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

[_roman_]

edit: It seems later someone wrote a tech news here about the same topic: https://www.techpowerup.com/332334/unofficial-12v-2x6v-power-connector-melts-nvidia-geforce-rtx-5090

PCGH - https://www.pcgameshardware.de/Gefo...r-Berichte-von-schmelzenden-Steckern-1465621/

I only parsed a bit the user comments.

It seems third party crap cables with ASUS PSU and nvidia gpu.

It seems the source for that article there was: https://www.reddit.com/r/nvidia/comments/1ilhfk0

That 24.99 US dollar cable looks like high value - expensive cable to be used with a ~ 100€ ASUS power supply unit with a 3500€ nvidia graphic card. (when someone knows how much each part is, name, brand and price just comment below. )

 

[D007]

Like I said...

If a $2000+ product cant protect it self from ignorance, idiocy, laziness or whatever you can think of calling it, then its a false design from the beginning.
With just a few extra $$ on the cost of the card it can be done. The video on OP shows this and its very simple.

It’s nice that you said that, I am very proud of you. Lol.
However, acting like a company should design everything it makes to account for modding is delusional IMO.
That’s like saying hey, I bought this nice car, let me go ahead and modify the hell of it out of it and then complain when I blow the engine.
That’s your fault, not the developers.

And the fact that this has been a known problem since the 40 series and people are still doing it, except now they assumed it would be safer with a 600 W power requirement when these were melting with just 400 to 500 W.
There’s just no excuse for it.
People should know better by now.

 

[Assimilator]

That’s like saying hey, I bought this nice car, let me go ahead and modify the hell of it out of it and then complain when I blow the engine.
That’s your fault, not the developers.

More accurately, it's like installing a non-standard part that blows up the engine, then blaming the company that manufactures the car.

And the fact that this has been a known problem since the 40 series and people are still doing it, except now they assumed it would be safer with a 600 W power requirement when these were melting with just 400 to 500 W.
There’s just no excuse for it.
People should know better by now.

It's a known possible problem since the 4000-series that has been addressed with an updated connector... which this person wasn't using.

 

[Darmok N Jalad]

I’ve been around long enough to recall when all CPUs were sold lidless and had no thermal protection. It wasn’t a widespread problem, but you would hear stories of people cracking the die when mounting a cooler, or if you didn’t have proper contact, you smoked the CPU as soon as you pressed the power button. Guess what happened? They started adding lids and thermal protections to CPUs. Now we’re at a place where the thermal monitoring is so advanced that CPUs basically overclock themselves in fractions of a second. What started as mitigations for user error ended up being a benefit to everyone.

And for what it’s worth, not all aftermarket parts void a car’s warranty. Many parts can be interchanged with non-OEM brands, and much of the automotive industry is built around such a model. The non-OEM part just needs to follow the standard.

 

[Zach_01]

It’s nice that you said that, I am very proud of you. Lol.
However, acting like a company should design everything it makes to account for modding is delusional IMO.
That’s like saying hey, I bought this nice car, let me go ahead and modify the hell of it out of it and then complain when I blow the engine.
That’s your fault, not the developers.

And the fact that this has been a known problem since the 40 series and people are still doing it, except now they assumed it would be safer with a 600 W power requirement when these were melting with just 400 to 500 W.
There’s just no excuse for it.
People should know better by now.

Yes using the not proper cable is equivalent to “I’ve modified the hell out of a car”.
Nice example. It’s so nice that you have to exaggerate the hell out of the example to make your point.
A car engine has dozens of parts that can be modified and be points of failure.

Here we are talking about a single point of failure that could have been addressed simply by monitoring it and not let the user load the card unless the proper cable and connection is installed.

If what this user have used is not what should be using then it’s his fault. But on the other hand it’s so easy and simple to protect the product you manufactured from (I said already on post12) ignorance, laziness and stupidity.
But if you don’t care about it, if you don’t care to spend a few bucks to add this simple safety then you keep relay on the user to do the right thing. And hoping to sell him another card maybe.

 

[evernessince]

More accurately, it's like installing a non-standard part that blows up the engine, then blaming the company that manufactures the car.

Except for the part where the third party cables follow the exact same spec as the stock cable.

A spec which by the way was sponsored by Nvidia. These arguments against third-party cables are just illogical. If the spec is bad then so too by extension are the OEM cables that follow that same spec.

It's a known possible problem since the 4000-series that has been addressed with an updated connector... which this person wasn't using.

12V2X6 only addresses partial connections. It doesn't address bend issues or the low safety margin (or non-existent in the case of the 5090).

Given that the user's issue didn't stem from a partial connection, it's essentially irrelevant whether they used a 12V2X6 or 12VHPWR connector. It's a non-factor in this case.

 

[D007]

Yes using the not proper cable is equivalent to “I’ve modified the hell out of a car”.
Nice example. It’s so nice that you have to exaggerate the hell out of the example to make your point.
A car engine has dozens of parts that can be modified and be points of failure.

Here we are talking about a single point of failure that could have been addressed simply by monitoring it and not let the user load the card unless the proper cable and connection is installed.

If what this user have used is not what should be using then it’s his fault. But on the other hand it’s so easy and simple to protect the product you manufactured from (I said already on post12) ignorance, laziness and stupidity.
But if you don’t care about it, if you don’t care to spend a few bucks to add this simple safety then you keep relay on the user to do the right thing. And hoping to sell him another card maybe.

Come on man. Bro used an old cable and he knew he shouldn’t have been using even for a 40 series card and it burnt out on a 50 series card and you’re trying to defend it. You are reaching so far you’re going to hurt something.

Except for the part where the third party cables follow the exact same spec as the stock cable.

A spec which by the way was sponsored by Nvidia. These arguments against third-party cables are just illogical. If the spec is bad then so too by extension are the OEM cables that follow that same spec.



12V2X6 only addresses partial connections. It doesn't address bend issues or the low safety margin (or non-existent in the case of the 5090).

Given that the user's issue didn't stem from a partial connection, it's essentially irrelevant whether they used a 12V2X6 or 12VHPWR connector. It's a non-factor in this case.

These statements are becoming more and more ridiculous. Were you there when they made that cable? Can you verify strict adherence to the specification? No, you can’t.

And even if so, it has been widely reported that the specific cable a lot of these people are using should not be used on 40 series and people continued to use them, even on 50 series cards that had much higher power requirements.
I’m sorry, but that’s just dumb.

I feel bad when people blow up their hard earned money, but to just sit here and act like it’s anybody else’s fault but the user is nonsensical.

“Hey everyone, don’t mess with that crazy guy on the corner, he shoots 50% of the people who talk to him”

Person proceeds to talk to the guy anyway and gets shot…. Who’s fault is it?

OK, I had enough drama for the day. I wish the best to everyone who this happened to and I hope they warranty your stuff.

 

[Vayra86]

Lovely.

Especially the back and forth about what is obviously a heavily underspecced piece of cable to pull 600W through, or even 450W. Its like, when there's a train headed right at you, you're gonna deny it too?

Pfew. I guess idiocracy is real, and cognitive dissonance > common sense. 'But they said its safe'... 'and then deny all responsibility'... 'and then update the connector anyway'. How fucking stupid can you get as a customer?! It gets even better when people suggest AIBs are finding ways to make the cable safer, too, and how great those ideas are. We're really in the land of the blind here simply because we can't stand the simple fact something is underspecced from the get go. Wow.

 

[Zazigalka]

The user chose to use a directly-connecting PSU<->GPU cable, rather than the included 'hydra head' adapter.
Because, logically: Adapter(s)=more impedance/resistance.

Whether flawed-thinking or not, I can 100% see why the user chose a short +12VHPWR/2x6<-to->+12VHPWR/2x6 cable over the included 4x8-pin to +12VHPWR/2x6.
Space constraints aside, it would *seem* to be the better choice, electrically.


(PCIE) 8-pins have burned, yes.
Still, doesn't change the fact that it takes *considerably less* 'going wrong' for +12VHPWR/2x6 to catastrophically fail,
over PCIe 8-pin or EPS 8-pin.
View attachment 384116

how exactly is that "safety factor" calculated ?

 

[lilhasselhoffer]

Let's play google detective.

12VHPWR is a 16-pin power connector that connects graphics processing units (GPUs) to computer power supplies. It was introduced in 2022 to provide up to 600 W of power to GPUs.

In 2023, a minor revision called 12V-2x6 (H++) was introduced. This revision changed the GPU- and PSU-side sockets to ensure that the sense pins only make contact if the power pins are seated properly.

The 5090 is capable of drawing 900 watts for 1 ms, according to Nvidia themselves.

The 5090 is rated at 575 watts "maximum."


All of the above is just the facts. Seems stupid that I say this, when the maximum is at once two different values, but if you play in the land of transient surges it all goes silly. So...you need a power supply that's chonky...which is why Nvidia basically says a 1 kw supply is mandatory. Cool.
You're welcome to quibble about stupid consumers...but if you have a massive burning issue like they did before, you redesign the specification to prevent the thing that caused it. It's almost like the 2023 revision was supposed to do that...but a connector designed for 600 might have problems pulling 150% of its designed power. It's the equivalent of not a modified car, but our nitro-methane fueled dragster shredding a belt because that fraction of a second when you stomp on the gas it literally experiences brutal momentary loads that don't allow you to realistically design for without silly expense might be a bat-crap crazy thing.
Let me also suggest that if they designed this connector for 2x safety factor, IE rating it at 600 watts and designing is for 1200, then your entire PSU would have to be capable of safely pulling nearly 2000 watts if there was anything other than basic connectors with the same safety factor (IE, 600 GPU+ 253 watts for the processor+ drives and motherboard)...and that's just an absolutely silly thing to say. 2000watts/110volts = 18.18 amps...you'd better not have any lights in the house running at the same time, or be on a GFCI. That kind of potential pulls is stupid dangerous...going through a tiny plastic connector. But, why even pretend. People are going to call this a consumer issue no matter what. It was the same thing before...until they redesigned the connector...right? They did that for no particular reason...right?

 

[Visible Noise]

Funny I wonder if you've also blocked fan favorite tech Jesus, HUB, LTT, Jayz2c or the master of BS MLID among others

If you're not selling at least some amount of "lie" you're not really Youtuber!

I’ve blocked all them, plus GN.

 

[Zach_01]

Come on man. Bro used an old cable and he knew he shouldn’t have been using even for a 40 series card and it burnt out on a 50 series card and you’re trying to defend it.

No I’m not…

If what this user have used is not what should be using then it’s his fault. But on the other hand it’s so easy and simple to protect the product you manufactured from (I said already on post12) ignorance, laziness and stupidity.
But if you don’t care about it, if you don’t care to spend a few bucks to add this simple safety then you keep relay on the user to do the right thing. And hoping to sell him another card maybe.

But that does not justify the lack of safety of this $2000+ product, caused by cheapness.
Greed
Save as many bucks as you can!

All in for the profits and everything else including safety can go to hell.

You are reaching so far you’re going to hurt something.

What is that suppose to mean?

 

[Chrispy_]

From 3rd party Cable and was used for the 4090 first. I am not surprised that it melted.

I agree that this is non-news.
The real news will be when the included 4x8 to 12v-6x2 adapters start to melt, or when a regular ATX 3.1 first-party 12v-6x2 cable melts.
That, IMO, is inevitable.

 

[BoggledBeagle]

There is nothing wrong about that connector.

Like everything, the plug can be made badly, for example too small, from too soft plastic, inner metal connectors from wrong metal alloy, which deforms to easilly.

It can be abused by the user as well.

Somebody plugs in poorly made plug in a cramped space, the plug deforms in the socket, the metal inner connector deforms permanently. As long it is plugged the same way, the connector can still work. But if is plugged in a different way, suddenly the metal connector tube over the pin is too loose, the mating surface is too small, current density too high and it warms up a lot.

Do you think it is a coincidence, that the outer connector melted? That one is subjected to highest mechanical stress and got deformed the most.

So please use products only from manufacturers, that put their name on the product, that have some quality management certificate, that can provide some documentation about their products. And then do not abuse and destroy the products by using them improperly.

BTW I hope that Asus will not prevent others from using their motherboard power connector for GPUs, because I believe it is the best and most practical solution.

 

[evernessince]

These statements are becoming more and more ridiculous. Were you there when they made that cable? Can you verify strict adherence to the specification? No, you can’t.

Classic argument from ignorance: https://en.wikipedia.org/wiki/Argument_from_ignorance

The burden of proving the cable wasn't to spec would fall on you as you are the one lodging the argument.

Otherwise, to imply that people had to be in the room is obviously ridiculous.

how exactly is that "safety factor" calculated ?

It's the ratio / percentage of max wattage vs rated wattage. For example 12VHPWR has an absolute max wattage of 675w with a rated wattage of 600w. That gives it a 12.5% safety margin or 1.125 safety factor.

So please use products only from manufacturers, that put their name on the product, that have some quality management certificate, that can provide some documentation about their products. And then do not abuse and destroy the products by using them improperly.

The problem is this has happened to cards with the official adapter on lower TDP cards like the 4090.

A lot of folks bought name brand cables and adapters and still had issues (cable mod for example). At least for those people they got their money back.

BTW I hope that Asus will not prevent others form using their motherboard power connector for GPUs, because I believe it is the best and most practical solution.

Yes, I do like the idea of a just an additional slot and finger on the GPU for power. That said I'd hope it comes with a good safety margin.

 

[D007]

Classic argument from ignorance: https://en.wikipedia.org/wiki/Argument_from_ignorance

The burden of proving the cable wasn't to spec would fall on you as you are the one lodging the argument.

Otherwise, to imply that people had to be in the room is obviously ridiculous.



It's the ratio / percentage of max wattage vs rated wattage. For example 12VHPWR has an absolute max wattage of 675w with a rated wattage of 600w. That gives it a 12.5% safety margin or 1.125 safety factor.



The problem is this has happened to cards with the official adapter on lower TDP cards like the 4090.

A lot of folks bought name brand cables and adapters and still had issues (cable mod for example). At least for those people they got their money back.



Yes, I do like the idea of a just an additional slot and finger on the GPU for power. That said I'd hope it comes with a good safety margin.

Ignorance? Ignorant by definition would be using a product that is known to fail and still using it anyway.
Either you’re ignorant or just plain stupid if you do that.
The burden of blah blah blah lol.
Hey bro, here’s a piece of crap cable, should I use it on my $2000 graphics card? Sure dude, you should just do that because that’s incredibly intelligent. Lol..

 

[evernessince]

Ignorance? Ignorant by definition would be using a product that is known to fail and still using it anyway.
Either you’re ignorant or just plain stupid if you do that.

In your last post and this post you seem to be under the impression that a 12VHPWR cable isn't up to spec with the 5000 series but as a matter of fact the 12VHPWR and 12V2X6 connectors are identical. There was only a slight change to the sense pin and terminal depth. A 600w 12VHPWR cable has equal power handling capabilities as a 600w 12V2X6 cable.

If you can plug in a 12VHPWR into a 12V2X6 GPU, they had better be compatible. That would be one of the dumbest engineering oversights to not change the keying to prevent issues. That's not the case though, they aren't keyed differently because they are identical and the specs back that up.

Hey bro, here’s a piece of crap cable, should I use it on my $2000 graphics card? Sure dude, you should just do that because that’s incredibly intelligent. Lol..

The reddit post provided a link to the cable and it seems to contradict the idea that it's low quality: https://www.moddiy.com/products/ATX...-16-Pin-to-16-Pin-PCIE-Gen-5-Power-Cable.html

It's more expensive than most other cables of the same type and if you look at the features you can see it even is the same more flexible style as the new cables included with the 5000 series.

Scroll down to the bottom of the page, their cable client list includes Nvidia, Apple, Cisco etc. A reddit search seems to only turn up good things for them as well.

The burden of blah blah blah lol.

You can't on the one hand ask for an insane level of proof (like your prior statement that one has to be on the factory floor when the cable was made in order to determine quality) and then not hold your own comments to any standard at all. Zero evidence and your reply to failure to produce any supporting evidence, the double standard is pretty obvious.

 

[D007]

In your last post and this post you seem to be under the impression that a 12VHPWR cable isn't up to spec with the 5000 series but as a matter of fact the 12VHPWR and 12V2X6 connectors are identical. There was only a slight change to the sense pin and terminal depth. A 600w 12VHPWR cable has equal power handling capabilities as a 600w 12V2X6 cable.

If you can plug in a 12VHPWR into a 12V2X6 GPU, they had better be compatible. That would be one of the dumbest engineering oversights to not change the keying to prevent issues. That's not the case though, they aren't keyed differently because they are identical and the specs back that up.



The reddit post provided a link to the cable and it seems to contradict the idea that it's low quality: https://www.moddiy.com/products/ATX...-16-Pin-to-16-Pin-PCIE-Gen-5-Power-Cable.html

It's more expensive than most other cables of the same type and if you look at the features you can see it even is the same more flexible style as the new cables included with the 5000 series.

Scroll down to the bottom of the page, their cable client list includes Nvidia, Apple, Cisco etc. A reddit search seems to only turn up good things for them as well.



You can't on the one hand ask for an insane level of proof (like your prior statement that one has to be on the factory floor when the cable was made in order to determine quality) and then not hold your own comments to any standard at all. Zero evidence and your reply to failure to produce any supporting evidence, the double standard is pretty obvious.

Ugh…
Aftermarket 12vhpwr cables bad..
OEM 12vhpwr cables good.
I can’t anymore lol.

 

[Why_Me]

Not all that many amps.

600W ÷ 12V = 50A. If spread evenly across 6 pins, 8.3A each.

18AWG chassis wire rated for 16A, 16AWG chassis wire rated for 22A. https://www.powerstream.com/Wire_Size.htm

Further reading: https://aorph.github.io/12vhpwr/

Improperly seated connectors resulting in poor mating contact surfaces are most likely the cause of overheating.

This ^^ .. it was always a Darwin thing. Some people just aren't meant to own a new Nvidia gpu.

 

[wolf]

Just want to check... this has happened to one 5090, using an aftermarket cable? Just want to be sure.

 

[dyonoctis]

No, nvidia chose to enforce it... at least with 3000 series they only used it on their own fe card, while aibs stuck to pcie (obviously) - but with 4000 series nvidia demanded aibs use that shitty connector aswell...

Besides the 3090TI, the 3000 series didn't use 12 VHPWR. The Nvidia 12 pin had a simpler design, the sensor pins will prevent you from plugging a 12VHPWR into a 3000 FE

 

[D007]

Just want to check... this has happened to one 5090, using an aftermarket cable? Just want to be sure.

I’m using the OEM connector on my 5090, overclocking to the limits and stress testing and it hasn’t happened to me. Just don’t use aftermarket connectors.

 

[ShrimpBrime]

Just want to check... this has happened to one 5090, using an aftermarket cable? Just want to be sure.

6 pins has melted
8 pins has melted

Oh no!! Don't look now!!

12 pins has melted.

Even if they made a 24pin motherboard connector connect to the fkn card, that would melt too.

I wonder why it's such a shocker.

deutsch connectors are just not great for circuit loads that have high resistance. Not in any application I've ever seen.

It melts, RMA. Get a replacement. I dont see what the big deal is. People RMA their Intel chips cause it blue screened once. Lol.

 

[eidairaman1]

The Nvidia Execs that gave the go ahead for 600W+ cards should be terminated and new blood brought in.

Start with jacket man

 

[freeagent]

It was originally Intel's idea.

Also:
View attachment 384171
Yes, that should at most result in black screen power off/restart and not melting but still.

Guy from reddit was using 1000W PSU with some 3rd party cables. Enough to say he's probably SOL on warranty claim.

According to this picture, a GPU should not get over 450w on a PSU rated for 1000w, and 600w only opens up with a PSU rated for 1200w.

How are the GPUs still getting 600w+? Are all of these guys using 1200w PSUs?