Nvidia RTX 4090 power connectors melting?

Some of us feel it's common sense (which we know isn't that common) for a company selling GPUs at this price to make sure the design of the power connector doesn't need special treatment and monitoring. You know, basically how it's mostly always been.
That's kind of what I said when I said it wasn't the best design. I plug mine in and don't monitor it since I diagnosed that it was a connector issue not being plugged in all the way.
 
I have Corsair's cable and it's rigid as hell and sticks out more than you'd think. It's a tight fit in my 5000D and it pushes against the glass ever so slightly. Probably not enough to matter, but I'm going to keep my eye on it. The instant other cable options start showing up (especially a right-angle one), I'll be hopping on board. For what it's worth, the 4-cable adapter my card came with was far less rigid and pokes out less than Corsair's 1-cable solution. It's ugly and kind of a pain in the ass, but it doesn't come in contact with my glass panel at all.
New solution from Cablemod.
https://old.reddit.com/r/nvidia/comments/10fo4it/you_guys_asked_for_a_preview_here_it_is/

ngrqa6muxxca1.jpg
 
I found the CableMod 4-1 adapter stiff and hard to bend. The CableMod cable was pretty but the amount of tension I had to put on it to get it routed properly in the P600S worried me. I am much happier with the BTOS Seasonic cable.
 
So they are still not out? Damn they hyped these up hard. I remember like 3 months ago it was releasing but that was release of a waiting list lmao. They really creating hard fomo for these I guess rofl.
Middle of next month is the email I got from them.
 
I'm 100% interested in that item. I'm also curious what Corsair provides with their new ATX 3 "Shift" PSU's that are due sometime soon. I have this one and like I mentioned, it's rigid as hell and non pliable. It seems sturdy, but it also probably won't fit in every case. They have a "premium" one that looks a hell of a lot softer, so I'll hope that they find a happy medium.
 
I'm 100% interested in that item. I'm also curious what Corsair provides with their new ATX 3 "Shift" PSU's that are due sometime soon. I have this one and like I mentioned, it's rigid as hell and non pliable. It seems sturdy, but it also probably won't fit in every case. They have a "premium" one that looks a hell of a lot softer, so I'll hope that they find a happy medium.
You have the Corsair cable? Idk if I would call it rigged. I had mine bent a little and 2 of the cables started coming out the back. Unless they changed something which I hope. I was surprised how easily it turned in to a molex connector.
 
You have the Corsair cable? Idk if I would call it rigged. I had mine bent a little and 2 of the cables started coming out the back. Unless they changed something which I hope. I was surprised how easily it turned in to a molex connector.

Yup. Ordered/received it last week. The cable itself feels like hard plastic. It has very little bend to it, especially as it gets closer to the connectors on both sides. Roughly about an inch and a half from the connectors has almost zero bend. It looks like the plug that goes into the video card has soldered cables, but I didn't exactly mess around with it much.
 
The design is faulty. Period.

Once somebody dies in a fire, it will be taken seriously.
I am agreeing with you, sorry if I didn't articulate that adequately.

Just stating if you have the cards that use this plug already, then you definitely have to check it plugs in all the way regardless of it clicking.
 
I am agreeing with you, sorry if I didn't articulate that adequately.

Just stating if you have the cards that use this plug already, then you definitely have to check it plugs in all the way regardless of it clicking.
Some thoughts on this with everyone chiming in that it was not pushed in all the way.

FWIW, I worked in industrial supply for 10 years before I got into HardOCP, and my primary job was selling thermoplastic hose and tubing, so I have a bit of exposure when it comes to melting plastic.

The only failures we have seen actually happen have been "forced" by the tech journos covering this issue.

My thought is that is it quite possible that failures could be occurring with the connector plugged all the way in. The plug gets super-hot, and due to the plug on the plastic expanding, it is pushing itself out of the socket elongating the capture clip or it coming unclipped, then once it has pushed out as far as it can, it sits there and "cooks," showing the line where the plug has moved out to. Once the connector cools off, you would have no proof of that capture clip elongated or failed.

Just thinking through this, I do not think all the investigation we have seen done on this is conclusive until you could actually see one fail that was fully seated.

That's my 2 cents, you may need change.
 
Some thoughts on this with everyone chiming in that it was not pushed in all the way.

FWIW, I worked in industrial supply for 10 years before I got into HardOCP, and my primary job was selling thermoplastic hose and tubing, so I have a bit of exposure when it comes to melting plastic.

The only failures we have seen actually happen have been "forced" by the tech journos covering this issue.

My thought is that is it quite possible that failures could be occurring with the connector plugged all the way in. The plug gets super-hot, and due to the plug on the plastic expanding, it is pushing itself out of the socket elongating the capture clip or it coming unclipped, then once it has pushed out as far as it can, it sits there and "cooks," showing the line where the plug has moved out to. Once the connector cools off, you would have no proof of that capture clip elongated or failed.

Just thinking through this, I do not think all the investigation we have seen done on this is conclusive until you could actually see one fail that was fully seated.

That's my 2 cents, you may need change.
I was in the plastics industry for over twenty years in quality control and have overseen everything from the injection molding process down to the assembly end of things. You may be on to something here.
 
Some thoughts on this with everyone chiming in that it was not pushed in all the way.

FWIW, I worked in industrial supply for 10 years before I got into HardOCP, and my primary job was selling thermoplastic hose and tubing, so I have a bit of exposure when it comes to melting plastic.

The only failures we have seen actually happen have been "forced" by the tech journos covering this issue.

My thought is that is it quite possible that failures could be occurring with the connector plugged all the way in. The plug gets super-hot, and due to the plug on the plastic expanding, it is pushing itself out of the socket elongating the capture clip or it coming unclipped, then once it has pushed out as far as it can, it sits there and "cooks," showing the line where the plug has moved out to. Once the connector cools off, you would have no proof of that capture clip elongated or failed.

Just thinking through this, I do not think all the investigation we have seen done on this is conclusive until you could actually see one fail that was fully seated.

That's my 2 cents, you may need change.


Nice insight. I am also wondering if this is one of the reasons some of the AIBs don't want you pushing lot of power through this even though they have more than enough power stages to push all you can. For example MSI seems really conservative when it comes to adding more power even on their water cooled card.

Some of the partners might be trying to avoid whatever then can due to too much power being pushed through one plug.
 
Some thoughts on this with everyone chiming in that it was not pushed in all the way.

FWIW, I worked in industrial supply for 10 years before I got into HardOCP, and my primary job was selling thermoplastic hose and tubing, so I have a bit of exposure when it comes to melting plastic.

The only failures we have seen actually happen have been "forced" by the tech journos covering this issue.

My thought is that is it quite possible that failures could be occurring with the connector plugged all the way in. The plug gets super-hot, and due to the plug on the plastic expanding, it is pushing itself out of the socket elongating the capture clip or it coming unclipped, then once it has pushed out as far as it can, it sits there and "cooks," showing the line where the plug has moved out to. Once the connector cools off, you would have no proof of that capture clip elongated or failed.

Just thinking through this, I do not think all the investigation we have seen done on this is conclusive until you could actually see one fail that was fully seated.

That's my 2 cents, you may need change.
Now there's an angle I hadn't considered. Thanks.
 
My thought is that is it quite possible that failures could be occurring with the connector plugged all the way in. The plug gets super-hot, and due to the plug on the plastic expanding, it is pushing itself out of the socket elongating the capture clip or it coming unclipped, then once it has pushed out as far as it can, it sits there and "cooks," showing the line where the plug has moved out to. Once the connector cools off, you would have no proof of that capture clip elongated or failed.

Just thinking through this, I do not think all the investigation we have seen done on this is conclusive until you could actually see one fail that was fully seated.
Hmm, sounds plausible. But wouldn't the plastic have to expand a lot for that to happen?
If the plugs on these adapters differs somehow to those plugs used on the 3090FE Cables, and/or the shape of that plastic sheath around the adapter contact, then these could behave differently. But I used a 3090 for 2 years and it was fine. But, that connector was the 'preview' version of the 12vhpwr connector, and didn't have the sense pins. So yeah, there could have been changes to the sizing of that plastic around the contact.

Seems like we would be seeing bulged plastic.... well, no, they get soft and melt... if heating softens the plastic, not seeing how that would push it out of the socket. If the plug got really hot, and was being pulled taught due to cable management, then the clip could let go, and you know the rest of the story. My theory anyway. GN showed that when pulled at a particular angle, some of the connectors can lose connectivity, so all of the current goes thru the remaining connections. Then you can get some heat buildup, then the plastic softens just enough for the clip to let go, it drops to fewer connected pins, then overheats and connector melts. More examples of a cable pulled taught. Those 4 big connectors all the same distance from the 12vhpwr connector is shit, makes cable mgmt difficult.

The clip should run the entire width of the connector imo.
That's my 2 cents, you may need change.
I do, gimme a dollar.
 
Dunno why they just didn't go with 3 regular 8 pin pcie connectors and be done with it. Its not as if they couldn't have went that route, they probably didn't like the optics of a card requiring 3 8 pins (which make it look like a power hog from an outward appearance) so decided to go with one shit 12 pin and still have it be a power hog in stealth mode.
 
But wouldn't the plastic have to expand a lot for that to happen?
If the 12VHPWR connector's clip is like that of the EPS and PCIE ones, no, probably half a millimeter would probably be enough for it to be able to slip.
 
Best write-up on dielectric grease I have ever read.

Also, of you are going to use this on a 12VHPWR plug, do NOT put the grease on the actual metal electrical connections. Just use a tiny bit on the outside edges of the male plastic connector housing. May very well help you get a full insertion.

I love male full insertion :p
 
Last edited:
Some thoughts on this with everyone chiming in that it was not pushed in all the way.

FWIW, I worked in industrial supply for 10 years before I got into HardOCP, and my primary job was selling thermoplastic hose and tubing, so I have a bit of exposure when it comes to melting plastic.

The only failures we have seen actually happen have been "forced" by the tech journos covering this issue.

My thought is that is it quite possible that failures could be occurring with the connector plugged all the way in. The plug gets super-hot, and due to the plug on the plastic expanding, it is pushing itself out of the socket elongating the capture clip or it coming unclipped, then once it has pushed out as far as it can, it sits there and "cooks," showing the line where the plug has moved out to. Once the connector cools off, you would have no proof of that capture clip elongated or failed.

Just thinking through this, I do not think all the investigation we have seen done on this is conclusive until you could actually see one fail that was fully seated.

That's my 2 cents, you may need change.

I learned the hard way a few years back to never trust 3rd party cables.

This 12v motherboard power extension came from Microcenter.

Worked fine for years until it suddenly couldn't keep up with an overnight handbrake transcode:

196494_IMG_20190223_165838.jpg


196496_IMG_20190223_165913.jpg


And yes, it was in all the way, its' only separated a little bit in this picture because I tried to pry it off to see if I could save the PSU modular cable, but that thing is was welded together.

Through some miracle of fortune it didn't kill anything (well, except the power cable)
 
I was just looking at the 90 degree adapters for the 12vhpwr connector, on aliexpress.

They have version that will angle either direction from the plug. So, the issue is, let's say you need one for your card that angles OVER the clip. How the hell will you be able to ever unclip it?

Solution to this, would be to make it so that the clip move to the sides. That way there are 2 clips, you have to squeeze both to get the connector out, and it is able to be reached on cables that angle in either direction. It should also make it a lot less likely that a cable would slip out very far. For this, the plug on the card would also have to be updated.

It's pretty obvious problem if someone had given a little more thought to the design.
 
We wouldn't be even needing stuff like the 90-degree adapter if someone had given a little more thought to the design.
We really don't "need" them, but they would tidy things up some. I honestly never had an issue with the stock dongle connector though it was cumbersome. Now that I've moved on to a PSU that is PCIE 5.0 ready I don't even feel the need to get a 90 degree adapter.
 
Pre-orders just went live:

https://store.cablemod.com/12vhpwr-angled-adapter/

But at $40 for a small power adapter, I can't help but feel we are being ripped off...

Small volume orders doesn't turn profit after design work very quickly, so price has to go up. Cuz let's be honest, it wouldnt be worth the work to try to get a profit on charging $10ea for that. And there will be a new video card and connector setup in the next version making your connector invalid very soon ill guess.
 
Small volume orders doesn't turn profit after design work very quickly, so price has to go up. Cuz let's be honest, it wouldnt be worth the work to try to get a profit on charging $10ea for that. And there will be a new video card and connector setup in the next version making your connector invalid very soon ill guess.

That is true I guess, but even "small volumes" is relative. Between the few 3000 series that had these connectors and the 4000 series to date, we are at least talking several hundred thousand potential customers.

Of course, not all of them are going to buy, but...
 
That is true I guess, but even "small volumes" is relative. Between the few 3000 series that had these connectors and the 4000 series to date, we are at least talking several hundred thousand potential customers.

Of course, not all of them are going to buy, but...
Cost of an injection mold eats a lot and is upfront. So yea they gotta price it to get a decent ROI timeframe and also care competing with others in this category. And they are competing with copies that will do little to no testing and possibly be a worse solution than nvidias. I cant imagine them selling to more than 10% of the total number of cards out there, prolly depends how much they spend on marketing and advertising too.

And yea they can buy in bigger volume to get price down but then it gets risky for how many do you want to sit on
 
Cost of an injection mold eats a lot and is upfront. So yea they gotta price it to get a decent ROI timeframe and also care competing with others in this category. And they are competing with copies that will do little to no testing and possibly be a worse solution than nvidias. I cant imagine them selling to more than 10% of the total number of cards out there, prolly depends how much they spend on marketing and advertising too.

And yea they can buy in bigger volume to get price down but then it gets risky for how many do you want to sit on

Lets say have ~50,000 customers out of what has been sold to date (with an unknown number in the future, depending on how much longer the plug gets used)

Hmm. Lets think about it.

So we'll have - what - 3 days of engineer time, 1 week of testing, and the mold, and leave a handsome amount of room for profit. I don't see any really complicated features on the part, so, I am going to guess the mold isn't going to be too complicated. Maybe $10k at most? (I've seen people mold threaded round parts for less, so this should be possible)

Maybe your development, testing and mold parts amount to some $50k?

Then you have to figure in the COGS, which is going to take some manual assembly for the conductors, as well as some plastic for the molded parts. Do this in a cheap labor market and you are still probably only talking some $3 bucks a piece.

At $10, you'd have $7 of profit per part, so your break even is going to be just north of $7,000 pieces.

At $40, your profit per unit is $37, so you break even at ~1,350 ea.

Now, I didn't consider sales, distribution, etc., but I think I'm in the neighborhood.

I think they just upped the price because they could, not because it was warranted by the work that went it to it, balanced by the volume.

And the thing is, people will pay. Especially since they make it available in many different pretty colors.

As we all know, the value of something is determined by what people will pay, not by the cost to make something.

Unfortunately we live in a time where FOMO has made the market completely irrational, and people are willing to pay absolutely crazy amounts of money for stupid things.

They probably assessed enough people would be willing to pay $40 that it would work, so why charge less?
 
They probably assessed enough people would be willing to pay $40 that it would work, so why charge less?

Probably, but I already own the regular CableMods cable, it was less than $40. I would pay maybe $20, and that would be a bigger market but maybe not as profitable. I'm not paying $40.
 
Lets say have ~50,000 customers out of what has been sold to date (with an unknown number in the future, depending on how much longer the plug gets used)

Hmm. Lets think about it.

So we'll have - what - 3 days of engineer time, 1 week of testing, and the mold, and leave a handsome amount of room for profit. I don't see any really complicated features on the part, so, I am going to guess the mold isn't going to be too complicated. Maybe $10k at most? (I've seen people mold threaded round parts for less, so this should be possible)

Maybe your development, testing and mold parts amount to some $50k?

Then you have to figure in the COGS, which is going to take some manual assembly for the conductors, as well as some plastic for the molded parts. Do this in a cheap labor market and you are still probably only talking some $3 bucks a piece.

At $10, you'd have $7 of profit per part, so your break even is going to be just north of $7,000 pieces.

At $40, your profit per unit is $37, so you break even at ~1,350 ea.

Now, I didn't consider sales, distribution, etc., but I think I'm in the neighborhood.

I think they just upped the price because they could, not because it was warranted by the work that went it to it, balanced by the volume.

And the thing is, people will pay. Especially since they make it available in many different pretty colors.

As we all know, the value of something is determined by what people will pay, not by the cost to make something.

Unfortunately we live in a time where FOMO has made the market completely irrational, and people are willing to pay absolutely crazy amounts of money for stupid things.

They probably assessed enough people would be willing to pay $40 that it would work, so why charge less?

Also they made 4 adapters, so some costs are shared some are not.

I think a lot of risk assessment put into it. Pretty close on my estimates on pricing, testing can be a bit more depending on what countries you are selling in and if they are doing all of it the 'right' way. Also whoever they are distributing through wants their cut as well and shipping costs (if baked into price).

Likely going to be an aliexpress part for $8 for a 2 pack at some point, it will be trash but it cuts into your customer base. So you want to make some profit as soon as you can because its diminishing returns very quickly.

And yes, you really do want to charge the most the market will allow, otherwise your are undercutting yourself. This can be dangerous though as the larger the gap between the high and and the low end, the more places can be filled in between.

And sometimes price is two edged. A low price can mean imply low quality and a high price can imply the opposite.

My knowledge of such things is much more limited, im an electrical engineer not marketing. I just get to sit in some of the fun meetings lol.

I really don't personally find $40 to be a horrible price. But for me personally i don't buy much for connectors for anything that isn't some official distribution of a major company so i'm used to paying for quality.

Unless this thing ends up being a pile of crap, in which case, burn them at the pyre.
 
Ok so cheap chi-na connectors, got it.
How do I plug that big cord in at the wall socket from my power supply?
 
What load and what pin, what power supply ,would be a good start to figure this out.
 
Back
Top