’m interested.
Two PCI-E cables can deliver 450 watts, because each one has three V+ cables + one sense pin. One V+ cable can deliver 75 watts. If you have six of them, then you end up with 450 watts. But if you use one sense pin of the PCI-E cables for power delivery you get an extra 75 watts and the MB socket can deliver at least 75 watts, too. So overall two PCI-E cables are just fine up to 600 watts.
The issue might lie in longer sessions with lower graphic demand but longer buildup of heat.
The problem with the heat will be from a poor connection with pins having less surface area in contact. This will cause a hot spot. It definitely will get worse over time so depending how bad the connection is it could happen in 1 hour with a really bad connection or take over a day of constant use.
What about putting some stress on the plugged in connector? That might be happening if you try to put your side panel on and it will probably put some stress on the pins or even pull the connector out slightly.
Is there any difference between this cord and the one for the 3090 ti besides the extra 4th cable???Because I got mine against the glass in my case but I don’t have to force my side panel on.