News User claims RTX 4090 16-pin power connector melted on both GPU and PSU side, despite running at 75% power

Status
Not open for further replies.
Failures of the connectors on both ends is a common cascading failure mode for power connectors with a high number of parallel circuits. The same thing happens to pralleled wires in industrial power wiring. I've seen it many times over my career as a design engineer. As one pin on connector starts to go high resistance (overheats), it drops the current in that wire and the other parallel conductors have to take up the slack. This puts more stress on the remaining usable conductors and connector pins. The weakest pin in the remaining parallel group then gets the hottest, which can be made worse if it is in the center of the housing (heating from the neighboring pins).

This issue has been going on for a long time, with many users making relevant comments on what is happening (including me). It appears that they did not select enough pins on the connector for paralleling and did not derate what the pins could handle on top of that.

From June 16, 2023 Tom's article:
https://www.tomshardware.com/news/12vhpwr-connector-melting-psu-side

The saving grace is that most GPU and PSU vendors are using UL approved connectors which should have 94V0 plastic. That type of plastic is hard to truly ignite and have a flaming fire, and if it does flame up, it goes right out once the heat is removed.

Since I started thinking about this issues more, it is much like the problem that automakers had trying to use legacy 12VDC power rails in cars as DC power usage went up (over 6-8KW). Once they moved to 48VDC rails, the weight of the wiring harnesses got smaller, and the reliability went up due to lower connection failures (reduced current stress). Not to mention the alternator efficiency went up (diode bridge losses as part of the total voltage output).

These GPUs could have the same solution. As far as that goes, why not bring 48VDC into the motherboards as well? It would be ATX 4.0.
 
"there hasn't been any closure on the meltdowns."

Well, there kind of has been. To date, there has been only 1 cause found, and it has affected every single one of the melted cables looked at.

I know, har-har, not plugged in... but that's just the simple truth. Time and time again, people just failed to plug it in all the way. There's talk about it being difficult in some cases, but that doesn't mean you should just stop trying to plug it in. If it doesn't go in, return it. Then Nvidia / OEMs can deal with why it wasn't plugging in, and you don't have to worry about melting your cables, and possibly damaging the PSU/etc.

In this case, it's a bit hard to say from the few pictures shown, but the burnt cables on the PSU side strongly suggest it was angled, and not fully inserted.

@vinay2070 There was... and the person suing had already posted photos online that showed the tell-tale signs of the cable not being properly plugged in, and then silence.

Reminder: People have pushed these cards beyond 1kw with cables barely even getting warm. The cable isn't the main problem. ( Could it be designed better to ensure it's plugged in? Of course. That clearly needs to be done. But at the end of the day, it's up to the user to make sure it is. It's impossible to idiot-proof something. ) Even the "bend-gate" was disproven as insufficient to cause the issue. It has to actually be unplugged an amount to cause any issue.

@Releximas Just make sure you have it plugged all the way in. If it's fully seated and latched, you're good to go.

Oh, slight caveat: There WAS the rather ironic cablemod adapter issue. People went out and bought the cablemod adapters for fear of their cables melting. The original cable was fine, but the cablemod adapter was faulty and had to be recalled.
 
Oh, slight caveat: There WAS the rather ironic cablemod adapter issue. People went out and bought the cablemod adapters for fear of their cables melting. The original cable was fine, but the cablemod adapter was faulty and had to be recalled.
It was a bit of both, they had similar design issues as the original 12VHPWR connector, and people could still plug them in wrong. The right angle adapter was there to solve the connector being too close to many chassis side panels.

They decided to recall them because they were covering repair costs for people that had problems. Cheaper to remove them from the market than it was to keep sending people's GPUs in for repair.
 
  • Like
Reactions: thestryker
"there hasn't been any closure on the meltdowns."

Well, there kind of has been. To date, there has been only 1 cause found, and it has affected every single one of the melted cables looked at.

I know, har-har, not plugged in... but that's just the simple truth. Time and time again, people just failed to plug it in all the way. There's talk about it being difficult in some cases, but that doesn't mean you should just stop trying to plug it in. If it doesn't go in, return it. Then Nvidia / OEMs can deal with why it wasn't plugging in, and you don't have to worry about melting your cables, and possibly damaging the PSU/etc.

In this case, it's a bit hard to say from the few pictures shown, but the burnt cables on the PSU side strongly suggest it was angled, and not fully inserted.

@vinay2070 There was... and the person suing had already posted photos online that showed the tell-tale signs of the cable not being properly plugged in, and then silence.

Reminder: People have pushed these cards beyond 1kw with cables barely even getting warm. The cable isn't the main problem. ( Could it be designed better to ensure it's plugged in? Of course. That clearly needs to be done. But at the end of the day, it's up to the user to make sure it is. It's impossible to idiot-proof something. ) Even the "bend-gate" was disproven as insufficient to cause the issue. It has to actually be unplugged an amount to cause any issue.

@Releximas Just make sure you have it plugged all the way in. If it's fully seated and latched, you're good to go.

Oh, slight caveat: There WAS the rather ironic cablemod adapter issue. People went out and bought the cablemod adapters for fear of their cables melting. The original cable was fine, but the cablemod adapter was faulty and had to be recalled.
This many people don’t just “fail to plug it in properly” for no reason. Even though I haven’t had the issue since I only have a 4080, I fully recognize the plug is NOT a good design with a nice firm click like it should be.
 
the solution is to add 48V to the 12V as we used 5V and 12V in older systems.

48V for GPU power and 12V for the rest...

Plus changing into USB-C types of connectors.
 
This many people don’t just “fail to plug it in properly” for no reason. Even though I haven’t had the issue since I only have a 4080, I fully recognize the plug is NOT a good design with a nice firm click like it should be.
Part of it is definitely laziness instilled due to how overbuilt PCIe 6 and 8 pin connectors are. I had a 6+2 improperly plugged in for months and the only reason I noticed is that I bumped my case and the PC crashed with the video card fans going to max.

I do think the majority of the blame is split between where the plugs are located on most cards, and the connector itself as you say not clicking in place in a fashion that's both easy to hear and feel.
 
Came to see the inevitable "it's not Nvidia's fault" posts. Was not disappointed.
This kind of crap exactly. The whole user error joke has gotten stale long ago. Does Nvidia have a single honest bone in their body?

Does there ever come a point where ya just come out and say ya know what, we made a mistake, the 16 pin cable was a dumb decision. It didn't work out. We're humans, we made a mistake. We can learn from our mistake.

Of course not, because that means recall which costs shareholders money. It's easier to just pass the buck. Let's make it the user's fault instead of owning our mistake. One can only hope they fix this on the 5090. If it ain't broke, don't fix it. Go back to the 8 pins so we can finally put this to rest.
 
They really need to redesign the connector to have a larger mating surface (similarly to the 8 pin connector), and then give it a screw locking mechanism similar to DVI and VGA cables.
The thumb screws have a proven track record of maintaining a solid connection, even with rather rough cable management. With the current 16 pin design, even when the plastic latch is clipped in place, it is possible to partially wiggle the cable loose, as well as have certain positions that even slightly stress the cable result in the connection working itself loose slightly, and leading to a melted connector.
Connectors like the 8 pin connectors have such a large mating surface that even with loose tolerances for the locking tab, the cable will maintain ample contact to not result in the connector melting unless there is a major defect with the cable or connector.
wE3l6Mk.jpg

F40UEkF.png
 
  • Like
Reactions: iLoveThe80s
This many people don’t just “fail to plug it in properly” for no reason. Even though I haven’t had the issue since I only have a 4080, I fully recognize the plug is NOT a good design with a nice firm click like it should be.
Except that's exactly what has happened. As I mentioned, this isn't just random "Oh, let's blame users" or anything.

Because the problem involves high temperatures, it leaves hard physical evidence on the connector showing how far it was plugged in. In lawsuits, forum posts, and every single cable that has been examined, it has been shown that they were not plugged in all of the way.
Came to see the inevitable "it's not Nvidia's fault" posts. Was not disappointed.

This kind of crap exactly. The whole user error joke has gotten stale long ago. Does Nvidia have a single honest bone in their body?

Does there ever come a point where ya just come out and say ya know what, we made a mistake, the 16 pin cable was a dumb decision. It didn't work out. We're humans, we made a mistake. We can learn from our mistake.

Of course not, because that means recall which costs shareholders money. It's easier to just pass the buck. Let's make it the user's fault instead of owning our mistake. One can only hope they fix this on the 5090. If it ain't broke, don't fix it. Go back to the 8 pins so we can finally put this to rest.
Just because you want to hate on Nvidia, does not mean this particular case is their fault.

Remember, it wasn't even Nvidia that originally found this.

In fact, people often conveniently forget that this isn't even unique to the 16 pin. Do a search, and you will find every single one of the power connectors has a problem with melting, ( Molex, 8-pin, 6-pin, etc. ) many of which are likely also failure to properly insert the connector ( or with cheap enough cables, you can get pins to come loose and retract into the cable ).

It's also worth reminding, that not a single alternative cause has ever been found. Many theories were tested, bent cables, inserted but wiggled a little loose, damaged pins, only using 3/4 of the cables. Each of those tested could barely even warm the cable/connector. The only thing that has ever resulted in a melted cable was it being unseated, and not just a little. ( Most likely people are almost plugging it in, then while closing the case/etc. it gets bumped/wiggled further out. )

Let me be clear, I still think PCI-Sig should do more to fix the connector ( it's not an Nvidia connector btw. Many companies are involved, including AMD and Intel. They just haven't adopted it yet, and now they will likely delay or even abandon just to not be associated with it, but that's separate. ) There's clearly some room for improvement. Make the main pins longer, but leave the sense pins. ( This gives room for error on the power pins, and would make it so failure to plug in all the way could disconnect the sense pins, and cause the card to go into emergency mode. ) More secure latch, with better feedback. I don't think it needs active attachment like screws, but maybe. Maybe a barrel nut, etc. Clearly, if all of the standard power cables have this issue, it's something that could use more work than the current friction with a small latch system that they all use.

I also want to be clear about my intentions. I'm not trying to white knight Nvidia, PCI-Sig or anyone else. They've got plenty of money, they don't need it. They make mistakes all the time. The 40 series was mostly a disaster. Of course the 4090 was expensive, classic top of the line premium... but what about the others? They were so cut down, yet still high priced. But I digress.

I just really dislike the fear-mongering when we have long known the specific cause of the problem, which is super easy to avoid, and largely unrelated to the cable itself. ( As mentioned, every single power cable in a computer has suffered the same issue. ) They should NOT have stated that the issue is still unknown, when it clearly is. There's a reason the class-action lawsuit fizzled out, that we almost entirely stopped hearing about the problem overnight when the cause was found, that no recall was forced, etc.

PS: Nvidia also didn't do nothing. Any video card affected could be RMAed for the issue, even 3rd party cards. That's despite them finding and seeing clear proof that it was failure to properly seat the cable in 100% of cases. Though, that's likely because it's such a small number of people affected. Best estimates I could find were approx. 0.05% of cases. At that point, it's easier to just RMA. If it had been a lot more, yea, I believe Nvidia would have started rejecting RMAs based on failure to seat the cable.
 
The connectors, as specked by NVIDIA, with pure copper or hi-copper bronze pins and thick high purity copper wire might present no overheating problem, but when the part is lowest-price sourced and they end up with high zine brass pins and cheap high resistance, under gauge low purity copper alloy wire....
This is not a new problem in PC connectors, but usually it just causes sporadic poor connections as plugs come loose having lost their spring and won't stay securely connected.
12 volts is safe for amateurs (a surprise is not an injury), 24 volts can burn fingers and indirectly kill the susceptible.
 
Just because you want to hate on Nvidia, does not mean this particular case is their fault. Remember, it wasn't even Nvidia that originally found this.

The fact that they didn't find it first means absolutely NOTHING.... and this has nothing to do with hating NVidia. Good grief.

In fact, people often conveniently forget that this isn't even unique to the 16 pin. Do a search, and you will find every single one of the power connectors has a problem with melting, ( Molex, 8-pin, 6-pin, etc. ) many of which are likely also failure to properly insert the connector ( or with cheap enough cables, you can get pins to come loose and retract into the cable ).

Yes, 8-pins have melted too. It happens, not nearly as much as 4090's. You would have to look pretty deep to find anything <=4080 melting. It happens, but extremely rare. TH also has a nice article about one particular computer repair shop that gets 100s of 4090's every day that are melted. Ya. It's definitely user error. I wonder why users keep making errors on plugging in a 4090, but ALMOST never anything <=4080. Mankind has been plugging electrical cables into sockets for over 100 years, yes suddenly when it comes to a single item, we just are too stupid to figure out how to do it. Good grief

It's also worth reminding, that not a single alternative cause has ever been found. Many theories were tested, bent cables, inserted but wiggled a little loose, damaged pins, only using 3/4 of the cables. Each of those tested could barely even warm the cable/connector. The only thing that has ever resulted in a melted cable was it being unseated, and not just a little. ( Most likely people are almost plugging it in, then while closing the case/etc. it gets bumped/wiggled further out. )

Mechanistic cause and responsibility/fault are not the same thing. I saw many of these tests, but none of them were done in real world situations. It wasn't just the cable being unseated properly, its the surface area of the metal pins inside the connector itself is too small. The pins are too small and there isn't enough surface area to offset the heat created by current. The pins of an 8-pin are physically bigger and longer, more surface area for the heat to spread out. It's not just proper seating, its also the fact that if the internal pins are not absolutely 100% perfect down to the subatomic level, there will be melting. If it slides or loses contact by an atoms-width: boom, melting. This is a design flaw, not user error

If they want to use a single connector, that's fine, but the pins inside must by longer, thicker, wider, larger surface area, and better bound to the internal connector itself, along with a clicking sound when its plugged in all the way on both sides. I've

There's a reason the class-action lawsuit fizzled out, that we almost entirely stopped hearing about the problem overnight when the cause was found, that no recall was forced, etc.

Yes, because every day PC gamers going up against the massive conglomerate of lawyers from a multi-billion dollar company don't stand a chance. There wouldn't be enough of them and enough money to bank-roll that lawsuit. Too many tech-tubers ran a 5 minute test on the 4090 and falsely determined it was user error. "Oh, mine didnt melt. Its your fault" ..... didnt help matters either.

That's despite them finding and seeing clear proof that it was failure to properly seat the cable in 100% of cases.

Heh, that's interseting. I wonder why users keep not properly seating a 4090, yet every other GPU in the last 20 years, we somehow knew how to seat just fine, even 3090Ti's, 4080s, etc etc. I wonder why users have such trouble seating a cable on a 4090. Good grief. Sometimes it just takes a little thinking

Oh.... and one more thing. I ran my 2080S, around the clock for 2 years, under full load, non-stop, with one single daisy-chained 8-pin cable (was doing video stuff). It never melted, and the GPU was perfectly fine, running max wattage pull for 2 years straight. At one point, I went on vacation for 2 weeks while that was running. I would go to sleep every night, and it was always running doing on full load. Never had an issue.

....Now I triple dog dare ANYONE to do that with a 4090. I dare you. Go ahead.... hashtag noballs. I dare you..... GN, LTT, and all other tech tubers who say "Oh you didn't plug it in properly".... Ok... Go ahead... plug in a 4090, put it on a 2 week run, uninterrupted, full load in your 5-million dollar brand new home..... and go on vacation for 2 weeks. Obviously us gamers are too stupid to plug in a cable, so seeing as all these "user error" techtubers are so perfect at it, put your magic cable plugging in touch to work, 4090, 2 weeks, full load. Go on vacation while the task in running. Put your money where your mouth is. I dare you. You won't.... hashtag noballs.

Enough of this user error garbage. Just stop doing it. Seriously. Enough is enough. @ NVidia Just own your forking mistake like a man and do better on the 50 series so we can move on. Good grief
 
Last edited:
  • Like
Reactions: Pierce2623
In the comments I was reading it's an FSP power supply and apparently they are having issues?

Hard to say since user didn't post pictures of case, ventilation, temps, etc..

But FSP may be popular in Asia but definitely not here.

Either way, for it to melt at the PSU I would suspect the PSU was at fault here with the limited info. I don't recall ever seeing other posts with the PSU ports melting and for that to happen, PSU protection is either faulty or non existent but could be wrong on that.
 
Except that's exactly what has happened. As I mentioned, this isn't just random "Oh, let's blame users" or anything.

Because the problem involves high temperatures, it leaves hard physical evidence on the connector showing how far it was plugged in. In lawsuits, forum posts, and every single cable that has been examined, it has been shown that they were not plugged in all of the way.



Just because you want to hate on Nvidia, does not mean this particular case is their fault.

Remember, it wasn't even Nvidia that originally found this.

In fact, people often conveniently forget that this isn't even unique to the 16 pin. Do a search, and you will find every single one of the power connectors has a problem with melting, ( Molex, 8-pin, 6-pin, etc. ) many of which are likely also failure to properly insert the connector ( or with cheap enough cables, you can get pins to come loose and retract into the cable ).

It's also worth reminding, that not a single alternative cause has ever been found. Many theories were tested, bent cables, inserted but wiggled a little loose, damaged pins, only using 3/4 of the cables. Each of those tested could barely even warm the cable/connector. The only thing that has ever resulted in a melted cable was it being unseated, and not just a little. ( Most likely people are almost plugging it in, then while closing the case/etc. it gets bumped/wiggled further out. )

Let me be clear, I still think PCI-Sig should do more to fix the connector ( it's not an Nvidia connector btw. Many companies are involved, including AMD and Intel. They just haven't adopted it yet, and now they will likely delay or even abandon just to not be associated with it, but that's separate. ) There's clearly some room for improvement. Make the main pins longer, but leave the sense pins. ( This gives room for error on the power pins, and would make it so failure to plug in all the way could disconnect the sense pins, and cause the card to go into emergency mode. ) More secure latch, with better feedback. I don't think it needs active attachment like screws, but maybe. Maybe a barrel nut, etc. Clearly, if all of the standard power cables have this issue, it's something that could use more work than the current friction with a small latch system that they all use.

I also want to be clear about my intentions. I'm not trying to white knight Nvidia, PCI-Sig or anyone else. They've got plenty of money, they don't need it. They make mistakes all the time. The 40 series was mostly a disaster. Of course the 4090 was expensive, classic top of the line premium... but what about the others? They were so cut down, yet still high priced. But I digress.

I just really dislike the fear-mongering when we have long known the specific cause of the problem, which is super easy to avoid, and largely unrelated to the cable itself. ( As mentioned, every single power cable in a computer has suffered the same issue. ) They should NOT have stated that the issue is still unknown, when it clearly is. There's a reason the class-action lawsuit fizzled out, that we almost entirely stopped hearing about the problem overnight when the cause was found, that no recall was forced, etc.

PS: Nvidia also didn't do nothing. Any video card affected could be RMAed for the issue, even 3rd party cards. That's despite them finding and seeing clear proof that it was failure to properly seat the cable in 100% of cases. Though, that's likely because it's such a small number of people affected. Best estimates I could find were approx. 0.05% of cases. At that point, it's easier to just RMA. If it had been a lot more, yea, I believe Nvidia would have started rejecting RMAs based on failure to seat the cable.
Dude the connector is a poor design. I’ve got a 4080 and it absolutely doesn’t give a good firm click like the old 8 pins.
 
Except that's exactly what has happened. As I mentioned, this isn't just random "Oh, let's blame users" or anything.

Because the problem involves high temperatures, it leaves hard physical evidence on the connector showing how far it was plugged in. In lawsuits, forum posts, and every single cable that has been examined, it has been shown that they were not plugged in all of the way.



Just because you want to hate on Nvidia, does not mean this particular case is their fault.

Remember, it wasn't even Nvidia that originally found this.

In fact, people often conveniently forget that this isn't even unique to the 16 pin. Do a search, and you will find every single one of the power connectors has a problem with melting, ( Molex, 8-pin, 6-pin, etc. ) many of which are likely also failure to properly insert the connector ( or with cheap enough cables, you can get pins to come loose and retract into the cable ).

It's also worth reminding, that not a single alternative cause has ever been found. Many theories were tested, bent cables, inserted but wiggled a little loose, damaged pins, only using 3/4 of the cables. Each of those tested could barely even warm the cable/connector. The only thing that has ever resulted in a melted cable was it being unseated, and not just a little. ( Most likely people are almost plugging it in, then while closing the case/etc. it gets bumped/wiggled further out. )

Let me be clear, I still think PCI-Sig should do more to fix the connector ( it's not an Nvidia connector btw. Many companies are involved, including AMD and Intel. They just haven't adopted it yet, and now they will likely delay or even abandon just to not be associated with it, but that's separate. ) There's clearly some room for improvement. Make the main pins longer, but leave the sense pins. ( This gives room for error on the power pins, and would make it so failure to plug in all the way could disconnect the sense pins, and cause the card to go into emergency mode. ) More secure latch, with better feedback. I don't think it needs active attachment like screws, but maybe. Maybe a barrel nut, etc. Clearly, if all of the standard power cables have this issue, it's something that could use more work than the current friction with a small latch system that they all use.

I also want to be clear about my intentions. I'm not trying to white knight Nvidia, PCI-Sig or anyone else. They've got plenty of money, they don't need it. They make mistakes all the time. The 40 series was mostly a disaster. Of course the 4090 was expensive, classic top of the line premium... but what about the others? They were so cut down, yet still high priced. But I digress.

I just really dislike the fear-mongering when we have long known the specific cause of the problem, which is super easy to avoid, and largely unrelated to the cable itself. ( As mentioned, every single power cable in a computer has suffered the same issue. ) They should NOT have stated that the issue is still unknown, when it clearly is. There's a reason the class-action lawsuit fizzled out, that we almost entirely stopped hearing about the problem overnight when the cause was found, that no recall was forced, etc.

PS: Nvidia also didn't do nothing. Any video card affected could be RMAed for the issue, even 3rd party cards. That's despite them finding and seeing clear proof that it was failure to properly seat the cable in 100% of cases. Though, that's likely because it's such a small number of people affected. Best estimates I could find were approx. 0.05% of cases. At that point, it's easier to just RMA. If it had been a lot more, yea, I believe Nvidia would have started rejecting RMAs based on failure to seat the cable.
You keep claiming 100% of cases were shown to be user error but that was only ever written by some online magazine. We don’t know if that’s really true. In fact, I highly doubt it. No product in the world has ever had 100% of failures caused by user error..
 
The fact that they didn't find it first means absolutely NOTHING.... and this has nothing to do with hating NVidia. Good grief.
In case it wasn't clear what I meant since I was a bit vague, I was saying Nvidia was not the one that found out and called it user error. That was multiple other sources, and not just "youtubers". Nvidia then later responded that they had found the same thing, noting that it was the case in 100% of cards that had been sent in.

This does mean something, because it means it wasn't Nvidia just trying to make up some excuse out of nowhere to sweep it under the rug. It was verified across multiple sources first.
Yes, 8-pins have melted too. It happens, not nearly as much as 4090's. You would have to look pretty deep to find anything <=4080 melting. It happens, but extremely rare. TH also has a nice article about one particular computer repair shop that gets 100s of 4090's every day that are melted. Ya. It's definitely user error. I wonder why users keep making errors on plugging in a 4090, but ALMOST never anything <=4080. Mankind has been plugging electrical cables into sockets for over 100 years, yes suddenly when it comes to a single item, we just are too stupid to figure out how to do it. Good grief
Well as you say, there is a strong bias towards higher power cards, independent of cable style. It takes 2 to tango, you need to pull enough power plus have a fault, before the cable will melt. And other cables melting isn't anywhere near that much rarer than 4090s. 4090 is already vaguely in the 0.05% range is all. For hardware, that's a pretty good rate. Apple/Samsung wish their phones were that reliable.

Also, I would point out that 1 repair shop is the only one saying it, and they're saying it on their youtube, with links referring people to themselves. That easily puts them below random youtubers ( as low as that is ), and puts them alone. No one has corroborated their story in any way, so I would file that firmly under "suspicious", much like the original melting reports, and the original speculations / etc. for the cause.

The difference with the failure to plug in, is there is simple physical evidence that anyone can check. When the cable isn't seated properly, it leaves a visible mark on the plug. If you see that mark, it's (primarily) the users fault.

This means you can happily ignore / distrust the youtubers, and Nvidia, and repair shops, and the users... and just look at it for yourself.

It's also worth noting, that there remains only 1 single method of reproducing the issue: Not properly seating the cable. Every single other theory has failed to reproduce the problem, typically barely even making the connector warm.
Mechanistic cause and responsibility/fault are not the same thing. I saw many of these tests, but none of them were done in real world situations. It wasn't just the cable being unseated properly, its the surface area of the metal pins inside the connector itself is too small. The pins are too small and there isn't enough surface area to offset the heat created by current. The pins of an 8-pin are physically bigger and longer, more surface area for the heat to spread out. It's not just proper seating, its also the fact that if the internal pins are not absolutely 100% perfect down to the subatomic level, there will be melting. If it slides or loses contact by an atoms-width: boom, melting. This is a design flaw, not user error

If they want to use a single connector, that's fine, but the pins inside must by longer, thicker, wider, larger surface area, and better bound to the internal connector itself, along with a clicking sound when its plugged in all the way on both sides.
Wow, complete BS here. "100% perfect down to the subatomic level"? There wouldn't even be a single working one. If you think the connector manufacturers are managing a 99.95% success rate when dealing with subatomic precision, you are out of your mind. In fact, pretty sure quantum mechanics expressly forbids that level of precision.

You may notice btw, most of your suggestions are exactly what I said. I then went and checked the details, and surprise! PCI-SIG already did the big ones, improving the pins, lengthening the power pins, and shortening the sense pins. You'll also notice those are specifically designed to prevent the problem of improperly seating the cable. If it wasn't an actual problem, they wouldn't be implementing the fix for it.

This fix has already been implemented in the latest versions of the cards.
Yes, because every day PC gamers going up against the massive conglomerate of lawyers from a multi-billion dollar company don't stand a chance. There wouldn't be enough of them and enough money to bank-roll that lawsuit. Too many tech-tubers ran a 5 minute test on the 4090 and falsely determined it was user error. "Oh, mine didnt melt. Its your fault" ..... didnt help matters either.
Actually, that is the norm in class action lawsuits, and they happen all the time. They have a few things going for them. First, is numbers. Second, nearly all attorneys working on class action lawsuits do so at little to no cost upfront, instead taking a ( large ) percentage of the winnings.
Heh, that's interseting. I wonder why users keep not properly seating a 4090, yet every other GPU in the last 20 years, we somehow knew how to seat just fine, even 3090Ti's, 4080s, etc etc. I wonder why users have such trouble seating a cable on a 4090. Good grief. Sometimes it just takes a little thinking
No, plenty of other cards also have not been properly seated. It just wasn't a brand new $1,600 monster GPU and getting splashed all over the internet. They absolutely were there though.
Oh.... and one more thing. I ran my 2080S, around the clock for 2 years, under full load, non-stop, with one single daisy-chained 8-pin cable (was doing video stuff). It never melted, and the GPU was perfectly fine, running max wattage pull for 2 years straight. At one point, I went on vacation for 2 weeks while that was running. I would go to sleep every night, and it was always running doing on full load. Never had an issue.

....Now I triple dog dare ANYONE to do that with a 4090. I dare you. Go ahead.... hashtag noballs. I dare you..... GN, LTT, and all other tech tubers who say "Oh you didn't plug it in properly".... Ok... Go ahead... plug in a 4090, put it on a 2 week run, uninterrupted, full load in your 5-million dollar brand new home..... and go on vacation for 2 weeks. Obviously us gamers are too stupid to plug in a cable, so seeing as all these "user error" techtubers are so perfect at it, put your magic cable plugging in touch to work, 4090, 2 weeks, full load. Go on vacation while the task in running. Put your money where your mouth is. I dare you. You won't.... hashtag noballs.
I actually have been daily driving a 4090 since a week or 2 after release, overclocked at that, and have run entire days of heavy gaming load, as well as benchmarking. People have pulled over 1,000w through the cable, which is MORE than double the default, and done just fine. I know 4 other people that have also been daily driving a 4090 for nearly the same time. ( I worked in the gaming industry, so I know a lot of hardcore gamers. ) My work had systems running them 24/7, no issues reported. Many cloud gaming servers, and other high-end servers can run hundreds of RTX 4090s, again 24/7, without issue.
Enough of this user error garbage. Just stop doing it. Seriously. Enough is enough. @ NVidia Just own your forking mistake like a man and do better on the 50 series so we can move on. Good grief
Reminder, Nvidia didn't make the connector. It was PCI-SIG that designed it, and ( last I heard ) 2 manufacturing companies that made it. Nvidia just bought the connectors and soldered them on. ( Well, they were also part of the large group that worked to design it, including AMD and Intel. )

And, they DID step right up. They offered an RMA for absolutely anyone with a problem, even if the card was actually a 3rd party card, like PNY, ASUS, Gigabyte, etc.

PCI-SIG then updated the design, the manufacturers updated the connected, and that is now the connector used on cards.
 
  • Like
Reactions: TJ Hooker
Dude the connector is a poor design. I’ve got a 4080 and it absolutely doesn’t give a good firm click like the old 8 pins.
That seems to be more of a manufacturing issue, because many do give a good click. But, yes, that is 1 of the areas that has been noted for improvement.
You keep claiming 100% of cases were shown to be user error but that was only ever written by some online magazine. We don’t know if that’s really true. In fact, I highly doubt it. No product in the world has ever had 100% of failures caused by user error..
It was actually reported by Nvidia themselves that every one they had received had physical evidence that it was not properly inserted. ( Then various sites / youtubers checked the ones they had, and confirmed the same. ) Additionally, images from users on forums/reddit/etc. that showed the side, showed evidence they were not properly seated. ( That last part is really key for me personally, that you can look at reports and when they show the side of the cable, you can see the warping, white line, or melted edge typically a few mm up the pins, showing it just was not inserted all the way. )

That wasn't meant as 100% of all RMAs all together, or anything. It was only specifically in reference to the burnt connectors, using standard cables. ( So not cablemod adaptors for example. I'm not sure what the exact specific issue was there, though judging by a post above here, it sounds like it was an extension of the failure to fully seat, made worse by adding in more connection points. )

True, even that probably isn't accurate at this point. That was just some months in to it.
 
Status
Not open for further replies.