Nvidia RTX 4090 power connectors melting?

Been running the cable mod wires and 180 degree adapter for a few months now. Wattage pull in the worst of games is 400-450w. The aluminum back of the plug doesn't really get hot to the touch.

The new cable mod adapters do nothing really to prevent user error from being the culprit nor does the video do anything other than the guy claiming that it "was" plugged in fully despite there no evidence to show it was or was not plugged in fully. He's ranting on fixing a few of these out of 130k 4090s sold now.

"melted" cable news has dried up and this is more of a click count video imho
 
Last edited:
Slade said:
Been running the cable mod wires and 180 degree adapter for a few months now. Wattage pull in the worst of games is 400-450w. The aluminum back of the plug doesn't really get hot to the touch.

The new cable mod adapters do nothing really to prevent user error from being the culprit nor does the video do anything other than the guy claiming that it "was" plugged in fully despite there no evidence to show it was or was not plugged in fully. He's ranting on fixing a few of these out of 130k 4090s sold now.

"melted" cable news has dried up and this is more of a click count video imho
Click to expand...
Perhaps, but it could still be a manufacturing defect that affects a small % of cards. Low enough that nvidia won't issue a recall over.
 
Furious_Styles said:
Perhaps, but it could still be a manufacturing defect that affects a small % of cards. Low enough that nvidia won't issue a recall over.
Click to expand...
If it's within the expected failure rate, then just rma and be done with it. It's not like cards don't fail for one reason or another. We've yet to have a story about a house burning down since the 4090 came out. Not sure about where this guy operates out of, but there's more than enough stock to stay on walls for weeks if not maybe months of 4090 stock locally here. I don't see any reason for a 4090 not to be warrantied out as nvidia did mention they would still warranty if this problem occurs despite user error.
 
Not that I'm concerned (I like my odds on this), but does the warranty cover the rest of your machine if the part burns up and wrecks your entire build?
 
I found it interesting how long they had to push it and how "unplugged" it had to be to get to the fire point on GN. Judging by all the cards display that had this problem, it appears that eventually the system hits a point where the card fails and probably locks up the system cutting off the load before it gets bad. Historically, I've had a power supply explode mid game a few decades ago and take the motherboard and video card with it and both were replaced under warranty. I didn't lie about what killed it either. Said I saw a flash from my box and mobo and video card died. I had a spare system and psu so I could do the testing. I've never tested a warranty against fire from outside sources but I suspect it would be a no.
 
Domingo said:
Not that I'm concerned (I like my odds on this), but does the warranty cover the rest of your machine if the part burns up and wrecks your entire build?
Click to expand...
Not a chance unless its a prebuilt. If your self built machine go down in flames from the cable issue, you'd most likely have to sue NVDA personally.
 
SomeGiantIdiot said:
Not a chance unless its a prebuilt. If your self built machine go down in flames from the cable issue, you'd most likely have to sue NVDA personally.
Click to expand...

That's kinda what I figured. I wouldn't be concerned about a fire so much as a $4000-5000 PC getting borked. Most people rocking 4090's are probably also rocking lots of high-end components.
 
Slade said:
Been running the cable mod wires and 180 degree adapter for a few months now. Wattage pull in the worst of games is 400-450w. The aluminum back of the plug doesn't really get hot to the touch.

The new cable mod adapters do nothing really to prevent user error from being the culprit nor does the video do anything other than the guy claiming that it "was" plugged in fully despite there no evidence to show it was or was not plugged in fully. He's ranting on fixing a few of these out of 130k 4090s sold now.

"melted" cable news has dried up and this is more of a click count video imho
Click to expand...

True while we can blame it on the user, may be they could have done a better job to ensure shit was more bullet proof when snapping in. These users existed with old connectors too that didn't melt like this. I hate to blame all the users for it, may be they should have had a more user error proof snap in mechanism.
 
NKD said:
True while we can blame it on the user, may be they could have done a better job to ensure shit was more bullet proof when snapping in. These users existed with old connectors too that didn't melt like this. I hate to blame all the users for it, may be they should have had a more user error proof snap in mechanism.
Click to expand...
FWIW, the adapter included with my Gigabyte card gave both a very noticeable tactile and audible click when it was pushed all the way in, and it can't be wiggled at all. It didn't feel numb going in (that's what she said) like shown in the GN video.
 
Armenius said:
FWIW, the adapter included with my Gigabyte card gave both a very noticeable tactile and audible click when it was pushed all the way in, and it can't be wiggled at all. It didn't feel numb going in (that's what she said) like shown in the GN video.
Click to expand...
FWIW, I used the nvidia cable that came with my 4090. There was not really a very noticeable click or feel too it. Knowing the issues, I did ensure it was in all the way and did not budge when I pulled or wiggled it, but it was slightly concerning at first like "Is that correct?". So I guess people not familiar with building PCs may not even notice an issue.
 
I used an aliexpress cable for about 2 months, it looks fine. Then my Cablemod 90 arrived, and I installed that, and now it has about 2 months of use. It looks good as well, checked it this morning.

I suspect in the above example, either the cablemod 90 was defective, or when it was inserted, it was not quite aligned and the pins in the card pushed on the metal contacts in the adapter, pushing them up, so that they never had proper contact. That second row that you can tell was untouched from heat damage after he pulled it apart. Those were likely not making any contact.
It's too bad he had to destroy it to get the plug off of the card. Problem determination is much more difficult. At least he replaced the plug for the user. Could also be the youtuber just looking for clicks. Haven't heard anywhere else about any more of these happening.

I wonder if there has been any new re-design work on the female side of these, the side in the cables/adapters?

We also don't know how much more of this has happened since the first spate of them. 1 or 2 does not a story make.
 
Last edited:
Last edited:
TaintedSquirrel said:
More popping up daily.
https://www.reddit.com/r/nvidia/comments/13p5jz9/4090_tuf_cable_burnt_fully_connected_cable_after/
https://www.reddit.com/r/nvidia/comments/13p2r3d/12vhpwr_adapter_melting_after_6_months/
https://www.reddit.com/r/nvidia/comments/13nmojh/4090fe_burned_corsair_sf750_platinum_corsair/

edit: Subreddit mods are deleting the posts.

GN is investigating again (@ 7:45)
Click to expand...


From one of those threads:
50-platinum-corsair-12vhpwr-ryzen-v0-ms5nruk3xc1b1.jpg
 
FrgMstr said:
So a very clear visual indicator of the plug not being fully inserted is a bad thing?

That said, still a bad design and we will see a lot more burning in a few years after they have been seated/removed through many cycles.
Click to expand...
I don't think its a bad thing, but we are starting to see more problems that appear to point to a design issue. Putting another product's marketing overtop of addressing the issue is very problematic in my opinion for consumers. But I guess if we exhaust all other possible solutions it could lead to an actual fix.
 
TaintedSquirrel said:
MSI has solved the issue with the power of... paint.

Click to expand...


Paint or yellow plastic as part of a 2 color injection molding process? It looks like yellow plastic to me; adding paint would add new potential sources of failure if it flakes off and multiple insertion/removal cycles result in the debris getting jammed into the inside of the connector pins.

The connector itself needs redesigned to either remove that 1-2mm of slack in the clip or to be able to carry maximum current even in the loose position. MSI can't do that, color coding a bad insertion for easily visible inspection is probably the best any OEMs can do until NVIDIA/PCISIG/etc revise the form factor.
 
FrgMstr said:
So a very clear visual indicator of the plug not being fully inserted is a bad thing?

That said, still a bad design and we will see a lot more burning in a few years after they have been seated/removed through many cycles.
Click to expand...
Outside of reviewer testbeds how many people regularly un/replug their connectors. Most of them are probably inserted once and then left there for 1 to 4 years until a new GPU is installed.
 
DanNeely said:
Outside of reviewer testbeds how many people regularly un/replug their connectors. Most of them are probably inserted once and then left there for 1 to 4 years until a new GPU is installed.
Click to expand...
I unplug mine and remove the card when I clean it out with the compressor, but even so, the card will have long outlived its usefulness before I reach the 20+ limit of the connector.
 
Purely anecdotal here, and based on a tiny sample, but I noticed something interesting when I picked up a Gigabyte 4070 the other week...

The fit of the power plug into the socket feels noticebly worse than on my Founders 3070Ti and MSI 3090Ti.

I have two adapters and two native cables, all different brands. All of them easily seated into both 30-series cards and made a definitive click, even if the click obviously felt a bit weaker than olde 8-pin plugs.

However when I plugged one of the native cables into the 4070, it was very difficult to get it to seat properly had to give it a big shove at the end and only got the tiniest of clicks. I could definitely see how it would be possible to think it's all the way in when it's not...

Like I said, purely anectodal, but I wonder if something is different in a bad way with the sockets on 40-series cards...
 
  • Like
Reactions: Wat
like this
I use the same 180 degree adapter and unlike his video with the wiggle, mine is pretty stable in there on my 4090. I've had it out just a day ago to swap some nvme drives around. Still no signs of melting since 4090 launch.
 
Today is the first time I need to clean out my PC since I got my 4090 and need to remove the card... wish me luck I get it connected correctly again and don't have a melted connector in a few weeks... lol. Its been flawless performance for 7 months now.
 
DanNeely said:
Paint or yellow plastic as part of a 2 color injection molding process? It looks like yellow plastic to me; adding paint would add new potential sources of failure if it flakes off and multiple insertion/removal cycles result in the debris getting jammed into the inside of the connector pins.

The connector itself needs redesigned to either remove that 1-2mm of slack in the clip or to be able to carry maximum current even in the loose position. MSI can't do that, color coding a bad insertion for easily visible inspection is probably the best any OEMs can do until NVIDIA/PCISIG/etc revise the form factor.
Click to expand...
Full length clips should have been on it similar to a 4+4 8-pin EPS connector. Or on both sides of the connector even.
 
While a redesign seems appropriate to mitigate this issue, it would seem to be that if you want to make sure it doesnt happen to you, pressing the connector in hard until there is an audible click, and doing a light tug test (holding the connector, not the wires) to make sure it is fully inserted, followed by a last sanity "push in" just in case the tug test, and then never bending the cables at a tight angle out of the connector should protect you from it happening to you.

Either way, if this type of extraordinary precaution is necessary it is a design fail. This shit should never have hit the market, and should have been recalled the moment these issues started appearing.

As a company you need to design things with your user base in mind, and given that these are largely not installed by trained operators in a factory, but by end users, the connector should have been designed to be "end user proof".

I still consider this to be a MAJOR fail on the part of PCI-SIG and Nvidia.
 
Going on 9 months now with zero issues. Pretty much the same stance from my last post on this. The design works, but the built-in safety margins for handling a sligtly sub-par connection or solder joint simply just doesn't cut it, hence all the reported issues. Still think they should have gone with a better design than what they arrived at (the new tweaks which now appears to be in the works don't look like much - hopefully it's enough). Still think they should have simply used 4 of the older type, proven 8 pin connector design rather than what they arrived at.
 
Last edited:
Supercharged_Z06 said:
Going on 9 months now with zero issues. Pretty much the same stance from my last post on this. The design works, but the built-in safety margins for handling a sligtly sub-par connection or solder joint simply just doesn't cut it, hence all the reported issues. Still think they should have gone with a better design than what they arrived at (the new tweaks which now appears to be in the works don't look like much - hopefully it's enough). Still think thay should have simply used 4 of the older type, proven 8 pin connector design.
Click to expand...
I disagree with the last point. While splitting it into multiple 8 wire chunks is less bad than the 23 wire monster of the main ATX mobo connector, that large of a connection area and that many wires is unwieldy mess and PITA for cable management even in consumer systems where you'd normally only have a single card. In compute workstations/servers where you might have 4 or 8 of them it was beyond excessive and a more compact form factor is needed.
 
Actual fire.

https://www.reddit.com/r/nvidia/comments/14svvhf/suffered_the_fate_of_the_burning_180_degree/

Was playing Diablo IV earlier tonight, decided to go grab a snack and come right back. I made it halfway up my stairs when I smelled something burning, came right back to see the fire and smoke from the GPU connector. Thankfully I smelled the smoke and was able to get to it in time, I don't know how far the fire may have gone had I not gotten there in time. RIP my white ASUS Strix 4090, will see what happens with warranty on it.

I have already opened asupport ticket with Cablemod and I am waiting to get a call from ASUS support on it.

I was using an EVGA 1200W Platinum PSU with a Cablemod 4x8-Pin to 12VHPWR cable and I made sure that the connectors on the adapter and the PSU cable were both seated fully and not in a wierd angle. This has been installed and running fine for acouple months now.

Waterblock is fine as far as I can tell, the soot wiped right off with a cloth, had to go to work and will drain and inspect it more tomorrow morning.

My main PSA is don't leave you system on when you're not using it of you are running a 4090 (I know I never will again). I normally hibernate or shut mine down but forget sometimes, this could have ended much worse had I not been there to deal with immediately.

Pics of the damage:


https://preview.redd.it/ljcquojzuga...ed&s=4d3ea619775650072734bc2f55e9b0102078310e


https://preview.redd.it/u9s47i10vga...ed&s=1ecbb5e534fe5f1c93d38f53518db3f430a80d84


this is after I messed with an exacto knife by the 4 sense pins to try and break the melted plastic to remove the adapter.

https://preview.redd.it/xl8lhqe0vga...ed&s=e9b609075efe155a18f2434182f0c2a54adc24ff


https://preview.redd.it/joc3zwp0vga...ed&s=91e396b79ba3318fd528955fcdc14bf843f485a5
Click to expand...

u9s47i10vgab1.jpg
 
TaintedSquirrel said:
Actual fire.

https://www.reddit.com/r/nvidia/comments/14svvhf/suffered_the_fate_of_the_burning_180_degree/



View attachment 581887
Click to expand...
Gruesome as that looks, it would be a pretty easy repair, assuming you could get a replacement part, which I assume must be possible if Cablemod and Der8auer can.

Fake edit: This one goes the wrong way for that Asus board, but is probably the right one for some other brand.
https://www.mouser.com/ProductDetai...=sGAEpiMZZMvlX3nhDDO4AFL5XOD9rcrf7uvLqTg5LRc=

I'd be tempted to solder a pigtail onto the board with a series of old skool 8 pin connectors, particularly if the board vendor refused to honor the warranty.
 
RazorWind said:
Gruesome as that looks, it would be a pretty easy repair, assuming you could get a replacement part, which I assume must be possible if Cablemod and Der8auer can.

Fake edit: This one goes the wrong way for that Asus board, but is probably the right one for some other brand.
https://www.mouser.com/ProductDetail/Amphenol-FCI/10160920-1240100LF?qs=sGAEpiMZZMvlX3nhDDO4AFL5XOD9rcrf7uvLqTg5LRc=

I'd be tempted to solder a pigtail onto the board with a series of old skool 8 pin connectors, particularly if the board vendor refused to honor the warranty.
Click to expand...
CableMod pays for a replacement card.
They get the dead card repaired and I think give it away. Saw a post about it.
 
Kudos to cablemods for sticking up for owners when gpu companies don't. The stats are interesting in the thread cablemods have for failures. The burnt connector pic in the post above has a suspicious gap in it indicating it wiggled out.
 
You must log in or register to reply here.
Back
Top