"nVidia finally responds to 4090 cables melting..."

Tengis

Supreme [H]ardness
Joined
Jun 11, 2003
Messages
6,076
https://www.extremetech.com/extreme...esponds-to-rtx-4090-cable-melting-controversy

This is apparently a “common issue,” according to the update; users may think it’s fully seated, and the card is operational despite it not being fully inserted. Nvidia provided a clear illustration of what this can look like. It says all the melted cables it examined showed a clear “wear line” that indicated it was not fully seated:

Edit: won't let me put the image in for some reasons
 

Attachments

  • aid_5413_000.webp
    328.8 KB · Views: 0
in other words, not Nvidias fault, user is to blame, user is out $2000. Gee, standard Nvidia SOP, if what I am saying is correct.
 
Nvidia hasn't declined any warranties that I've seen. So the user is out the use of his card for a few days while they RMA it.

Times Nvidia deserves our hate, this doesn't seem like one of those times.
Hate no... laughter yes.
I mean they designed a power delivery setup that allows the card to operate north of 500 watts when it isn't actually plugged in. I mean what are the sense pins for if a barely contacting connection is given a green light.
Anyway ya this is old news at this point I guess.... push those suckers in till you think you'll break your card. Don't worry you probably won't break anything and they probably won't deny you an RMA if you do. ;)
 
You’d think a card that costs $2 grand would come with a failsafe that protects the user and hardware from an incident like this.

Nvidia’s only honoring RMA’s because they’re wanting to avoid legal issues, but what I think is messed up is they’re using utilizing their position to blame the end-user; and their fan base is giving them a pass. Imagine if it were AMD, if people still complain about driver issues, something like this would be talked about years from now.
 
You’d think a card that costs $2 grand would come with a failsafe that protects the user and hardware from an incident like this.

Nvidia’s only honoring RMA’s because they’re wanting to avoid legal issues, but what I think is messed up is they’re using utilizing their position to blame the end-user; and their fan base is giving them a pass. Imagine if it were AMD, if people still complain about driver issues, something like this would be talked about years from now.
Happens every generation.... with the 3000s. The cards burning themselves out was the fault of Amazon and their MMO. This generation its user error. I guess when the 5000s explode it will be the AIBs fault... giving Nvidia the excuse they have been looking for to fire them all and be a single source company.
 
Nvidia’s only honoring RMA’s because they’re wanting to avoid legal issues,
I not sure you are serious here (I imagine not), it is what 0.04% of an rare Halo product having the issue but ultra talked about, honouring RMA's would be an easy no brainer virtually free thing to do even if they were no legal obligation to do so (which would be hard to imagine to be the case, hard to imagine that would not go under some warranty for such a new out of the box product) for PR reason and to help sales of future card with said connector.

The stain on the company to not do so would be worth 10 times of the price of replacing 0.04% of 4090 outthere.
 
I not sure you are serious here (I imagine not), it is what 0.04% of an rare Halo product having the issue but ultra talked about, honouring RMA's would be an easy no brainer virtually free thing to do even if they were no legal obligation to do so (which would be hard to imagine to be the case, hard to imagine that would not go under some warranty for such a new out of the box product) for PR reason and to help sales of future card with said connector.

The stain on the company to not do so would be worth 10 times of the price of replacing 0.04% of 4090 outthere.
I think we also have to consider that its 0.04% SO FAR. These plugs are badly designed. Based on what we know of how these fail... its only a matter of time before another chunk of users who have it "mostly" plugged find another Amazon MMO that pushes 550 watts for an hour and BURNs their plug that hasn't burned yet. Or cables that are running fine now... flake a little zinc and end up with a run away short. There are also plenty of people that have then plugged fine right now. That won't have them plugged properly in 6 months or a year from now when they open their case to add a new SSD or something. A little jiggling on the power cables on the back securing their SSD and wooops, next gaming session fire works.

I truly hope that all the bad press isn't forgotten in a month. The info needs to be in circulation so that people remember to always check their stupid Nvidia power connection every time they open their case. (not unplug and replug just check it isn't loose) Last thing you do now before you put the side of your case back on... give that nvidia cable a good push. lol And yes... all the cable checking is going to cause more fails. All they had to do was add a click, I really don't understand the thinking on that design oversight.
 
Hate no... laughter yes.
I mean they designed a power delivery setup that allows the card to operate north of 500 watts when it isn't actually plugged in. I mean what are the sense pins for if a barely contacting connection is given a green light.
Anyway ya this is old news at this point I guess.... push those suckers in till you think you'll break your card. Don't worry you probably won't break anything and they probably won't deny you an RMA if you do. ;)

2 of the sense pins are strictly to inform the card how much power the PSU can deliver.

The third is a power good signal from the card to the PSU. It's optional.

The 4th is intended as a cable connected check. It's optional.

Page 27/28:
https://www.cybenetics.com/attachs/52.pdf
 
Luckily the cards I don't think were damaged. The dongles were melted but I believe the connectors were in tact. You would just have to clean the melted plastic and replace the dongle. I run a 4090 rated braided dongle on a 4080 and it doesn't even get warm.
 
I think we also have to consider that its 0.04% SO FAR.
Even if it end up being a giant 2% of them, it would not be simply fear of legal issue that would motivate NVIDIA, is it not obvious that the bad press of not honouring warranty for something so talked about would also be a motivation.

It seems almost trivial to be true.
 
1h5jn8.jpg


LEAVE NVIDIA ALONE!
 
I would just make the clip the full width of the plug, vs a little tiny one in the middle, thus lower the chance for wiggle to even happen to the point the plug begins to come out, if it is attached, or prevent as Nexus showed, the contacts not being made well on all points.
Even move to a metalic clip or something?
I mean, you do got to push it in pretty dam hard to get it to click as they showed, so could see why people might think it is in, since the clip is not well visible to confirm, but I mean, ya, anyone installing any hardware these days should know all PSU's plugs "clip" in and you can usually hear, feel it.
 
Last edited:
I would just make the clip the full width of the plug, vs a little tiny one in the middle, thus lower the chance for wiggle to even happen to the point the plug begins to come out, if it is attached, or prevent as Nexus showed, the contacts not being made well on all points.
Even move to a metalic clip or something?
I mean, you do got to push it in pretty dam hard to get it to click as they showed, so could see why people might think it is in, since the clip is not well visible to confirm, but I mean, ya, anyone installing any hardware these days should know all PSU's plugs "clip" in and you can usually hear, feel it.

And after attaching a cable is it not customary to give it a tug and say "that's not going anywhere", after said tug produces no movement? I mean if it moves then it's not tight enough and you have to begin the ritual again.
 
And after attaching a cable is it not customary to give it a tug and say "that's not going anywhere", after said tug produces no movement? I mean if it moves then it's not tight enough and you have to begin the ritual again.
Customary to smart people, yes. But as we have all seen the years go by, the vast majority of people have been getting dumber and dumber.
 
Happens every generation.... with the 3000s. The cards burning themselves out was the fault of Amazon and their MMO. This generation its user error. I guess when the 5000s explode it will be the AIBs fault... giving Nvidia the excuse they have been looking for to fire them all and be a single source company.
And don't forget the "space invaders" of the 2000 series.

Nvidia just needs a good slogan to better communicate the issue. How about: "Don't Fondle That Dongle!"
 
Customary to smart people, yes. But as we have all seen the years go by, the vast majority of people have been getting dumber and dumber.
I wouldn't say smart, I mean the same ritual is followed to tie anything down to the back of a truck and I can tell you some stories there involving a beat-up flatbed and some takeoff assists.
 
Happens every generation.... with the 3000s. The cards burning themselves out was the fault of Amazon and their MMO. This generation its user error. I guess when the 5000s explode it will be the AIBs fault... giving Nvidia the excuse they have been looking for to fire them all and be a single source company.
No, the cause was poor soldering around EVGA's MOSFET circuits, it was an EVGA QC issue. New Worlds lack of a frame filter on their menus was an oversight for sure and would have caused other problems later down the line that they really should have found far earlier in the development cycle, this would have put large amounts of stress on a card for sure, but there was a reason it was only the EVGA cards that were affected by the problem.
 
Do you have proof of that?
https://www.thefpsreview.com/2021/0...x-3090-new-world-failures-on-soldering-issue/

https://amp.hothardware.com/news/evga-root-cause-geforce-rtx-3090-failures-new-world-mmorpg

“After conducting a series of tests and close examinations, EVGA believes it has discovered the reason why some of its GeForce RTX 3090 graphics cards were giving up the ghost when running Amazon's New World MMORPG. What it boils down to, if EVGA's analysis is correct, is sloppy soldering on certain areas that would not necessarily be visible to the naked eye.”
 
https://www.thefpsreview.com/2021/0...x-3090-new-world-failures-on-soldering-issue/

https://amp.hothardware.com/news/evga-root-cause-geforce-rtx-3090-failures-new-world-mmorpg

“After conducting a series of tests and close examinations, EVGA believes it has discovered the reason why some of its GeForce RTX 3090 graphics cards were giving up the ghost when running Amazon's New World MMORPG. What it boils down to, if EVGA's analysis is correct, is sloppy soldering on certain areas that would not necessarily be visible to the naked eye.”
I thought we were talking about Space Invaders cards, my bad.
 
I thought we were talking about Space Invaders cards, my bad.
Easy mistake to make the X090 cards do seem to have a history of failures, but that one was faulty memory modules, I don't know if it was ever explained why it resulted in that particular pattern, I would be interested to know what generated that specific character, I assume it is some sort of internal debug thing but would love to read a breakdown on the specifics should one be available.
 
Easy mistake to make the X090 cards do seem to have a history of failures, but that one was faulty memory modules, I don't know if it was ever explained why it resulted in that particular pattern, I would be interested to know what generated that specific character, I assume it is some sort of internal debug thing but would love to read a breakdown on the specifics should one be available.
I finally got the whole story, but it was promised off the record. Hence me asking above.
 
I finally got the whole story, but it was promised off the record. Hence me asking above.
Ah, the best I could find after my 2080TI failed was something about how the space invaders each corresponded to an SM or a subset of one and blah blah blah technobabble but that was Reddit troubleshooting at its drunkest.
 
I had a 2080TI die space invaders style. I got it replaced under warranty. nvidia warranty has been good about things.

PR is not driving replacement for nvidia, simple economics is. Companies like GM Ford etc do this math all the time. You look at the cost of lawsuits, recalls, and the failure rates, and potential sales losses and then go with whatever is cheapest. GM got bit hard by this with the faulty problem ignition where they failed to weigh the publicity value correctly when considering that the problem cost lives as well. nvidia is simply looking at the cost of repairing these cards (a connector is really the repair here as no cards died from the bad connection) along with a total 50 repairs so far along with shipping costs and maybe offsetting the cost of an AIB selling what will be just b-stock sales.

With the updated information for amateur pc builders to CHECK THE DAMN CONNECTION, a lot of new owners will be looking at eternal shame of their pc building skills being put into question. Male pride will pretty much keep the rate of these failures lower than the initial first month.
 
I finally got the whole story, but it was promised off the record. Hence me asking above.
I bet it was someone who didn't wash thier hands after going to the bathroom. They have those signs for a reason...
 
Still funnier than Norovirus... Worst new years dinner ever...
I've worked on cruise ships before, it's one of the most common sicknesses in confined spaces such as a ship. And all because people don't wash their hands properly after taking a shit. :meh:
 
I've worked on cruise ships before, it's one of the most common sicknesses in confined spaces such as a ship. And all because people don't wash their hands properly after taking a shit. :meh:
Ski resort in my case but yeah.

They got the whole event sick, the hotel apparently forced their kitchen staff to come in sick or not. It was spreading through their employee housing. Made for a really bad drive home.

Wash your hands and don’t work when your leaking.
 
Last edited:
from the PCI-SIG consortium, not that it amounts to much of anything in the grand scheme of things aside from their governance body washing their hands of the scenario and situation herein:

1669921156789.png
 
I bet it was someone who didn't wash thier hands after going to the bathroom. They have those signs for a reason...
Yes they have those signs to remind you that most people will normally not wash their hands (even with signs up) and you should not take risks with your food safety!
 
Back
Top