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

[Endymio]

I work on cables all the time.

My mechanic works on engines all the time. He's never designed one.

Because household wiring is inside walls made of wood and other flammable materials ... There are various causes and points of failure,

I based my figure on a prior estimate that 50% of home house fires are due to electrical appliances/devices, heir power cable, or the receptacle, not to internal wall wiring. Not that it's relevant. If that inside wiring didn't overheat due to a fault somewhere, it wouldn't matter how much flammable material was nearby.

....there is exactly one connection type that we’re actually taking about here, and it’s a newborn compared to what you keep trying to bring [up]

That's exactly my point. Home wiring has had more than a century to get this right; that standard has been revised many, many times, and we still get millions of these fires each year.

You’re asking us to be equally outraged by household wiring issues.

No, I'm asking for you to apply critical thinking. Fifty amps is a lot of current in a small package, and a tiny handful of failures is not indicative of a systemic design fault, particularly when half of them are from known damaged or out-of-spec cabling.

 

[ty_ger]

My mechanic works on engines all the time. He's never designed one.

Okay?
But he has received training, right?
And has reference materials, right?
And he has special tools, right?
And he can follow instructions in order to reach the end result, right?

It's the same. You use special tools and special materials and follow the product design doc and datasheet to achieve the end result.

This spec, just happens to suck.

 

[R-T-B]

Your wall socket if you live in NA is capable of like 1200w, maybe 1500.. and 12 amps. 12 of em.

15a actually, and for most new installs its 20a.

 

[freeagent]

15a actually, and for most new installs its 20a.

Well that's what I meant

I hate math.

 

[eidairaman1]

I work on cables all the time.

Ever seen 0000?

Heavy heavy

Just throwing this out there.. because maybe it needed to be mentioned.

Your wall socket if you live in NA is capable of like 1200w, maybe 1500.. and 12 amps. 12 of em.

Typical US outlet circuit, 110-125VAC, 15AMPs= 1875Watts.

PIE. 125*15

 

[LittleBro]

So, we've seen a total of SIX confirmed issues (10 total suspicious/unconfirmed). While this is a problem, I'd say it isn't as rampant as some youtubers and forum members make it out to be........

10(6, really) out of THOUSANDS of cards in the wild. I'm more worried about Gigabyte's GPU gloop and my mount in a SUP01 than this, lol.

Be aware of a fact that Reddit threads regarding Nvidia are very strictly moderated. Many people posted issue there and in an hour link to post was not working.
So, they might be a bit selective on what to publish and what not. In other words, Nvidia-related things on Reddit get often redacted and censored.

Also, many users might not be aware of the melting.

https://overclock3d.net/reviews/gpu_displays/asus-saved-our-bacon-we-had-12vhpwr-12v-2x6-cable-issues/

RTX 4090's 16-Pin Connector Melted After One Year of Usage

We now know it can take years for symptoms to crop up.

www.tomshardware.com

See? Connector melted only partially and thus was not reported immediately. And imagine how many similar cases are out there.

Stop. Just stop with the insanity. The specification requires even flow. Uneven current distribution happens -only- if the cable is faulty or improperly mated.

Dude, even when users followed all speficic steps to properly seat the connector, a shit happenned. I agree, it happens when connector is improperly mated.
But how can you achieve such thing when this connector is specified to be living on the edge? You have no safety factor to compensate for potential mating difficulties.
Users have no feedback whether they reached proper connector mating level. They don't measure resistance between card and PSU. Users fully insert the connector. Yet shit happens.
How can you know that you not only fully inserted the damn connector, but you also achieved damn proper mating level? How? How can users know?
Only very few GPUs are equipped with per-pin current sensing. Let's take ROG Astral for example ... imagine user never installs ASUS GPU Tweak utility.
How would this user know there's a problem with mating? Because without that utility nothing will pop up and tell the user.

Connector is living on the edge, it requires specific precision and conjunction of starts (luck) to be seated properly.
With more robust connector (bigger safety factor) this precision won't be neccesary and larger mating surface would compensate for improperly mated connector.
Imroperly mated that users is not aware of.

And yet there hasn't been one single fire from it. Meanwhile, there are several million home fires a year worldwide (250K in the US alone) due to faulty 120v cables, and tens of thousands of deaths. Odd you don't seem outraged over that.

Look up video from Gamer's Nexus, when they totally melted the connector by normal use. There was a smoke and plastic was bubbling.
Now imagine they don't stop the GPU load, this happens while user is AFK, whole rest of PC case can catch fire.
Are we gonna wait until shit hits the fan when we already know there are issues and we're gonna put blind eye to them just because shit hasn't hit the fan yet?
Safety is about precautions. You're better safe than sorry. You try to predict what can happen and establish safety routines based on prediction outcome.

Lot of residential fires in USA are caused by overloads of circuits, especially overload of wall outlets.
Using multi-outlet extensions instead of wall outlest for appliances such as clothes dryer or stoves is also quite popular in USA.
Why I don't seem outraged over that is because I have my opinion on 120V standard in USA.
But in PC, there's just DC (rhymes, doesn't it) and is used in same way everywhere (global standard).

And what happens when the failure occurs before the GPU? As just one example, a crimp or damaged insulation can cause a ground short within the cable itself, causing extreme flow from one or more pins on the PSU. No amount of GPU-side sensing will detect that.

Sure, it would be great to have per pin sensing even on PSUs. I'd like to have it, for the sake of safety.
There's just per rail sensing. If such connectors as 12VHPWR or 12V2x6 will not cease to exist,
I'm pretty sure we'll see per pin sensing on PSUs in future. PSU will not cost $200 but $500 and more.

This is nothing new and yet this issue only exists with this connector. Pretty obvious what that tells us.

Exactly.

 

[Bomby569]

All I got to say is - if this was a design issue - we should be seeing a LOT of incidents. Like, hundreds - possibly thousands of them. I can't fathom how it's a design issue (which means it should plague ALL cables) with failure rates at 0.05% or something.

there were design flaws in so many products through history, even things as important as bridges for example. Just because they don't all have issues right away doesn't mean the problem of design isn't there. It's very common to go and fix them after the fact so the flaw doesn't end in disaster. Some may fail right away, some may take years or never, it depends on several factors. People don't all use the cards in the same exact way.

Probably one of the most well know cases is probably the automotive, recalls are made all the time for design flaws because some cars fail, sometimes a very very small percentage, but no one is willing to let the rest gamble on the chance the design flaw will turn into a problem or not. In some cases the cards could be scrapped before anything would ever happen.

You're making a absurd connection: because they aren't failing at insane rates, there is no design flaw. That's not a thing.

 

[EarthDog]

Fifty amps is a lot of current in a small package

It's also 12V, not 120V... so there's that.

And imagine how many similar cases are out there.

If a tree falls in the forest and nobody is there to hear it......

It's not good, but, it's still working and, didn't cause a fire..........yet. So, it's, technically, not an issue.

 

[dragontamer5788]

It's also 12V, not 120V... so there's that.

Wires and connectors don't care about voltage. Current is what wires and connectors care about.

Voltage is delivered to the load device. That's why high voltage lines can be 380,000 Volts. You pretty much can arbitrarily increase voltage forever and wires will be fine. (Eventually the voltage gets so high it will punch through the plastic housing. But computer parts don't deal which such voltages).

Computers care about voltages because any more than 20V likely will permanently destroy the fragile nanometer sized transistors and they only operate in the 1.5V region.

 

[LittleBro]

Wires and connectors don't care about voltage. Current is what wires and connectors care about.

Voltage is delivered to the load device. That's why high voltage lines can be 380,000 Volts. You pretty much can arbitrarily increase voltage forever and wires will be fine. (Eventually the voltage gets so high it will punch through the plastic housing. But computer parts don't deal which such voltages).

Computers care about voltages because any more than 20V likely will permanently destroy the fragile nanometer sized transistors and they only operate in the 1.5V region.

Actually, there is voltage limit for wires and connectors because of chance for electromagnetic induction (surge build-up) or other unwanted interference. There's a minimum spacing distance required for pins in a connector for given voltage. This minimum spacing safety distance is applied to international overhead lines as well. It's a complex topic regarding electromagnetic fields, surge and flashovers. Each connector should have stated max. operating current and voltage.

 

[ty_ger]

I don't know if someone else already shared this somewhere, but this is pretty cool.

 

[HTC]

I don't know if someone else already shared this somewhere, but this is pretty cool.

Lipstick on a pig ...

 

[ty_ger]

Lipstick on a pig ...

Sure, but it does provide a solution to a problem so many want to pretend doesn't exist.

And I like how easily he provides an example of why balancing the power from the PSU side isn't effective or practical.

 

[eidairaman1]

How about gpus being limited to 300 Watts in the consumer sector, make these companies get creative with designs.

 

[HTC]

Sure, but it does provide a solution to a problem so many want to pretend doesn't exist.

And I like how easily he provides an example of why balancing the power from the PSU side isn't effective or practical.

NOT a solution.

A solution would be to COMPLETELY REWORK the 12v high power connector.

 

[ty_ger]

NOT a solution.

A solution would be to COMPLETELY REWORK the 12v high power connector.

Of course. But it is a solution to stuff burning. If the current exceeds the pin rating, the PSU shuts off. It's automatic, hands off, and stops a problem in its tracks.

 

[Sunny and 75]

the PSU shuts off.

and takes my gameplay progression along with it, great!

Spoiler

 

[eidairaman1]

Of course. But it is a solution to stuff burning. If the current exceeds the pin rating, the PSU shuts off. It's automatic, hands off, and stops a problem in its tracks.

I presume ocp, opp, ovp are on the input side of the psu and not on the output

 

[Frizz]

Sure, but it does provide a solution to a problem so many want to pretend doesn't exist.

And I like how easily he provides an example of why balancing the power from the PSU side isn't effective or practical.

NOT a solution.

A solution would be to COMPLETELY REWORK the 12v high power connector.

Of course. But it is a solution to stuff burning. If the current exceeds the pin rating, the PSU shuts off. It's automatic, hands off, and stops a problem in its tracks.

Because being the first to have something like this will automatically make this PSU the default go to for anyone using 12VHPWR. This connector is the smartest thing NVIDIA made and its clearly their way of giving back to content creators and PSU manufacturers . Then again, no-one wants their house to burn down.

The solution is awareness as it's technically not a design flaw, don't run the cable at it's max capacity 24/7, undervolt, no aggressive bending etc. etc.

EDIT: also buy a fire extinguisher, smoke detector in PC case etc. etc.

 

[freeagent]

I think the current connector is fine for about 400w, but they should use a different connector if they are planning on exceeding 400w.

Or better yet, use their super massive AI peen to build something that can fully take advantage of a 350w budget.

 

[Frizz]

I think the current connector is fine for about 400w, but they should use a different connector if they are planning on exceeding 400w.

Or better yet, use their super massive AI peen to build something that can fully take advantage of a 350w budget.

550-600w is burn area but also maximum capacity, which is what the 5090's runs at out of the box. Undervolting to 450-500w dramatically and significantly changes connector temps and voltage sag. 2x 12VHPWR connectors limited to 300w would be PERFECT!

 

[eidairaman1]

I think the current connector is fine for about 400w, but they should use a different connector if they are planning on exceeding 400w.

Or better yet, use their super massive AI peen to build something that can fully take advantage of a 350w budget.

Yeah anything over that might as well run a canon plug and have input ampeage control on the gpus

 

[ty_ger]

and takes my gameplay progression along with it, great!

Spoiler

Better than burning 5 minutes later and taking your gameplay progression along with it.

They said another option is monitoring and warning earlier. We'll see what they implement.

It's not a good solution, but its better than the GPU designer has been doing. It's an attempt. It's something Seasonic can do.

And the explanation of why the options on the PSU side are so limited, is helpful.

 

[Hesychius]

Why would anyone use a third party cable and not the one that comes in the box of the gpu or power supply box? Humans make no sense to me.

Ha, yeah... Its almost as if they do it intentionally so they can whine about it happening again on forums, while looking to sell their GPU to a prominent Youtuber for 'analysis'.

 

[boppingXQWB]

So much for user and third-party cable blaming.
First-party MSI cable included with 5090 melted.
Source:

https://www.reddit.com/r/pcmasterrace/comments/1kuf6wu/rtx_5090_power_connector_melted

Alternative source: https://videocardz.com/newz/another-msi-yellow-tipped-power-cable-burns-on-geforce-rtx-5090