It's happening again, melting 12v high pwr connectors

[Panther_Seraphin]

"Jay also mentioned that Nvidea went to a common rail with the 4090 and newer which in his opinion was a step backward from previous cards where they split the power, amongst the pins .."

This is where I have the biggest problem and is dependant on so many things.

From what I have seen on the FE models the input is basically all a single plane which in theory should mean all pins have equal draw in normal operation. However the problem occurs when there is flaws/failures in pins/connections where the card doesnt care and can then draw the same amount of power through the plane BUT its not coming from far fewer pins that designed for and is where the simple sense pins are near useless from a safety aspect.

It looks like some of the AIBs at least are using multiple shunt resistors which means they can monitor more than 1 single point of power input. This in theory should mean if a connector fails there is a chance the card shuts down before catastrophic damage to the connector/recepticle similar to how cards wont power when 1 or more 6/8 pin is missing on older designs. Its not fool proof as a bad connection will still allow power to be drawn. But will they have this logic in their power circuitry is another question.

 

[evernessince]

To coin a phrase entirely from scratch: "you can't handle the truth". If you have a significant imbalance in individual wire current, then you have a significant imbalance in resistance -- which means the cable pin and/or connector mating is the problem. If current load actually was on a per-pin basis, this issue would be happening on most or all cards.

Now you can blame NVidia if you want for not designing in a larger safety factor, but the fact remains that if you have an in-spec cable properly mated, you won't have issues.

We have already seen numerous examples of OEM cables / adapters melting from people who specifically took extra precautions when plugging in the cable, ensuring the bend is acceptable, ect. Go and take a peak at the Nvidia reddit, lot of people are checking their 4090s (and even some 4080 supers) and finding a damaged connector.

The connector simply cannot handle this level of wattage and it only gets worse as it ages. As has been pointed out multiple times now, there is a complete lack of safety margin and no protections when something does go wrong.

Step aside, we have real tests now

I really wish he had just held out until a more comprehensive video could have been done. No 5090 and I expected some level of deeper dive given his level of expertise. I also expected him to do more testing with the OC to at least simulate higher wattage.

I guess we have to wait for the deeper dives. GN probably has the best shot if they are gathering failed cables, there have been a lot of those popping up lately.

 

[ir_cow]

@evernessince I think he is showing that the melting cables is user error and the PSU melting is probably just a bad unit.

 

[Frick]

The AMD 295X2 had a TDP of 500W; spikes of 600W+ were brief. And even still, there were many reports of melted connectors and cables with that card, no?

Because it exceeded spec, which in that configuration was 375W.

 

[Caring1]

I just made a measurement on my 5080 with a new Seasonic PSU and the results for all the 6 cables are between 4,8 and 5,0 A. Repeatedly. Perfectly balanced.

And if I had a 5090, nothing would change. It would be balanced as well, I am convinced. For a like new cable, there is no problem. Only abused cables have a problem. Do not abuse your cables, people!

View attachment 384740View attachment 384741

You might want to watch this:

Edit, posted before I read the entire thread.

 

[BoggledBeagle]

Jay said: "That is a design flaw".

No, that is a manufacturing flaw. Corsair makes junk.

 

[nguyen]

Took a picture of the 12VHPWR (2x8pin on the PSU side) cable that came with Corsair H1200i PSU, this cable is new, I was going to use it for the 5090

Looks the the 3rd pin from left to right is slightly more recessed than the rest.

I will inspect the stock adaptor and see which one is of better quality

 

[mama]

To coin a phrase entirely from scratch: "you can't handle the truth". If you have a significant imbalance in individual wire current, then you have a significant imbalance in resistance -- which means the cable pin and/or connector mating is the problem. If current load actually was on a per-pin basis, this issue would be happening on most or all cards.

Now you can blame NVidia if you want for not designing in a larger safety factor, but the fact remains that if you have an in-spec cable properly mated, you won't have issues.

Maybe the spec needs review....

 

[Outback Bronze]

Looks like OC3D had some issue's too:

ASUS saved our bacon - We had 12VHPWR/12V-2x6 cable issues - OC3D

Edit: Gees here's another one of a 4090 after 2 years of use: Here's what's happened to the 12VHPWR power cable of our NVIDIA RTX 4090 after two years of continuous work

Shit, I'm gonna have to check my 12VPWR

 

[BoggledBeagle]

Also, that Nvidia stays silent at the moment. Speculation why:
1. The issue is not instantly deniable ('user error').
2. They need more time to understand the issue.
3. They need more time to understand the number of melting event and statistics (that's what you get for paperlaunching - to little statistical information).

They are just waiting for manufacturers to get their sh*t together and stop making junk cables.

 

[Bomby569]

To coin a phrase entirely from scratch: "you can't handle the truth". If you have a significant imbalance in individual wire current, then you have a significant imbalance in resistance -- which means the cable pin and/or connector mating is the problem. If current load actually was on a per-pin basis, this issue would be happening on most or all cards.

Now you can blame NVidia if you want for not designing in a larger safety factor, but the fact remains that if you have an in-spec cable properly mated, you won't have issues.

shit happens in manufacturing, and this happens with reputable cables, proper cables, original cables from good brands.
What happened is that Nvidia decided to play cheap with the hardware, watch buildzoid's video, this would not happen on the old series cards, you "can blame Nvidia" and you should blame Nvidia.

 

[BoggledBeagle]

Looks like OC3D had some issue's too:

ASUS saved our bacon - We had 12VHPWR/12V-2x6 cable issues - OC3D

Does he mention a brand of the spotty cable?

Does he have a contract with Corsair???

 

[LittleBro]

They are just waiting for manufacturers to get their sh*t together and stop making junk cables.

It's not just about junk cables. It's the specification/design. There are also reports of original Nvidia cable getting melted.
There's too narrow margin for user and/or manufacturing errors. 12V2x6 is litteraly living on the edge.

Since RTX 3090 Ti had no problems with melting connectors even with 3rd party junk cables, this argument is not valid.
Also, they had time since melting connectors on 4000 series to do a proper re-design of the connector to increase safety factor.
12V2xž is in terms of size of plastic parts and metal parts same as 12VHPWR. Sensing pins aside, difference is only +0.25mm in length of pins. That is barely a re-design.
If anybody here really thinks that +0.25mm in pin length will save their connector from melting with currents well above 9.5A, well, good luck with that.

No one can blame 3rd party manufacturers or PSU cable makers for the root cause of the problem. It's good there are other manufacturers of that cable apart from Nvidia. I take it you know what Apple charges for their damn stupid cables. 35 € for primitive USB-A to lightning 200cm cable. Now imagine that you need new cable after 10 pluggin cycles and one would cost like $100.

 

[Dredi]

Looks like OC3D had some issue's too:

ASUS saved our bacon - We had 12VHPWR/12V-2x6 cable issues - OC3D

Those numbers would be worse on other GPU SKU's too. The astral model they used has shunts on the board which actually even out the current distribution somewhat by simply existing. It's difficult to say how much, but likely the amps would have been way past 10A on the pins with good connection. The shunts are 2mOhm, so an end of life cable with 1-5 mOhm resistance distribution would become a 3-7 mOhm resistance distribution, so from 500%. increase to resistance to a mere 233% increase from pin to pin.

In effect the shunts would be a change a hypothetical current distribution of 5A 5A 5A 5A 5A 25A to just 7A 7A 7A 7A 7A 15A or so.

 

[N/A]

Took a picture of the 12VHPWR (2x8pin on the PSU side) cable that came with Corsair H1200i PSU, this cable is new, I was going to use it for the 5090

Well the question is did he pull the cable or the connector. because those tiny stoppers can dig in the plastic and get totally squished and give way.

J2c must have been pulling and wiggling the cable all day long. that's why it's recessed like that.

 

[9087125]

Why is logic so difficult for some people?

Because, as science demands through its naturalistic and reductionist doctrines, the human mind is literally an ontological shadow of the biological brain which is itself by design irrational. Logic is actually poorly understood in the everyday world, even among professionals in many fields. The biggest misconception of Logic is the notion it gives us the truth, when barring the problem of whether or not truth is even possible, it doesn't necessarily give us the truth in many situations. That is the role of something much more broader and rigorous. What logic does is tell us which arguments have an invalid form and thus false conclusion. This can only work however if the language being used is very well defined and grammatical correct with proper argumental structure, which as this thread demonstrates usually never is. Logic takes our arguments and maps them to something akin to arithmetic so our arguments must be of perfect quality and well defined, if not, logic can only remian silent, logic cant give us a truth or false answer when people have different definitions and methodologies and refuse to play the game of agumentation or use poor argumentation. To makes matters worse, since inductive reasoning is what we are interested in, we rely on people being able to behave critically and analytically to draw on evidence and facts and conform to the scietific method, which again, as this thread demonstrates people dont do a very good job at. If people fail to understand evidence and fail to put their biases aside and continue to accepot a dogmatic rule they are not playing the game of science.

 

[Degreco]

Jay... just in it for the clicks.
In his first video on the issue two days ago, he tells us to ignore the cable quality argument.
In the next video, he says it could be the cable quality argument.

Then he holds up a Corsair cable on camera. The next day, forums are suddenly filled with "Corsair makes crap," etc.

Laughable. I'm not working for Corsair—just an enthusiast here. That said, my two 4090 builds (HXi1500 and HXi1000) beg to differ. They are working flawlessly with the native 12VHPWR Corsair cable that came with the PSUs.

BTW I did not inspect those cables before installation. And I’m completely sure that if I had, I would have found small differences between them. This is normal. Finding minor manufacturing variations in cables is not a big deal - just grab any two cables of the same standard in your home, inspect them with a magnifying glass, and look at the pins, lengths, or other details. They all have small differences.

The real issue here is design responsibility and quality of 12VHPWR. The connector's designer - Nvidia and PCI-SIG - is responsible for accounting for manufacturing tolerances, building in headroom, failsafes and predicting failure rates. The cable manufacturers then follow the specifications to get the certification. It’s not the tail wagging the dog - Nvidia is telling the industry (PCI-SIG included, I fear) what to do.

So, most PSU makers had to integrate this connector without much say in it. I'm sure Corsair, Seasonic, Asus and other PSU manufacturers (ok, let's say designers, but then again Nvidia is also just a GPU 'designer') saw the issues early on. They also have good engineers and surely didn’t just order the connector from CWT without question. In fact, you can clearly see Corsair’s distrust of Nvidia’s connector in how long they avoided adopting it on the PSU side. Their native cables use PCI-E 5.x 12VHPWR/12V-2X6 on the GPU side but 2x 8-pin PCIe (2x 300W) on the PSU side. Corsair explicitly stated that they did this to mitigate risks they identified in lab testing—user errors being just one of them.

Also, you can clearly see Asus' distrust in the connector (the Astral).

Now, let’s be real: the 12VHPWR and its so-called successor, 12V-2X6 (there’s no real difference between H+ and H++ when everything is properly connected) are fragile standards for 600W. Ironically, the 12VHPWR standard would have been a great thing for SFF builds, where saving space is critical - but at 600W, it’s too unreliable.

PCI-SIG should have spec’d this stuff for 375W max, in my opinion.

 

[BoggledBeagle]

... you can clearly see Asus' distrust in the connector (the Astral).

So, they trust the connector on lower models, and distrust it with Astral? If it is so, they think that the connector is OK, and the cable sensing is just a gimmick to support astronomical price of the Astral cards.

 

[Degreco]

So, they trust the connector on lower models, and distrust it with Astral? If it is so, they think that the connector is OK, and the cable sensing is just a gimmick to support astronomical price of the Astral cards.

I don’t know why Asus differentiates their products or what their reasoning is.

But the fact that they implemented a current readout for each individual pin on the 12V rail of the 12V-2×6 on their fastest card should tell you they’re worried about something. Why else would they go to those lengths if there wasn’t a real concern?

 

[Endymio]

We have already seen numerous examples of OEM cables / adapters melting from people who specifically took extra precautions when plugging in the cable

Simply because you "took extra precautions" when plugging in a cable doesn't mean the cable is up-to-spec, or isn't faulty.

there is a complete lack of safety margin and no protections when something does go wrong.

What "protections" do you want? Individual heat sensors on every cable pin? Can you imagine the cost of that?

What happened is that Nvidia decided to play cheap with the hardware ... you should blame Nvidia.

I'll let you in a little secret you're apparently unaware of. Nvidia makes graphic chips, not power cables. And cable standards aren't set by NVidia:

"...After early issues with RTX 4090 connectors melting, PCI-SIG, the standards organization responsible for the 12VHPWR connector, has now updated it to a new 12V-2x6 connector...."

 

[Bomby569]

I'll let you in a little secret you're apparently unaware of. Nvidia makes graphic chips, not power cables. And cable standards aren't set by NVidia:

"...After early issues with RTX 4090 connectors melting, PCI-SIG, the standards organization responsible for the 12VHPWR connector, has now updated it to a new 12V-2x6 connector...."

you should know it's nvidia that makes the board designs, and it's not a secret, that's why asus couldn't even undo all that crap, all they could do was put a bandaid on it

 

[Dragam1337]

Simply because you "took extra precautions" when plugging in a cable doesn't mean the cable is up-to-spec, or isn't faulty.


What "protections" do you want? Individual heat sensors on every cable pin? Can you imagine the cost of that?


I'll let you in a little secret you're apparently unaware of. Nvidia makes graphic chips, not power cables. And cable standards aren't set by NVidia:

"...After early issues with RTX 4090 connectors melting, PCI-SIG, the standards organization responsible for the 12VHPWR connector, has now updated it to a new 12V-2x6 connector...."

They don't make the connector by themselfs, no, but they do entirely decide on what connectors to use on their gpu's ! And they chose to go from a completely safe design, to what is a fire hazard, and enforce aibs to use that same fire hazard connector !

Jay... just in it for the clicks.
In his first video on the issue two days ago, he tells us to ignore the cable quality argument.
In the next video, he says it could be the cable quality argument.

Then he holds up a Corsair cable on camera. The next day, forums are suddenly filled with "Corsair makes crap," etc.

Laughable. I'm not working for Corsair—just an enthusiast here. That said, my two 4090 builds (HXi1500 and HXi1000) beg to differ. They are working flawlessly with the native 12VHPWR Corsair cable that came with the PSUs.

BTW I did not inspect those cables before installation. And I’m completely sure that if I had, I would have found small differences between them. This is normal. Finding minor manufacturing variations in cables is not a big deal - just grab any two cables of the same standard in your home, inspect them with a magnifying glass, and look at the pins, lengths, or other details. They all have small differences.

The real issue here is design responsibility and quality of 12VHPWR. The connector's designer - Nvidia and PCI-SIG - is responsible for accounting for manufacturing tolerances, building in headroom, failsafes and predicting failure rates. The cable manufacturers then follow the specifications to get the certification. It’s not the tail wagging the dog - Nvidia is telling the industry (PCI-SIG included, I fear) what to do.

So, most PSU makers had to integrate this connector without much say in it. I'm sure Corsair, Seasonic, Asus and other PSU manufacturers (ok, let's say designers, but then again Nvidia is also just a GPU 'designer') saw the issues early on. They also have good engineers and surely didn’t just order the connector from CWT without question. In fact, you can clearly see Corsair’s distrust of Nvidia’s connector in how long they avoided adopting it on the PSU side. Their native cables use PCI-E 5.x 12VHPWR/12V-2X6 on the GPU side but 2x 8-pin PCIe (2x 300W) on the PSU side. Corsair explicitly stated that they did this to mitigate risks they identified in lab testing—user errors being just one of them.

Also, you can clearly see Asus' distrust in the connector (the Astral).

Now, let’s be real: the 12VHPWR and its so-called successor, 12V-2X6 (there’s no real difference between H+ and H++ when everything is properly connected) are fragile standards for 600W. Ironically, the 12VHPWR standard would have been a great thing for SFF builds, where saving space is critical - but at 600W, it’s too unreliable.

PCI-SIG should have spec’d this stuff for 375W max, in my opinion.

100% agree.

I've been wanting to replace my psu, but not buying something till this sh1t show is solved, and a new safe connector is put in use, with new psu's supporting that new connector.

 

[ShrimpBrime]

NVidia is directly responsible for the design and use of the melting 12 pin connectors period.

They are strong arming the industry. The exact reason EVGA stopped doing business with NV.

 

[Bomby569]

I've been wanting to replace my psu, but not buying something till this sh1t show is solved, and a new safe connector is put in use, with new psu's supporting that new connector.

i miss the days when a PSU was just a PSU, look if it's any good and the wat. Now even PSU's are a damn mess

 

[Degreco]

Corsair officially reacted to the JayzTwoCents Video on Reddit:
https://www.reddit.com/r/Corsair/comments/1iovz6i/_/mcnjpig