RTX 5090 FE Molten 12VHPWR

[zehoo]

Anyone buying a new card that uses the 12VHPWR should really be factoring in buying a thermal grizzly wire view to go with it. I hope AMD aren’t stupid enough to force use of it for RDNA 5.

 

[Vegas P11]

The problem is that he tests a lot of graphics cards and keeps plugging and unplugging them; this, combined with the pins on the awful 12VHPWR connector—which shouldn’t be moved—is a recipe for melting. With the adapter that comes with the graphics card—the one Nvidia recommends—the pins aren’t fully encased in plastic, as they are on some third-party cables, and when the cables move, the pins move—which means the contact surface will always be different.

So, even though the connector is fully inserted, it won’t make a full contact = overheating.

Spoiler: Lian Li connector.

View attachment 805892
View attachment 805894

As Daniel said in the video, he leaves the nVidia connector attached to the 5090 at all times and just removes the card from the computer. He also stated that he does check it to make sure that it did not come loose. Even if that was not the case, the connector design should not be so fragile that it melts if it is even the tiniest bit loose. It is just a bad design for a connector carrying so much voltage.

 

[AVATARAT]

As Daniel said in the video, he leaves the nVidia connector attached to the 5090 at all times and just removes the card from the computer. He also stated that he does check it to make sure that it did not come loose. Even if that was not the case, the connector design should not be so fragile that it melts if it is even the tiniest bit loose. It is just a bad design for a connector carrying so much voltage.

Yes, he said that in the video, but I’m not talking about unplugging the connector—I’m talking about removing the card from the slot.
Look, this “guidance” is given only because the pins shouldn’t move—as I explained above, if the cables aren’t in the “right” position, the contact area will be less, which leads to a problem.



Yes, we all know this needs to be fixed, but there’s no sign of that happening anytime soon.

 

[MaZa]

It's been a while since he got it to happen, so he's not entirely inept. Still, watching some TechTubers annoys me most of the time. I just saw one a few days ago on my algorithm for a laptop, and he hated that the USB4 was on the left side of the laptop instead of the right. I figured he was left-handed, but no, he just preferred it on the right for aesthetic reasons. It's f'ing annoying that USB ports are on the right-side for right-handed people who still like using a wired mouse with their laptop.

Daniel however isn't really a techtuber in traditional sense. He is a school teacher who moonlights on youtube doing PC hardware reviews because it is his passion, not his fulltime job. Refreshingly devoid of clickbait titles.

 

[Vegas P11]

Yes, he said that in the video, but I’m not talking about unplugging the connector—I’m talking about removing the card from the slot.
Look, this “guidance” is given only because the pins shouldn’t move—as I explained above, if the cables aren’t in the “right” position, the contact area will be less, which leads to a problem.


View attachment 805901
Yes, we all know this needs to be fixed, but there’s no sign of that happening anytime soon.

Yeah, I am well aware and I check my 5070ti constantly to make sure it is still seated all of the way although I never move my PC. Even though a 5070ti doesn't carry nearly as much power as the 5090 I am still a bit cautious with the 12VHPWR connector.

 

[TheHig]

My buddy just sent me this pic today. Asus Tuf 4090, Cable Mod adapter melted last night after a year or so of normal use. He bought the 4090 at Microcenter on release and is a build it and forget it guy. He does not know if it’s a revised Cable Mod adapter. Regardless it sucks as the card is out of warranty he thinks and new cards of similar tier are stupidly expensive.

 

[ZeroBarrier]

Daniel however isn't really a techtuber in traditional sense. He is a school teacher who moonlights on youtube doing PC hardware reviews because it is his passion, not his fulltime job. Refreshingly devoid of clickbait titles.

Funny when they guy who doesn't do this professionally is the one to get it to melt when all the professional tech tubers couldn't make the fucker melt no matter what they tried.

 

[GoldenTiger]

Funny when they guy who doesn't do this professionally is the one to get it to melt when all the professional tech tubers couldn't make the fucker melt no matter what they tried.

View attachment 806058

Yep. It's almost like it's an extremely rare defect that is wildly sensationalized to sound common.

 

[Gideon]

Yep. It's almost like it's an extremely rare defect that is wildly sensationalized to sound common.

It's not rare, way to many people have had the issue. It probably affects 3% of the cards sold so far, which would be triple the defect ratio you normally want of 1% or less. However their is no way to know when it will happen and if the older the card gets the more likely it could fail. They need to engineer it to be more robust and make human error not cause a catastrophic failure. It's a weak area and it needs to be improved and it deserves the attention it gets, it would likely get less if Nvidia was not known for sweeping problems under the rug. Nothing is perfect but this plug is a obvious weak link.

 

[Tsumi]

My buddy just sent me this pic today. Asus Tuf 4090, Cable Mod adapter melted last night after a year or so of normal use. He bought the 4090 at Microcenter on release and is a build it and forget it guy. He does not know if it’s a revised Cable Mod adapter. Regardless it sucks as the card is out of warranty he thinks and new cards of similar tier are stupidly expensive.

View attachment 805945

There is a service that replaces these headers, not sure what they charge but surely less than the cost of a new card. If you or your friend knows anyone handy with a soldering iron (or better yet has professional soldering tools), they can replace the header as well.

 

[Ididar]

It's not rare, way to many people have had the issue. It probably affects 3% of the cards sold so far, which would be triple the defect ratio you normally want of 1% or less. However their is no way to know when it will happen and if the older the card gets the more likely it could fail. They need to engineer it to be more robust and make human error not cause a catastrophic failure. It's a weak area and it needs to be improved and it deserves the attention it gets, it would likely get less if Nvidia was not known for sweeping problems under the rug. Nothing is perfect but this plug is a obvious weak link.

Do we know it is 3%, or is that all guesswork?

 

[AVATARAT]

High-temperature long-term corrosion of copper in aerobic environment​

Just for the reference.

Do we know it is 3%, or is that all guesswork?

Officially, probably yes, but what goes to unofficial service centers isn't included.

 

[tunatime]

My 5090 on water would feel warm by the plug. Added a fan and it was still warm on the black plate and by the plug. So I put an old tr water block with a glob of paste right on the back plate and the plug is now room temperature and the core temps actually dropped about 2-4c also. I really think a lot of it is that people run these cards in cases with bad airflow and it gets it to that danger zone where it starts deforming.

That said it's a stupid plug that need to die

 

[sharknice]

My 5090 on water would feel warm by the plug. Added a fan and it was still warm on the black plate and by the plug. So I put an old tr water block with a glob of paste right on the back plate and the plug is now room temperature and the core temps actually dropped about 2-4c also. I really think a lot of it is that people run these cards in cases with bad airflow and it gets it to that danger zone where it starts deforming.

That said it's a stupid plug that need to die

Yeah they should just use usb-c

 

[DanNeely]

Yeah they should just use usb-c

Only worth it if they bump the voltage up. And at that point they can go back to a 6/8 pin style molex again because the amperage will be back in the safe zone again.

 

[tunatime]

Only worth it if they bump the voltage up. And at that point they can go back to a 6/8 pin style molex again because the amperage will be back in the safe zone again.

What funny/sad is the amps per power pin is actually lower on a dual 8pin pulling 600w then on the 12vhp or whatever it's called

 

[littlegreenmenfromhome]

Do we know it is 3%, or is that all guesswork?

He is ignorant and have no sources to back up his claims:
https://www.pcworld.com/article/138...ufacturing-debris-behind-12vhpwr-melting.html

The issues affects “0.05 to 0.1 percent.” that is how many dolts should buy a console, not a PC.


View: https://youtu.be/ig2px7ofKhQ?si=Vco_6eI98L5DMCoz&t=1600

 

[Ranulfo]

He is ignorant and have no sources to back up his claims:
https://www.pcworld.com/article/138...ufacturing-debris-behind-12vhpwr-melting.html

The issues affects “0.05 to 0.1 percent.” that is how many dolts should buy a console, not a PC.


View: https://youtu.be/ig2px7ofKhQ?si=Vco_6eI98L5DMCoz&t=1600

Nice, he sort of blames it on users not plugging it all the way in. I thought there was another video where Steve does put most of the blame on user error. Either way amusing and nice find. The fact that these cables even have this problem makes it a design issue, parts quality, and now we know card design issue on Nvidia's side for power delivery.

 

[Gideon]

Do we know it is 3%, or is that all guesswork?

This was asking a large group of users on failures. However this was based as of October 2024. they continue to happen and the manufacturer is not going to be super honest about the actual failure rate. If they decide it's user error they won't count it as a failure even though they opted to warranty a card for good will. So a solid number to point too, I do not. However the fact it continues to be a issue and you still see new postings of melted connecters, makes me believe the failure is higher than the touted .05 to .10 Nvidia answer touted by some. Also the hotter copper gets the faster it oxidizes and gains resistance which will kill the connecter.

 

[Ididar]

This was asking a large group of users on failures. However this was based as of October 2024. they continue to happen and the manufacturer is not going to be super honest about the actual failure rate. If they decide it's user error they won't count it as a failure even though they opted to warranty a card for good will. So a solid number to point too, I do not. However the fact it continues to be a issue and you still see new postings of melted connecters, makes me believe the failure is higher than the touted .05 to .10 Nvidia answer touted by some. Also the hotter copper gets the faster it oxidizes and gains resistance which will kill the connecter.

I get it, but there's the flipside to the coin that people who have a problem are more likely to respond to surveys than people who have no problem. And, yes, companies will lie as well to hide things they don't like. Either way, I recognize it is a problem but I'm not convinced it is 0.5% or over 3%. I just can't see that anyone has data reliable enough to say one way or the other.

 

[Niner]

I also believe that people are using aftermarket cables just to get that certain look and those are more than likely lackluster cables as well. IMO there are a lot of variables to take into account and I also agree the connector could be better.

 

[AVATARAT]

I also believe that people are using aftermarket cables just to get that certain look and those are more than likely lackluster cables as well. IMO there are a lot of variables to take into account and I also agree the connector could be better.

There’s nothing to believe here—the problem started with the RTX 3000 series and is getting worse with every new generation. There’s plenty of information about these issues.
If Nvidia wanted to, they could fix it, but I’m not sure they have any desire to do so.

 

[littlegreenmenfromhome]

View attachment 806819

This was asking a large group of users on failures. However this was based as of October 2024. they continue to happen and the manufacturer is not going to be super honest about the actual failure rate. If they decide it's user error they won't count it as a failure even though they opted to warranty a card for good will. So a solid number to point too, I do not. However the fact it continues to be a issue and you still see new postings of melted connecters, makes me believe the failure is higher than the touted .05 to .10 Nvidia answer touted by some. Also the hotter copper gets the faster it oxidizes and gains resistance which will kill the connecter.

No poll numbers or link, this is not "evidence" just a useless picture.

 

[littlegreenmenfromhome]

There’s nothing to believe here—the problem started with the RTX 3000 series and is getting worse with every new generation. There’s plenty of information about these issues.
If Nvidia wanted to, they could fix it, but I’m not sure they have any desire to do so.

Citation needed.

 

[AVATARAT]

Citation needed.

View: https://youtu.be/kb5YzMoVQyw

 

[littlegreenmenfromhome]

Nice, he sort of blames it on users not plugging it all the way in. I thought there was another video where Steve does put most of the blame on user error. Either way amusing and nice find. The fact that these cables even have this problem makes it a design issue, parts quality, and now we know card design issue on Nvidia's side for power delivery.

NVIDIA sells | 1 million GPU's with 12VP plug per month, we do not see 30.000 post per month of people crying over it the numbers simply do not match the absurd claims being made in this post.
And yes the number of inciendt is even smaler then the first decile of I.Q. so th epower is seemlingy very muppet-proof.

Numbers of forums are often misleading because if we followed the vocal AMD and Linux users postings, you would think each had 50% or higher number of market share...and not just 5% of sales and 3% user numbers.

 

[littlegreenmenfromhome]

View: https://youtu.be/kb5YzMoVQyw

Timestamp to numbers of failures as data rules and cringe whining sucks.

 

[AVATARAT]

Timestamp to numbers of failures as data rules and cringe whining sucks.

The main issue is with the RTX 5090—what proof do you have that 1 million RTX 5090s have been sold?

Spoiler: Existing solution.

XT150 Connector With 6mm Gold Connector

Specifications：
Item: XT150 Connector
Length: 23mm
Cathode Plug Length: 20mm
Net weight: 25g
Color: Blue / Red / Black (Optional)
Recommend Usage: XT150-F(Battery) / XT150-M(ESC)
Quantity: 1 Pair
Working Current: large current for 100 A, peak 150 AAnode Plug
Size: 6.0mm Bullet Connector
Metal Material: Gold Plated, Brass Gold, Plated Brass
Plastic Material: PA
Flame Class: UL94 V0
Contact resistance: 0.2mohm
Connector size: 6.0mm
Rated Voltage: DC500V
Rated Current: 100A
Burst Current: 120A
Recommend Use Times: 1000 times
Working Temperature: -200C to 1200C

 

[Gideon]

I get it, but there's the flipside to the coin that people who have a problem are more likely to respond to surveys than people who have no problem. And, yes, companies will lie as well to hide things they don't like. Either way, I recognize it is a problem but I'm not convinced it is 0.5% or over 3%. I just can't see that anyone has data reliable enough to say one way or the other.

The failures are mostly going to be on the high wattage cards, like the 4090 and 5090. So you are talking about a small pool of cards vs the entire lineup that has this connecter. Unfortunately Nvidia does not break down sales for each card, so it's a bit hard to know how many are out there. Also plenty of people have issues and don't go to forums to post it. I say it happens too often otherwise this story would have died some time ago. Hopefully they will finally decide to fix it with a new version that evenly distributes the power on each wire or has a thermal monitor at least.

 

[littlegreenmenfromhome]

Timestamp to numbers of failures as data rules and cringe whining sucks.

Jon Peddie Reserch data is not some secret, unlike your ficticious numbers:

NVIDIA consumer desktop GPU shipments per quarter
From Jon Peddie Research (AIB market):
https://www.jonpeddie.com/news/q425...-11-million-units-with-a-cagr-to-2029-of-5-9/

Total AIB market: ~11–12 million units per quarter

NVIDIA share: 92–94%

NVIDIA consumer desktop GPUs per quarter:
→ ≈ 10–12 million units

NVIDIA GPUs useing 12VHPWR / 12V‑2×6
Ada (40‑series):
RTX 4090
RTX 4080 / 4080 Super
RTX 4070 Ti / 4070 Ti Super
RTX 4070 / 4070 Super
Blackwell (50‑series):
RTX 5090
RTX 5080
(Likely) RTX 5070 Ti / 5070

These models represent ~25–30% of NVIDIA’s total consumer desktop shipments.
The rest are 4060/4060 Ti/4050 OEM units using PCIe 8‑pin.

Here are some numbers:
https://videocardz.com/newz/inno3d-...hpwr-rmas-across-europe-since-rtx-4090-launch
https://www.pcworld.com/article/137...-geforce-rtx-4090-12vhpwr-cables-melting.html


Now i predict you come with ZERO data and will just cry, please prove my point

 

[Ranulfo]

Timestamp to numbers of failures as data rules and cringe whining sucks.

Timestamp didn't work but that video shows that this is all or mostly on Nvidia and that this connector unless properly built out really can't handle more than 300w through it. We've seen these problems crop up since the 3090. The fact that we even have to have this discussion is proof there is enough of a problem. For years any power problems were usually bad psu's or bad wires or user error. Now we have warnings not to bend cables too much and Seasonic telling customers (briefly) to use a heat gun or hair dryer to warm the cables up before bending them no more than 20degrees (I forget the exact number).

 

[Tsumi]

Jon Peddie Reserch data is not some secret, unlike your ficticious numbers:

NVIDIA consumer desktop GPU shipments per quarter
From Jon Peddie Research (AIB market):
https://www.jonpeddie.com/news/q425...-11-million-units-with-a-cagr-to-2029-of-5-9/

Total AIB market: ~11–12 million units per quarter

NVIDIA share: 92–94%

NVIDIA consumer desktop GPUs per quarter:
→ ≈ 10–12 million units

NVIDIA GPUs useing 12VHPWR / 12V‑2×6
Ada (40‑series):
RTX 4090
RTX 4080 / 4080 Super
RTX 4070 Ti / 4070 Ti Super
RTX 4070 / 4070 Super
Blackwell (50‑series):
RTX 5090
RTX 5080
(Likely) RTX 5070 Ti / 5070

These models represent ~25–30% of NVIDIA’s total consumer desktop shipments.
The rest are 4060/4060 Ti/4050 OEM units using PCIe 8‑pin.

Here are some numbers:
https://videocardz.com/newz/inno3d-...hpwr-rmas-across-europe-since-rtx-4090-launch
https://www.pcworld.com/article/137...-geforce-rtx-4090-12vhpwr-cables-melting.html


Now i predict you come with ZERO data and will just cry, please prove my point

It's quite simple- the engineering (or lack thereof) of the 12v-2x6 connector is laziness to pinch pennies on $1000+ cards. No self-respecting electrical engineer would tie all the pins to the same bus without some sort monitoring device or load balancer between them. We shouldn't be letting nVidia get away with this- they should be held to a higher standard.

 

[littlegreenmenfromhome]

Still no numbers.

 

[tunatime]

Still no numbers.

I'm an electrician I can make my own numbers based on 25 years of experience what's going to last for years and what's a weak point. This is a weak point. If everything works well we are pushing 7-10amps per wire which is already high for #16 so no matter what Nvidia says they can't change ohm's law and the cable/plug will heat up and degrade over time as it's just to much current though #16 .....then add on a possible 10-20%+ imbalance and it will be outside the safety limit

Put it this way I wouldn't even use #16 in control work Nvidia doesn't get a pas for doing bad work.

 

[littlegreenmenfromhome]

I'm an electrician I can make my own numbers based on 25 years of experience what's going to last for years and what's a weak point. This is a weak point. If everything works well we are pushing 7-10amps per wire which is already high for #16 so no matter what Nvidia says they can't change ohm's law and the cable/plug will heat up and degrade over time as it's just to much current though #16 .....then add on a possible 10-20%+ imbalance and it will be outside the safety limit

Put it this way I wouldn't even use #16 in control work Nvidia doesn't get a pas for doing bad work.

And still no numbers, just anecdotes and I will take the PCI-SIG and electrical/electronics + mechanical engineers over an simple electrician anytime 25 years or not.

When someone got numbers I will listen but until then this is another big nothing burger, backed by zero data.

 

[tunatime]

And still no numbers, just anecdotes and I will take the PCI-SIG and electrical/electronics + mechanical engineers over an simple electrician anytime 25 years or not.

When someone got numbers I will listen but until then this is another big nothing burger, backed by zero data.

Well because you made it personal calling me a simple electrician I personally think you must be retarded and I will trust my brain that understands the physics of electrical over what Nvidia says. Lucky this is the USA and you are free to believe what you want and stay retarded...just don't ues a thermal camera to look at your cable or you might stop believing the Nvidia Kool aid that the plug is well made

 

[CaffeineMan]

View attachment 806819

This was asking a large group of users on failures. However this was based as of October 2024. they continue to happen and the manufacturer is not going to be super honest about the actual failure rate. If they decide it's user error they won't count it as a failure even though they opted to warranty a card for good will. So a solid number to point too, I do not. However the fact it continues to be a issue and you still see new postings of melted connecters, makes me believe the failure is higher than the touted .05 to .10 Nvidia answer touted by some. Also the hotter copper gets the faster it oxidizes and gains resistance which will kill the connecter.

Hey, I filled out this survey when GN asked. It was a quick google doc. I wondered whatever happened with that.

 

[Tsumi]

And still no numbers, just anecdotes and I will take the PCI-SIG and electrical/electronics + mechanical engineers over an simple electrician anytime 25 years or not.

When someone got numbers I will listen but until then this is another big nothing burger, backed by zero data.

How about the NEC (National Electrical Code) then?

NEC 310.4 states only conductors of 1/0 AWG or larger can be connected in parallel. An exception can be made for smaller wires as long as they meet three conditions: A. They are in the same raceway or cable. B. Each conductor can carry the entire load current. C. Overcurrent protection ensures no conductor exceeds its ampacity if one or more parallel conductors are disconnected. nVidia's 12v-2x6 implementation completely fails B and C. Granted, NEC code is for household wiring but I think they would know a thing or two about wiring things properly. They definitely aren't "simple electricians."

 

[littlegreenmenfromhome]

Well because you made it personal calling me a simple electrician I personally think you must be retarded and I will trust my brain that understands the physics of electrical over what Nvidia says. Lucky this is the USA and you are free to believe what you want and stay retarded...just don't ues a thermal camera to look at your cable or you might stop believing the Nvidia Kool aid that the plug is well made

Still no numbers but a lot a hurt feelings.

It is becoming quite the trend in this thread.

 

[littlegreenmenfromhome]

How about the NEC (National Electrical Code) then?

NEC 310.4 states only conductors of 1/0 AWG or larger can be connected in parallel. An exception can be made for smaller wires as long as they meet three conditions: A. They are in the same raceway or cable. B. Each conductor can carry the entire load current. C. Overcurrent protection ensures no conductor exceeds its ampacity if one or more parallel conductors are disconnected. nVidia's 12v-2x6 implementation completely fails B and C. Granted, NEC code is for household wiring but I think they would know a thing or two about wiring things properly. They definitely aren't "simple electricians."

Failure rate data numbers?

Still no numbers.

It is becoming quite the trend in this thread