My RTX 4090 Melted After 18 Months.
ฝัง
- เผยแพร่เมื่อ 21 เม.ย. 2024
- I bought my RTX 4090 the day they dropped, October 12th 2022 and after over 18 months of use the 12VHPWR power connector melted on both sides, the GPU and power supply.
Music by Karl Casey @WhiteBatAudio
#nvidia #rtx4090 #12vhpwr - เกม
![[FULL EP.01] LOL: Last One Laughing Thailand Thailand หัวเราะทีหลังปังกว่า | Prime Thailand](http://i.ytimg.com/vi/t4KunBRKn3s/mqdefault.jpg)
Man that sucks. A premium gpu should not have these types of issues. Period.
Any gpu should not have these types of issues.
Everyone knows about the melting get wrekt
Period. For this reason, I can really only consider the XX70 and below cards. Living in Asia, the prices are insane and more importantly, warranty is a shit show. Choosing amd wouldn’t be ideal since I do 3D stuff and the support there isn’t very good. The gpu market now, truly sucks in so many aspects.
@@shaunbay2 For the same reason I didn't get 4090 instead I went for 8 pins 4070, no worries after that.
@@quintrapnell3605 its not that common compared how many 4090s are in the wolrd. i have my for 2 years and 2 months now. way are u salty ?
This is why I love the power delivery on my 7900xtx, triple 8-pin connectors will never have this happen.
Preach. My 7900XT been rock solid
My cpu 8 pin melted cuz it wasn’t properly connected. So it is not the connector but human error..
Never? Just google melt/burn pcie 8pin.
Now i’m worried about my 7900xtx. My triple 8-pin cord is very short so I have an extremely aggressive bend with a lot of tension. Now I think I need to find a longer cord.
Take care of your GPU. You don't want to lose your 800$ plus card just to save 20$ worth of cable. @@Nazgul265
The pin to pin connection in that design is horrible. Why on earth they thought that changing something that was working fine to begin with over to this horrible design was warranted or even intelligent is beyond me.
As DerBauer demonstrated, they took no provision for redundancy/extra power... 8pin is rated for 150W but can hold 220W+, and power gets split over 2 or 3 cables when you need high loads, so my 7900XTX won't ever meld with 3x8pin even if I were able do a crazy 600W OC
But for the 4090, it's 600W on 1 cable with no provision for holding any more power, so it ends up overheating and burning way more easily
@pascaldifolco4611 Right, it's almost as if the engineers all looked around at each other and said, "Hey, this current connection system is way too safe and over redundant. Let's make something terrible instead, wouldn't that be amusing!"
@@pascaldifolco4611I keep my 4090 limited to 60-65% power target, (DerBauer video: 4090 power target makes no sense)
@@greggreg2458 it's a $1600 product, you shouldn't have to undervolt/underpower your gpu and shit like this shouldn't even be happening even for regular users.
@@pascaldifolco4611 Funny because lack of provisioning is exactly the same thing nvidia started doing with VRAM amounts since a few years ago. They go out of their way to make sure you're getting just the bare minimum in terms of stuff that's quite and easy to futureproof or give it some space left, and sometimes even less than minimum
Surprised? Pushing all that power through that tiny connector and expect not to happen is madness. 12VHP connector is dumb.
It's not though it's designed for that exact purpose initial design did not take into account that users were not following proper practice or using it for semicommercial use cases I explain in my own post
he said the unplugging and replugging was just a theory but its documented paper from the PCISIG.... smh when people don't care about their money.
Man this connector is designed with failure in mind, because I don't understand what engineer made this and said it's fine, because it's not, as a electrician who works in automation and stuff this just doesn't seem right, what was ever wrong with 8pin pci-e power.
fun thing is the old one uses THICKER wires but are rated for "less" per wire... the PLUG/PINS are thicker too but rated for less as well... kinda curious
Man I am glad I also grabbed that 2 year accidental damage warranty from Micro Center too. I've had the card for a year now with no problems but it's always in the back of my mind. Glad you caught it quick!
Yeah Mirco center has good warranty that will immediately replace or refund your money back if they can’t fix it. If I were buying a 4090 I would be buying the extended warranty form Mirco Center.
Sucks to hear man. GamersNexus seriously needs to revisit this topic and properly inform people. They initially wrote it off as user error and I don't think they have ever redacted that statement. At this point it is clear as day that is more than just user error.
Steve is a hypocrite you are willing to call out others for mistakes yet can’t acknowledge your own.
GN has been a terrible source of news for the last few years. The other big one was the issue with the x3d chips, they presented an extremely rare issue that required multiple failures as something common. Yes it was an issue, but it required a cpu to fail in a very specific way, they had to reach out to the audience and buy broken cpus in order to find one to replicate it, then presented it as something that was happening to everyone. With the 4090, they simply went with what they could replicate over a few hours, many of these are taking months to fail, not a couple hours, so writing the whole issue off as user error because that was the only way to replicate it in time for a video is absolutely a dangerous thing, people think "oh yeah, it's in all the way, no way it could melt or burn my house down". Repair channels are reporting dozens to hundreds of 4090s with melted connectors are coming in every month, not plugging it in all the way can't be that common of a user error, and if the connector is that prone to user error, there is still a problem with the connector.
Not sure what you're talking about. From what I recall they replicated the issue and showed how the 12v cable was not well designed to give the user feedback that it was plugged in all the way or not. That little warning light saying if it was not plugged in all the way could be off even if it was not snug, which would cause the issues.
The only take away he gave to consumers was take extra precaution to ensure everything was plugged in snuggly, then they should be good.
initially? but there were like 3-4 more videos that tackled that issue, even covering igors lab that found out that connector doesn't even take back all the power and it returns via PCIE and distorts signals on your mobo while going back via other cables back to PSU
overall it is more or less like that: it can happen due to user error, there's no definitive indicator that connector is all the way in, half of the connectors have worse design inside plug that goes loose and metal to metal connection is reducing since it's not that snug (metal is some kind of "pincer" and can come apart), so mostly it's design problem, which can be in most cases alleviated by user being extra careful
other thing is safety factor, 12vhpwr has SF of 1.1, should be at least 1.25 and 8-pin has like 1.8, 8-pin will never melt since it's overengineered to that degree,
so for people who doesn't know what that is: 12vhpwr is 600W, max is 660W, should be 750W, 8-pin is 150W, max is 270W
this video doesn't have good info either, " I've heard about this that you can only plug it in and out 20 or 30 times although this has been more of a theory than something officially acknowledged officially even if you read the instruction on my new power supply it doesn't say anything about plugging" it is officially in cable technical specification, that was announced, it should be safe to use those 20-30 times before degrading
12x6 cable might be a little better, but both cables have same flaws, very low safety factor, new one with shorter sensor pins might prevent not plugging all the way, but if cable gets wore off, then it will fail all the same, most probably it can happen from excessive "wiggling" and insides would just loosen
@@1Grainer1 Northridgefix has illustrated that this has nothing to do with user error but with bad design. Cards with connectors seated perfectly are still burning up till this day. Cablemod was also unfortunately caught up in the middle of this all.
Sorry to hear that Terra, hopefully next gen doesn't have these issues
It's not user error, I bet that's a very small percentage. Mine died after a year, the PC isn't even in the same room as me nor do I ever touch anything on it. I made sure there were no bends, no gravity moving the cable over time, etc. Yet it burned after a year. I had a pretty popular thread on Reddit, maybe you've seen it. I posted proof of purchase and all that stuff. ASUS Germany offered a replacement or a full refund for the original price (which was overpriced at the time) so I ended up getting a Suprim X for that money (best air cooled 4090). Everything was done in a day, I received the money in like 2 days. Was some straightforward woman no bs just straight to business no blaming and no comment on the missing warranty sticker (I replace the paste on all GPUs)
The Suprim X has been doing great for 5 months or so. Still using the same PSU. I also went for a Corsair 600W 12PWR connector because the NVIDIA trash connector is what burned the previous 4090. Mine only burned on the GPU connector, nothing on the PSU side. That guy that repairs GPUs on TH-cam is getting 200+ 4090s per month with burned connectors. Clearly something is wrong there. I'm also gonna mention that I've delidded CPUs for years, have had many different custom watercooling loops and messed around with many things. I know how to plug in a damn cable. If anything, I'd break the connector off the card if I pushed it harder than I already did. This time I documented the whole process (angles, showing that it's fully plugged in, date etc) we'll see if it dies again, tho it will probably last until the 5090 release date anyway, hopefully they've resolved this issue with the next gen.
And remember, don't stress over this stupid GPU. If it burns, it burns. I don't give a shit. They'll replace it.
Lol yeah I know how to plug a cable too. On top of that I work with electricity and controls, low and high voltage. I know the importance of a tight connection. I think I remember a post like this on Reddit though, glad you got it take care of in a satisfying way. I actually didnt have the Micro Center warranty for the PSU so got on the phone with Asus because it was a TUFF 1000W model, they sent me a shipping label and they are going to see if PSU can be fixed and if not send a replacement. I asked the woman on the phone about the GPU since it was an Asus 4090 as well if they would've repaired it or send a replacement for a melted connector and she said they would've took care of it.
Did Asus try to give you the run around until you kicked the hornet nest on Reddit lol?
@@TerraWare I didn't tell them about the thread at all, I don't think they were aware of it. Some tech websites picked it up but it was after a few days. ASUS didn't blame me for it, they immediately tried to help. I was expecting a total mess but for once it was hassle free.
In EU you send the card to the store you bought it from and they deal with it, but I decided to just deal with ASUS directly myself to save time. I don't know what's wrong with these cards, I've never had any high end cards do this before. This shouldn't be a thing at all.
I still find it ridiculous that I was aware of this issue since the very beginning after reading many Reddit threads as it was happening in real time, I did my best to prevent it and it still happened. Then there's pictures of PC builds with people bending the hell out of the cables / pushed into glass panels but mine died? Perfectly plugged in, no bends, open case (test bench) no panels, nothing. Oh well. I'll make a thread on Reddit if it dies again.
@@mcgmgc Another issue I had with my original Tuff 4090 was that one of the two HDMI ports didn't work right. Screen would be flashing off and on in one port and worked fine on the other. Not sure if you experienced anything like that. Funny thing is I have a Strix 3080Ti which too has 2 HDMI ports and one of them sometimes randomly black screens.
My new TUFF 4090 replacement both HDMI's work just fine. 4K/120 thankfully. That's one thing I like about ASUS cards is they have 2 HDMI's. My Nitro+ 7900XTX has 2 HDMI's and DP ports
I'm pretty sure Gamers Nexus owe people an apology for the "it was user error".
It clearly wasn't. They gave Nvidia a free pass over this when the connector cards should've all been recalled.
Watch the NV cultists spin this into "it's nothing but user error" and squeal like banshees on any minor problem that'll come to an AMD card.
Nvidia pays him good buck for that.
If it is not user error, then what is it? Genius.
@@LordAngelus design of connector. This guy who is talking in video said he done everything right on his part.
@@LordAngelus Bad design? Brainlet.
If that's the case then every PSU with those 8pin will also need to be recalled.
Well,
This is why I sticked to 7900 xtx. No 12VHPWR connector no worries man :)
I have the white strix 4090 Its stays really cool. I've had my 12VHPWR replaced cause the connector started to melt. I have the corsair 600W cable to plugs into two 8-pins on the psu side. Really considering vertically mounting my gpu so the cable doesn't bend.
Were you using any program to overclock the card or raising the voltage target?
Well, I had an MSI Suprim X, which I used with the adapter until I got a Cablemod cable. Both of them worked fine. Sadly, I then got one of those Cablemod angled adapters and a few months later, while on D4, it melted. After following the reports on those, I would say 5% of melt was bad use of cable by owners (forcing in small cases, tight bends, treating them like 8 pins), 90% down to those Cablemod adapters and a new one, 5% with the sense pins, although this one is the one that will be taking the lead now that the adapters had the recall.
I got a replacement card, a Gigabyte Master (I tried a few and bizarrely, this has been the best), and then bought an ATX 3 Asus PSU so I could use the native 16pin/12VHPWR, thinking this would be the best way. The Asus cable that came with it looks nice, but it did feel very thing and not robust. After about 3 months, I started to have (once a month) a black screen and fans to 100%. I thought "Oh here we go again" but after reading its an issue that has been around for decades, but happens a lot more with the new 16pin because the sense pins are not great. Sure enough, its such a big deal that Cablemod have made a cable that bypasses the sense pin to stop the black screen and 100% fans and after a few months of testing, it works.
The interesting thing is, the sense pins on the stealthcable are the same as my original Cablemod 4 to 1 cable, so that's why that cable worked great, and if your using the Nvidia adapter, it does not bother with the 16pin sense layout so it avoids the problem.
It seems that things are causing the melt.
Tight bend.
Plugging it in and out (this is why Nivida said cable was good for 20 times)
16pin cables that have loose sense pin cables (some you can move with easily, some are really wrapped up)
Not pushed into card fully and not pushed into PSU fully.
Using the Cablemod angled adapter (this melted like 90% of what I have seen online. 100s upon 100s in forums every month until the recall)
I do think that the sense pins are now the biggest issue because they can falsely report draws, which could make your GPU/freakout, or worse, perhaps ask for power that is not needed. What adds to that is so many are worried about their cables that they remove them to check, but everytime you do that you could be loosening the sense pins, this is another reason why this new stealthcable is a good workaround.
Yeah the 20 times in and out thing is dumb. They should've made the plug bigger, with bigger pins. Same as the standard PCIe pins.
You'd want a revision of something to be an improvement on all regards. Its what I think.
@@TerraWare
The 4090 feels like a Steam "Early Access" game lol
I do think the 5090 will have way, way less issues. The new 16-pin that will follow will be more robust, and the pin layout should help. On top of that, I think owner use will improve, although that was just a small part of it. Its funny to think that by checking our cables we might be making things worse. If, however, it does still happen in large numbers, AMD really needs to go for broke and try to win over unhappy Nvidia owners.
Scary to hear man hope my aero OC never dies like this i really dont wanna deall with all this u went through..
Well just make sure your cables are tight and no aggressive bends near the end of the plug. It's really all you can do. I did unplug my card quite a few times for the type of content I make so that couldve been a contributing factor. It just sucks that its something you have to be concerned about but as a regular user you should be ok.
You didnt expect to make this video and i definitely didnt expect to see it too😅 Man it sucks, at least you got it replaced quickly, hopefully this will be the last event for your 4090, hopefully nvidia will improve its new power connector on upcoming rtx 50 series.
Good that you got it sorted out and replaced. I'm not impressed with this new plug-ins reliability.
Yeah it was definitely not ready for release to the public imo.
The 12vhpwr cable is rated for only 30 plug/unplug cycles iirc.
Friend of mine had a 4090 and fixated the cable as tight as possible with extra cable ties to the GPU so it would not come lose at all. Well.. the cable fused with the GPU and there still was no gap at any angle, it was as flush as it could have been and so it also got fused. The cable ties did not touch the wires, only the plug. And the cable was not bent, so stress from improper bending was not the case. In fact it was similarly mounted like yous. He also never unplugged the GPU. 12VHPWR is just a mess and doesn't hold reliably up to its specs.
Boy, this is stressful. I need to check the connector
glad i have my 4090 undervolted still no issues yet
Did you overclock the gpu? Or have a gpu holder?
having fire hazard at home .. no way 40 series will find a place in my house
For my 4090, I kept my power limit to 80% and OC the memory clock to +1500 and set the curve editor freq to 2650mhz at 910mV
That's so unfortunate. I also own a TUF 4090 and I hope it's not gonna fail. It's powered by Corsair HX1200, and a direct CableMod 12vhpwr -> 4x PCI-E (on PSU side). I also agree that the TUF is a great card, mine specifically has massive coil whine
I've had my 4090 since it was released and had to order a 12vhpwr adapter cable from bequiet because the adapter cable that Nvidia included with the graphics card wasn't even properly degraded and didn't even go into the 12vhpwr port on the graphics card and I didn't have one Power supply with a 12vhpwr connection. I still don't have a 12vhpwr power supply, but with the adapter cable from bequiet in combination with my now slightly older bequiet Dark Power Pro 11 1000w Platinum power supply, I haven't had any problems yet.
So with burnt connector, they can rma? I thought they never did that
it seems there's a very tiny gap of the connector on both side from your new connection, it sucks if such a tiny gap did fry it up, and it's a bad design when it need such stupidly high force to just kept it from melting
What was the PSU model you had in the moment of melt ? Also did you use original cable from the PSU or it was a non brand cable ?
Asus Tuff 1000 watt. Used the cable that came with psu
Damn, that sucks. I just went to see how the cable of mine is, and thankfully it's still looking brand new. My 4090 is the INNO3D ICHILL X3 one, and it's about 16 months old now. I did do an undervolt on it since day 1 though, as in stock settings it was sometimes pulling 400W and getting my case pretty toasty, and with the undervolt it very rarely goes above 320W. Hope it stays alive for a long time lol, at least by the time the 6090 releases 😅
Yea I hope so too. I'm sure I didn't help my case with removing the card to replace it with the 7900XTX for my videos. I'm assuming thats something you dont do so as long as you dont disturb it should be fine.
But I'll tell you something I probably shouldve mention in the video. That burning fabric smell I smelled before finding it was burning. I smelled that same smell 2 nights before but my neighbor had a fire going so I thought it was the neighbors fire and maybe it was but I checked my PC and at the time it looked fine.
I say this in case you smell something funny check and make sure everything's alright. It sucks we even have to worry about this.
@@TerraWare Yeah, I'm definetily going to increase my paranoia as soon as I smell anything toasty lol. I've only unplugged the GPU one time since I bought it to upgrade my CPU
I'm surprised it doesn't happen more often. The leads in these connectors don't seem thick enough to handle the current.
Have you thought about undervolting your 4090?
I got a cooler master v750 ATX 3.0 for a 4070Ti Super (limited by case, 9L SFF Build). Cooler Master added instructions of NOT UNPLUGGING THE 12VHPWR NATIVE CABLE MORE THAN 10 TIMES on their "how not to bend the connector" sheet. Cooler Master has a 90 bend at the connector out of the box, going to send them an email to ask them about that tomorrow. Seeing all these melting connectors, it made me worried if I should go for a 4070 instead, or maybe I can just undervolt the 4070Ti Super from 285W to 230W. But still, I absolutely hate this new connector thingy. I have to buy a 2kg CO2 fire extinguisher and put it next to the pc. I would have gone to AMD if I can but no AMD new GPUs that are 7800XT or above, are under 250mm in length, so I am kinda conflicted right now.
I was going to get the 4090 but stopped bevause of this. Now getting the 4080 Super plus upping the CPU and Ram and still saving a nice chunk of money.
why did they had to put this connector also in the 4070 super since it only pulls out 220 watt?
eventhe non-super also has it
The socket/unsocket cycle life is an official ATX standard the same as cable bends and so on and they have always been there and 8pin was the same specs there, this goes back to molex days and before if you're old enough, Intel publish the specs by memory. Seen plenty of 8pins burnt out and personally think users are more the issue than the plug as they ignored all these cable management standards for so long that many are now making mistakes with either socketing too many times be it for making videos or to check the condition, overly ""managing"" cables AKA bending the snot out of stuff for looks and it has been so long since I seen anyone make a custom brace in a case or anything to add in a clip for a strain relief or anything like that. It used to be common 20yrs ago amongst the DIY build community to take special interest in actual cable management for practical purposes but now coiling all your power and signal cables together with a zip tie under the desk is the in thing....
Best thing to do if you're going to be swapping in and out high power cards is to get some spare cables and follow the guidelines for the plugs and toss them after the cycle limit and inspect both ends each time as the male ends will eventually wear too and have a tolerance you can measure 🖖
Yeah you could be right which is why I mentioned my use case and be transparent. Now the spec may be the same for 8pins but I have swapped those a ton as well. I think the 12vhpwr is more prone to failure because of it's design.
This much is obvious based of what we've been seeing. The 8 pin is more robust.
Did you have it power limited to 70%? I am worried about this happening, so I have it running at 70% power, and it's about 5% performance drop at most.
No 100%
Out of interest were you running the card OC / Stock or Power Limited to 350w or lower?
Stock and I often use framerate limiters as I prefer a more consistent frametime/1% low. The game I was playing at the time, the one on this video, No Rest for the Wicked, had the GPU fully maxed out pulling 370 watts.
The only time I have run the 4090 overclocked has been for a few 4090 vs 7900XTX videos on my channel.
@@TerraWare I was just curious of the avg wattage levels when gaming. To be honest at 4K@120hz vsync/framecap can depending on game still pull over 450w. I have had mine power limited to 350w since launch as I found beyond that the perf to power scaling was not great, maybe a 5-10fps increase with power and voltage sliders maxed out, better to frequency curve optimise with 350w being the target.
I think you are right the cause of this was most likely due to switching out the gpu's more than the average user would, but still shows that 12vhpwr is a flawed connector. Hopefully 12v-2x6 does resolve all potential causes of melt.
That sucks, if you didn't have an extended warranty prob be SOL. By the way, I beleive you even used the power cable that came with the power supply, nothing 3rd part?
That sucks man. I’m glad that this happened this way it could have been worse.
I agree a $2000 GPU card should be idiot proof and IMO should have amazing warranty OTB.
Question shouldn’t the power adapter also affect 4080 supers. Don’t they use the same adapter?
Yeah all RTX 4000 GPU's use the same adapter although there are some AIB 4060's and 4070's I think that use regular 8 pin connectors. I have seen a couple 4080's with melted connectors although they are very rare from what I know. It's 4090's that are most at risk probably because of the amount of power they can draw.
This is probably the 10th vid ive seen on the problems with power connectors on the 4090
Any card with this connector is a potential problem. Drawing more power than any other card will make the problem much more likely to present itself though, so 4090 should still be the most common.
Ive been dealing with random black screens and occasional black screen with gpu going to 100%. No event viewer info. Ended up finding the nvidia adapter was the issue. Was fully seated, no burns or signs of burning but just causing black screens. Replaced with a psu 12vhpwr cable and its perfect again. This is after a year of owning it
Oh wow. That's very interesting. I haven't experienced that.
I wonder if CableMod has any recourse now. You know, since they were blamed and had to recall all those connectors.
Glad you caught it early so there was no fire, and that you got a replacement fast and painless. I really hate this connector, I own a 4070 ti super, so I am probably safe but still it will be at the back of my mind. Having a gpu pull 400w is also a mistake by Nvidia that contributes to this in my opinion, not just the connector itself. 4090 would still have been a total beast and outperform everything if it was tuned by Nvidia to be a 350w card.
Yeah the 4070TI super should be good as it pulls around 300 watts depending on the model (Some 285, some little more)
Honestly Nivdia should have stuck with the 8 pin connectors for the 4000 series. Honestly I hope that one day they get sued for this.
I have attached a thermal sensor to the connector and then connected to my icue core, all lights red 55c shutdown 65c
I wonder if you limited the voltage to the card whether that would prevent the melting?
Hard to say. It could've prolonged the inevitable some more maybe.
@@TerraWare It seems as though the melting is caused by too much current going through the conductors, so if you were to limit that it should solve the issue. The problem is, you're deliberately gimping your card. On the other hand, I think the card usually maxes out a bit over 500W in terms of power draw. That probably varies from game-to-game, though. There's a rumor the 5000 series will use a new connector. Have you heard anything about that?
@@batman.darthmaul You're not wrong about the current, the 4090's draw more of it that has to pass through the connector. My use case differs from most people since I do make videos comparing cards, like the 4090 vs 7900XTX, I was unplugging the cable on the PSU side way more than an average user would but not on the GPU side.
I think what's contributing to it is how fewer wires there are in the 12VHPWR vs using 4 8-pin connectors which also have thicker terminals. I think over time, by unplugging but also just using the card, connector warms up when cards drawing full power, it cools off when it isn't. Over time the female part of the connectors expansion and contraction loosens up and increases resistance which = heat. Than it's a spiraling effect.
I think this is a contributing factor although speculative on my part.
As for the RTX 5000 I don't really know a lot about. I tend to not take leaks all that seriously and I don't go looking for them. I think this is definitely a big enough problem or a stain in how this connector is perceived by the enthusiast community that I dont think Nvidia will use the same or if they do it will be an updated one. I'm not sure.
FYI the reason you hear no/little coil whine in this clip is because the card is running at 2700mhz. Most cards are silent at that clock. It's when you let it run freely at 3000++ where the cards usually start to go crazy. Mine included. Quiet at 2750, very noticeable coil whine at 3000.
plus higher frame rates. my founders super was screaming some when i first got it when going above 120fps. Thankfully it broke in and I forgot that it had coil whine at the start.
Yea you definitely lucked out. I purposefully bought a 4080 to try and avoid this problem. I hope the 50 series avoids this issue of overheating for any of the cards.
4080 can also suffer from this, the connection is trash
@@bittripper3530 I haven't heard mass reports of the 4080 having the same issue as 4090.
@@marco12377 still happens
@@marco12377 I agree they are not as big a problem as the 4090 but Northwest repair has had them in with burnt connectors. Nvidia should have used two connectors (at least)
This is wild 4090 + intel cpu hard times this days blow up 💰
F.
The new connector is one big deterrence for me from buying RTX4000 series, other than the price.
Scary! I have this exact same GPU.
Make sure the cables are plugged in correctly and no extreme bends. It's all you can do really. I did swap my 4090 out for the 7900XTX for comparison videos which could've contributed to some extent. From now on I'll just do it at the adapter.
@@TerraWare Roger that!
You're the last person I would expect to encounter this issue taking into account how well you take care of your stuff since it's a near daily routine for you to shuffle around hardware. Only goes to show that it really is a quality issue at this point and Nvidia should have recalled the cards back at the beginning and re-release it with a normal pcie connector till they fix it because I do agree that with the power draw these cards have it's nice to have a single cable.
Btw does the pcie adapter help with the risk of it melting? Anyways as far as the 12vhpwr connector having a limited plug - unplug cycle it's probably bs unless you're yanking it around with brute force. The older pcie def doesn't have issues, I always took my rigs apart fairly often for cleaning and never ever had anything fail on me besides a backup ssd I had for storing movies/music.
Enjoy your new card man! Temps are absolutely amazing (nice to see the memory temps have been massively improved as well with the 40 series over the 30 series) and I'm a sucker for that cooler as well, It's the one I'm using for my 6800xt.
Well the PCIE adapter helps with the fact that I don't have to disconnect the 12vhpwr plug on the GPU or the PSU. If I want to use my 7900XTX for comparisons I can just unplug the the PCIE side of the adapter on the GPU remove it and use the 7900XTX and when done put the 4090 back in.
There probably is some truth with the plugging and unplugging of the 12vhpwr as the connection would become a bit looser over time which could lead to a less tight connection, higher resistance = high temperature in the connector.
The regular 8 pin PCIE connector is a tried and true method lol.
Crazy thing is you don't even need 4 8pin connectors on the 4090. 3 8pins is enough for my 7900XTX and that thing pulls more power than the 4090. So that's what I am doing. I already ran a bunch of 3D Mark benchmarks with the 12VHPWR cable and will now try the same with only 3 8pins with the power adapter. It will be the same I bet.
It's not quality issue, it's a design issue, they have revised it but still. Any and every 3rd party adaptor made eventually being recalled off the market. Ridgefix confirmed this, he fixed 200 cards a month
@@rzkrdn8650 I've heard that the plug has been revised and is in the new graphics cards. I'd like to know if mine falls in that category since my replacement is a brand new card.
The AD102 silicon is the newest revised A1 variant that's about all I know.
@@TerraWare Still won't save you. Don't bother to check. Your only best course is just lowering the power target (95% ish), plug it snug and pray. Or sell, haha.
I had a 4080 and using derbauer's adapter, it's what i did.
@@rzkrdn8650 I could return it save that money towards the 5090 lol. I have a 7900xtx
I have the 4090 Martrix comes with newer connector with 6 shunt resistors on the card to monitor the pins , any one of them acts up it shuts the card down instantly.
Is that the crazy ROG one with the AIO?
Best of luck with your replacement GPU. If you own a 4090, consider under-volting / reduce power limit it in MSI afterburner (kinda like what intel is recommending to keep their new k series cpus from exploding). Unfortunate to have to do this, but not there's any other way to keep this defective connector from ruining your experience.
This is why I skipped the 4000 series. It will be fixed for the 5000 atleast I think? But will see.
Damn that sucks bro.
I have the same exact card and I saw Northridge fix is repairing tons of them. I check it every so often but I fear it will eventually have this issue. I don't have a protection on mine since I got it from Amazon, but who knows what ASUS will do with these. I know they've turned down some RMAs for it and others were accepted. If it dies before 3 years of use I'll definitely be upset but there isn't much a consumer can do other than monitor it and be careful. I particularly like this card for the Dual HDMI option with the three display ports. On the 30 series several manufacturers had that option but for 40 series it was almost exclusively ASUS.
I don’t like ASUS because of their practices.
Thats why i usually pick hardware, that one step below top tier, cos top usually have some problems with heat/noise etc. Learned that a long time ago, with few burned top GPU's and same with CPUs, better buy 1 tier lower.. My 4080 doesnt eat a lot of energy, silent and stays 60~c even in heavy games.
That sucks. Hope your replacement doesn't do the same.
any problems like the RTX 4070? non-super and all that. just RTX 4070 with this connector?
No 4070/super should be fine. They don't draw a lot of power. Its mostly 4090s that are prone to this. Some 4080s maybe
@@TerraWare okay thank you!
I bet you could zip tie the cable to the card and still havw this issue
What if people would just undervolt the 4090 the heck out of it?
Like in Adrenalin (I have the RX 6600 and RX 6750 XT) one has lots of options to undervolt the card, reduce clock speeds to barely stable settings and reduce overall power delivery to the card. I had to fiddle with this since my UPS is only 350 Watts till it is overloaded and is basically to small for the RX 6750 XT - no problem full performance with the RX 6600. So now the RX 6750 XT just draws a bit more power than the RX 6600 and runs similar speeds. That 250 Watt card now draws just around 135 Watts.
Similar thing one could do with MSi Afterburner since it has slides as well to undervolt and restrict power the heck out of the card. I believe the slider goes down to only 85% of the power draw. That should reduce the power draw of the 4090 from 405 Watts to around 344 Watts.
You turn it to an 4080 or 4070 but hey you prevent burning your 4090 like that.
It should be stable still.
Never seen it happen on the psu side as well.
This is the reason I went for a 7900 XTX, especially sionce I live overseas and I bought it from the US. This means that I have to cover shipping myself if I want to claim warranty. With my luck, I'd get a burning one
And here's the worst part. Even if your psu and other parts are saved, and seller does you a solid honours a warranty - it will most likely happen again. If it can happen once, it is likely to happen again.
Not likely considering this is a different type of melting than the one the title suggest it is and is the first report I've seen of this kind of melting. This is likely either the cable got damaged at some point or some utterly insane OC, and lsat a possible defect but considering it lasted 18 months... This happened behind the connector at the cable. All the other melting incidents happen at the connector and the front side where the pins are, at that.
What a blunder NVidia made with this 12VH*Danger* crappy connector
Ive been having rog strix rtx 4090 oc for 6-7 months no problems no melting cable yet..... as fr psu have titanium rated rog thor 1600 w psu heres praying it stays alive😥
i expect my cards to last at least a decade with a handful of unplugs during maintenace, and no need to consider the connector being strained or 'mis seated'. this is some new trash that was rushed because business. meanwhile, nvidia is making news for another reason. i hope the next series or the one after that has less serious issues. 1 year stability does not mean anything, these cards are all compromised. i hope your replacement serves you well so you can 'upgrade'/replace it asap.
Always knew you liked BBC, Terra.
I would have guessed a year ago this problem is mostly due to people being overly cautious with plugging it in. But too many people still have the issue. Or course even a flawed cable or connector can survive with less power being run though it for much longer than one with same issues but with 4-500w for hours a day. Part of the reason I undervolt it. Only draws 200-250w and get over 100fps in the games I play mostly. It’s stupid tho to have to do this. Especially with such a high end card. Might just dump it and get a 5080. Use my old 3080 till it comes in.
I have an RTX 4080 with a new PSU and I use it about 13 months now but I don’t know that the 4080 has this melting problem but I’m certainly afraid from it.
I think I have seen one report of a 4080 connector melting. It's pretty much always 4090's being effected by this.
User error you do not plug the connector well. LOL 😂😆
those plugs are rated for 30 plugs an unplugs I think
It seems a bit strange to me that the original PSU didn't trip when the connector on the PSU side starting melting. The fact that it's fused with the PSU means there's a problem there, and I suppose the PSU protection should have cut the power by then. Hmm.....maybe that PSU needs some looking into.
As for the 4090, it would seem to me that the melting is not confined to any particular model of GPU or connector. I don't have all the data, but from the various videos I've watched thus far, there doesn't appear to be any single model that's been spared. Yes, I've seen those where the cable is bent immediately after the GPU and sensors attached didn't show any abnormally high temps or power drawn, so the thinking is that test example should not result in a melted connector if it had been left that way for daily use. Of cos, since it was a test, we'd never know what might actually happen in the long run. So, whether using the original 12+4 cable, whether plugging it in straight, whether it's bent, whether it's fully pushed in, whether it's drawing 450W or 600W, it doesn't seem to matter. As long as it's a 4090, it's only a matter of time before the connector melts. I'm no electrical expert and I know they say the plastic and copper wires in those connectors are rated for certain values, but personally, I'm of the opinion that regularly pushing 350W and more over those 12+4 things over an extended period of time is causing stress failure, resulting in the subsequent melting of the connectors. Each time we load the GPU, it draws around that amount or more. Over time, stress fatigue occurs and then the material fails and starts to melt. Maybe not 100% of all 4090s, but when you're paying so much money, do you want to risk finding out otherwise? Has anyone had a 4080 Super melt thus far? They draw significantly less than 350W sustained max.
And if I remember my basic electrics correctly, the thinner the wire gauge, the higher the resistance it has, all else being equal. And if you "force" a higher current thru thinner wires, logic tells me what should happen in due time.
I think it's a fatigue thing too ultimately. I do work with electricity although am not a licensed electrician but the way it makes sense to me the pins are so thin that plugging and unplugging can loosen how tight the connection is, there's also less surface area than 8 pin cables and a lot less cables carrying the current. Making this plug more prone to fatigue.
On top of that you have the connector warming up during high GPU use scenarios than cooling off when not in use, rinse and repeat hundreds of times that connector isn't as tight as it used to be which increases resistance even more. The higher the resistance the more of that electrical energy is converted into heat until it reaches that spiraling out of control point and it melts.
I admit in the video I did unplug the cable on the PSU multiple times so that I could benchmark other GPU's so am sure that contributed to it and I have heard from multiple comments in this video that the plug is rated for 30 times plug and unplug, just like 8pin PCIE cables. However I have unplugged 8 pin pcie cables hundreds of times and never had any issues like this. Clearly the design of this plug and the amount of power the 4090 draws is too much for it to be on the same level of robustness as the 8 pin pcie cables if that makes sense.
I'm not against having one plug to rule them all. I like the idea. It just seems this one wasn't very well design because any replacement should be at the very least as rigid as the old.
Class action lawsuit!
I’m for it.
Ok, I watched the video and this is so much worse than I expected. I haven't seen anyone report a melted connector on the power supply. You are the first person I have seen this happen to. All of the other reports are from the connector on the card itself melting. If your 4090 was always connected, and you just unplugged the cable from the power supply. That means not even the connector on the power supply is safe from melting. If the power supply catches on fire, its game over. I don't think it has much to do with cables being plugged in at this point. This all seems to come down to safety factor. der8auer talked about this in his 12VHPWR cable video. The reason why your 7900 XTX didn't cause the cable to melt was because it was only pulling 300 watts from the two 8-pins. 660/300 = 2.2. With your XTX, the safety factor is at 2.2. Not far from 8-pin power connectors. 660/450 = 1.467, 660/400 = 1.65. With the 4090, the safety factor went down to between 1.65 and 1.467 assuming you never pulled over 450 watts with it. That is a large reduction from the typical 1.9 to 2.5 of the 8-pin power connector. Nvidia should have just put two of these 12VHPWRs on the 4090. That would have avoided all of these issues.
Yeah that makes sense. The 12vhpwr from the 7900XTX was only powering 2 of the pcie cables and the third had its own socket. The 4090 ran entirely of that one connection. It goes beyond just user error.
Putting 2 12VHPWRs defeats Nvidia's purpose of the connector as it's only purpose is that you can power the GPU using just 1 cable.
@@bzrt554 Unfortunately that is the whole reason why a problem exists. The safety factor is slim to none here without two connectors.
because most use the adapters so the powersupply side is the old PCIE plugs which are WAY BETTER
@@LiLBitsDK These newer pcie 5.0 psu are 12pin on both sides
der8auer theory is logical. Thread count and gauge of wire. I have the same Tuff series gpu and i decided to use the cord that came with the gpu . More wire leads to split the load thus creating less heat and no melting threat . It is a great gpu had mine for a year so far.
It's literally basic common sense all of us that aren't even electric engineers understand yet somehow people working at nVidia thought it would be a good idea to squeeze the same amount of power that normally goes through 32 larger pins (4x8 pin connectors rated at 150W) through 12 smaller ones.
That's roughly 35% of the amount of pins that are like ~60-70% of the size of the normal ones, so basically less than ~25% of the power contact area you'd use when using 4x8 pins. Insanity. Even if you compare it to 3 8 pin connectors it would still be less than half of the power contact area. I can't even lol. So unnecessary, so avoidable and so stupid at the same time
Its crazy youtube nowadays, can we trust any soul anymore... everything influenced by greed and money.
I guess the only way to fix this issue is undervolting. Whether it is a simple power limit or voltage curve tuning, I think it does not really matter. What matters is power draw/voltage. Looks like this connector just can not sustain more than 400W of power delivery for relatively long time. And depending on some factors (like room temperature, GPU voltage, GPU power draw, a certain game behavoiur, the length of a gaming session etc.), it can melt anytime.
P. S. Coil whine is definitely a card issue (if it the card, which whines, because sometimes it can be PSU). But it can be reduced/increased, depending on what PSU you are using. So if your card has significant coil whine, try a different PSU. Higher quality PSUs with lower ripple level tend to decrease coil whine intensity. This is based on my own experience.
So what's the point paying for a premium GPU like 4090 which in the end you do undervolting.....buy a 4080 super . Just saying....
@@mariotop742 Fair point. But any undervolted 4090 is still significantly faster than even OCed 4080 Super. You can smartly undervolt 4090, so that it won't loose a single fps. Depending on a certain sample, you can even gain a few % of fps with undervolting. I personally always undervolt all my GPUs, so that would not be a problem for me. But for majority of inexperienced users this is indeed a big problem.
undervolting won't help... the issue is that the "poor connection" causes heat... it's not the amount of power being drawn... which the cables can handle just fine...
that's why I didn't buy flagship .. 3 yrs back i bought rtx 3070 Founders edition and now upgraded to rtx 4070 ti super gigabyte eagle oc 16gb at launch .
3080s , 3080 tis and 3090s killed the ppl psu wd transient spikes and now 4080 and 4090s are killing itself with high power connectors . im still running my high end machine on 9 yr old cooler master v650w 80 plus gold 😊
My 4090 runs since day 1 almost on a daily basis above 400w... in same games like Avatar and Metro Exodus 500 plus W... no issues so far.
Using Corsair PSU 1500watts and their 600w 12vhp... cable.
Q: Were you running your 4090 at stock settings or OC? I find that most ppl who report the burn up never mention this and most ppl do OC their gpu's. I have mine power limited down to 90% as it runs cooler and performs better but I make no claim that its safe from melting but I would imagine that its safer than running stock or OC. I also have temp probes attached to the gpu plug so I can always see whats going on and its setup to shut down if it hits 70C. Currently my plug never goes above 48C running my most demanding game CP2077 4K RT/PT ULtra settings.
EDIT: BTW I have the same card, bought my TUF back in Oct 2023 before prices skyrocketed for the holidays. Furthermore I believe you and I also agree this is Nvidia's fault / design flaw. Thx for the vid.
You can OC a 7900xtx to 3000mhz and beyond and launch the power limit to +30% on some models and you will never have to fear for your card. This connector is a firehazard and needs to be recalled period. And his card was at default power limit so your argument doesn't stand.
The only time I've run the 4090 overclocked is in a few 7900XTX vs 4090 videos but I not always use it stock 99% of the time but I also use framerate caps.
For example Dragons Dogma 2 is a game I've been playing a lot lately I cap my FPS to 90 at 3440x1440 with the FG mod and DLSS gpu draws 160 to 200 watts.
I was playing DD2 that morning and than I threw on No Rest For the Wicked at 4K. That game had Yue GPU to 100% pulling 360watts.
@@NoiseStaticBlurVery true. I have a Nitro+ 7900XTX I showed on this video. I OC that one to pretty close to 3ghz it pulls 460 watts all day long.
Hell some people just remove the power limit outright or flash a watercooled card's Bios to raise the limit. No melted connectors in decades of overclocking on 8 pin either. God knows how many of these were slightly loose connections too. 12VHPR is a bad connector.@@TerraWare
I also have a 4090, vertically mounted, custom water cooled, PSU > 3 8pin cables > Nvidia adapter > 3rd party 12pin extension > 4090.
My 4090 is limited to 300w, around 60-65% power limit, and i think this will save my 4090 from this stupid connector. (watch Derbauer video "4090 power target makes no sense)
I had mine in set to 70% all this time but I’ll gave a 60% a try
One of few reasons why i skipped the 40 series cards entirely. Just gonna keep pushing my old 3080ti until next gen. Hopefully the 50 series cards don't have this problem.
Unless you go for a 4090 it's not really worth upgrading from a 3080Ti anyway. I have a Strix 3080Ti. It's practically a more power hungry 4070 Super.
Yet Nvida has not addressed this issue 😮 Its just mot worth the risk as it cause issues down the line after purchase.
This is why i went with the 7900xtx.
To all those people who keep saying plug it in all the way... you guys are high as !@#$.. this wasnt a issue in past generations before. You dont hear about this with older 8 pin and 4 pin gpu designs..
Release connector with a safety factor of 1.1 vs the old pcie 8 pin with it's safety factor of 1.9, connectors start melting, blame user error cause nobody saw this coming.
Additional info, that means the connector only has 10% overhead before failure, even a hair or dust or wear or weakening springs in the plug can pretty easily cause a 10% degradation.
i have a good question. why 4080 doesn't have that issue?
Clearly power consumption plays a role. More power draw higher temperatures from electrical wire/connector resistance.
I'm just chilling with a 3060ti but I refuse to buy a gpu featuring timebombs like these. What the hell is Nvidia doing?
Not sure what I'll do if I need to upgrade.
So u think it could happen on an rtx 4080 too?
I suppose, I think I have seen one 4080 melt in a post, its usually 4090's because of the amount of wattage they can draw.
My EVGA Gtx 1080 ti sc black gaming used 369.9 watts loving life
Don't use that adapter, its alot worse than the native connections.
Not a good look for Nvidia and their fastest GPU. As a fellow 4090 owner, I feel for you.
I really hope this crap is fixed by the time the 5090 is out.
your card definitely has coil whine..
It sucks but they've been doing it since day one. Makes a product that cost them $300 to make, they use a new and untested connector type pumping 3-4 times more power through similar sized wires... It's going to melt. The best part, they sell it for $2000 😂
Design flaw that Nvidia doesn't want to hold their hands up to. These video's add another from a guy who's been repairing 25 4090s a week made me steer clear of this card even though I had the money to include it in my new build.
I hear ya. I got mine day one, well before these scenarios began to pop up and by well before I mean like a week or two before lol.
Would I buy one had I known. Yea probably with the warranty which thankfully I got a new replacement one.
I have 7900xtx
Yea 4090 get more fps in RT vs 7900 but if RT is off not so much bigger diff around 15-20 fps :D
It's funny someone argued against me in the comments not to long ago on a video by techyesCity, and I said the failure rate is about 90% within 18 months and he didn't believe me, even though the statistics are now starting to suggest it indeed is reality. I'm sad to see this and hope you can get it fixed. P.S the "A" version aka the Quadro version comes with the normal power connector .... and nvidia even had the gall to show it being "tortured in a video on how durable it is" .... I dunno if that video is still up but man what a kick in the face
They wont use this connector on workstation products lol but I want to know where you got that 90% figure from.
@@TerraWare most repair shops do about 100 to 300 a month, times the repair shops that offer the service times the amount of products sold. Don't forget this is a very "expensive product" if an average is 200 repair requests on this connector a month, this would indicate that this is indeed a ball park figure that is acurate. Say this is roughtly 5x you would end up around 1800 a month of failing units in total out of the 110.000 units sold in the USA in total in 2 years, you would end up with roughtly a failure rate of between 85% and 93% depending on total figures each month repaired. Since Nvidia puts the "40 series" sales figures as a whole and does not publish per country data its hard to really figure out how many 4090's exsist in the united states currently. But 110.000 would probably be acurate.
It is not a theory the mechanism does have a limited detachment count, it is clear in the submission of the cables for approval. It is a shame this happened, but I think people respond with alternative views is because they don't remove the cable once it is in situ. The newer versions of the reference cards have shorter sense pins, I know I have one of each. The card should have been released with the shorter sense pins, but that's another discussion. I know you will not agree, but it is design and user error. The card is not designed to be removed more than a very small amount of times, it is designed for consumer purposes not mixed commercial purposes. That said, whilst I disagree with the main premise of your video I sympathize with the destruction of your expensive purchase if people have a reference Nvidia card return will have no issue they are quick to RMA
Yeah this just showed this connector is flawed design wise when it's melting on the psu side as well, i hope they ditch this connector for next gen onward, otherwise it's intel or AMD for me.
People also have the 8 pin melt on the psu side before. Just google about it.
Hardware Nexus will say its your fault.