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

[BoggledBeagle]

this is the stupidest thing I have heard on the internet today I don't know what drugs you are on but consider getting some professional help

The next time please cite a statement you do not agree with and then explain your position and arguments, thank you.

 

[OneMoar]

The next time please cite a statement you do not agree with and then explain your position and arguments, thank you.

Nope you need to not be an idiot to get that privilege from me now get out

 

[Vayra86]

if one wire is doing all the load what do you think it's going to happen? what do you want, one of those industrial wires the size of your arm to solve the problem?

of course the cable is the one that fails, not the gpu, it's the wires on the cable that have to handle to lead not the gpu, but that doesn't mean it's a cable issue.

If i hit you on the head with a hammer, i assume it's your fault for not having a better more robust head, is that how it works? after all the hammer is just fine, your head is the one with a problem

Heavier wire and more connector surface area or multiple connectors or lower power per wire; all valid solutions.

 

[OneMoar]

Why?

This is how the 5090 FE wiring works:
View attachment 384473

So the PSU has 6 connectors passing 12V to the 5090, from a single 12V power rail.

The 5090 then combines all of that back together into rail before it hits the single shunt resistor on the card.

The difference in resistance between 23 amps (0.5 Ohms) and the lowest - 2 amps, (6 Ohms) is equivalent to over 1000 feet of AWG 16, so it would be hard (without a serious manufacturing issue on the PCB it's self) to have this huge increase in resistance. This should also be testable as you can test the resistance from the 12V-2x6 connector on the GPU to the single shunt resistor with a multimeter.

If you aren't seeing this significant increase in resistance between each of the pins, then the fault CANNOT be with the GPU specifically, and must be related to some combination of connection quality across the connectors.

still 100% the fault of the card for being designed in such a way to allow that failure mode.
if you can't design your device to take advantage of the connector you selected guess what thats also a design flaw
there should have been at least two power rails with two control i/cs with sense circuits on all power delivery pins

its not like the issue where the connector wasn't plugged in it was plugged in and still suffered a extreme current bias scenario due to bad power design of the card that has NOTHING todo with the connector thats 100% the card. and this is gonna keep happening on FE cards

 

[A Computer Guy]

I do not like the 12 pin connector design (should be much more robust), but unless it is abused or manufactured wrong, it simply works as intended. I also do not like that only one of these connectors is used on a 600W car.

You do understand that simply doubling the connector isn't a fix right? Without some sensible circuitry to prevent overloading of individual connectors/wires it seems this problem will only get more prevalent (for high wattage GPU's) as connectors wear out overtime. Resulting in melting potential for cards not equipped to detect/handle the problem or cards that can detect the problem will simply stop working as the connection becomes unreliable.

 

[Toss]

Easy fix - 70% Power limiting 5090 and you will get perfect 400W TDP card!

Performance loss is only 10%

 

[JAS_design]

If I had to bet real cash money, I'd say the problem is likely a cascade of fairly small issues which join together to cause a big issue. Take away any of them and the chances of failure will massively decrease.

The GPUs don't balance the load across the individual wires/pins well enough or at all; the GPUs may not have the positive pins connected to a big enough copper plane to effectively cool the pins; the smaller 3.0mm pitch connectors are more susceptible to incorrect installation and/or poor contact, leading to high resistance; last, but by no means least, 100+ watts through a single 3.0mm pitch crimp contact is A LOT no matter what you do.

Even if the 12VHPWR / 12V-2x6 had been designed using the same 4.2mm pitch MiniFit Jr connectors as the original 6 and 8-pin PCIe connectors, I think the huge amount of power being forced through so few pins would still cause some problems. The 4.2mm pitch 6 and 8-pin connectors are rated more conservatively to give a wide safety margin, but if they're pushed as hard as the new connectors...

 

[BoggledBeagle]

An engineer not checking a state of a cable he repeatedly uses for testing should return his diplomma. Seriously.

I just realised, that if he had been checking the cable, he might have learned, how long did the connectors work well. He might have also learned, that the cable was faulty (or less than perfect) from the very beginning. But unfortunately he learned little - only that it does not work now.

I have one positive takeaway from his video though: you can check easilly, if you have some serious current imbalance, just by feeling the cables, if they are all warm the same.

You do understand that simply doubling the connector isn't a fix right? Without some sensible circuitry to prevent overloading of individual connectors/wires it seems this problem will only get more prevalent (for high wattage GPU's) as connectors wear out overtime. Resulting in melting potential for cards not equipped to detect/handle the problem or cards that can detect the problem will simply stop working as the connection becomes unreliable.

Well, doubling the connectors halves a danger of seriously overloading one wire.

 

[TechKilledMe]

I noticed that there is an argument on whether the problem is due to a cable fault, user error, a plug fault, or a socket fault. Have you thought of a design error behind the socket? Buildzoid's video below makes a case that a circuit design error (either weak or no current balancing across the pins in the 12V-2x6 or 12VHPWR connection) on the video card behind the 12VHPWR socket or 12V-2x6 socket is likely to be the main problem instead of the problems being argued about in earlier posts.

The fact there is no current balancing on this connector on the 40 and 50 series is insane. Is there really no way for the GPU to correctly pull amperage from the pins evenly? DerBauer was showing within 5 minutes a massive load imbalance on a few pins causing massive temp spikes.

 

[BoggledBeagle]

DerBauer was showing within 5 minutes a massive load imbalance on a few pins causing massive temp spikes.

His video is irrelevant. He used abused (used to test many GPUs) and damaged cable. Normal people using new or like new undamaged cables will not experience this.

 

[mkppo]

Shocking that it's been three generations and nvidia somehow made it worse every generation instead of learning from it.

This deserves a new award.

 

[BoggledBeagle]

Shocking that it's been three generations and nvidia somehow made it worse every generation instead of learning from it.

This deserves a new award.

Perhaps they presumed that producers are able to make a connector, that works, and that consumers will not damage their cables?

 

[mkppo]

Perhaps they presumed that producers are able to make a connector, that works, and that consumers will not damage their cables?

When a cable supposedly gets damaged with 5 mating cycles as opposed to 30 and is running on the edge even with a perfect installation, the bare minimum they should've done is put two of these for 600W and load balance the cables.

That's learning from your mistakes. Right now, they're making it worse every generation.

 

[wNotyarD]

Perhaps they presumed that producers are able to make a connector, that works, and that consumers will not damage their cables?

One should always assume the user will do the most foolish, dumb and brute things to your product, and harden the project accordingly to take the beating and live on.

It is not what's happening here.

 

[evernessince]

buildzoid already answered that. that is a solved mystery and has nothing to do with the cable.
unless i missed your point entirely

I think you need to really watch the video again then. You can measure all you want on the GPU side, that still doesn't change the fact these cables are subpar for what they need to be doing.

Both of you are correct. The lack of load balancing is 100% an issue but there are likely other factors that contribute to it like the connector durability, cable gauge etc.

Catastrophic failures like this are usually the result of multiple points of failure, similar to the few 7800X3Ds that were bubbling up due to a combination of low quality memory controllers, lack of motherboard safety protections, memory choice, bad quality silicon, etc.

This issue would likely not exist if the cable was higher quality (or greatly mitigated as it would require more of the pins to have poor / no contact) and likewise if there was proper load balancing on the GPU end. Of course this is based on what we know now, notwithstanding any updates we may discover.

Instead of wasting R&D dollars on cables and connectors, Nvidia should be focusing on reducing power consumption in GPUs without losing or lowering their graphical abilities.
Ideally they should be no higher than 300W.

It wouldn't require R&D, Nvidia had much better connectors and load balancing on it's past GPUs. What takes R&D is redesigning a connector over and over because of poor power circuitry and connector / cable design decisions.

I'm throwing my marble in the circle by saying that the 12V-2x6 isn't necessarily a bad design, but it's a HORRIBLE design for what Nvidia is using it on. 12V-2x6 on a 300W card? Not so bad. That's manageable. But we're stuffing it into a 575W card all by itself. Nevermind that by all means, we are an advanced enough society to have protection circuitry on our very expensive electronics, and beyond that redundant protections. There is a reason that PCIe 8-pin lives on as 'more advanced' standards come and go. It does its job. And it does it within its means. And if it can't push enough juice, just put another one next to it so they can share the load. It's reliable, and it keeps designers from demanding unreasonable performance out of a single cable.

The problem is that it's rated for up to 600w. There's also the fact that the cable still puts pressure on the pins, the low number of mating cycles, and there have also been a steady stream of reports of the cable backing out of the connector over the years. I saw a handful of them on the Nvidia reddit just today as people were checking their cable.

There's just too much wrong with the cable itself to the point where even if they dropped it to 300w I'd still vastly prefer to old connector. At that point you aren't even getting the proposed benefits in the first place. Honestly I'd rather just have a motherboard power connector for the GPU.

Easy fix - 70% Power limiting 5090 and you will get perfect 400W TDP card!

Performance loss is only 10%

4090s burned up around that TDP. This would likely only delay the issue, not deny it. You'd have to go even more aggressive then that when power limiting a 5090. After all, you have to consider that the entire 400w can go through a single wire.

 

[Vayra86]

Easy fix - 70% Power limiting 5090 and you will get perfect 400W TDP card!

Performance loss is only 10%

Sounds almost as genius as buying that 14900K to disable your e cores for more performance.

 

[Bomby569]

Heavier wire and more connector surface area or multiple connectors or lower power per wire; all valid solutions.

the wire was so hot that melted the plastic, because there was no balance load and one of the several wires carried all the load.
I'm not saying that a ticc ass wire wouldn't solve it, but it's also not the problem here.

 

[Vayra86]

the wire was so hot that melted the plastic, because there was no balance load and one of the several wires carried all the load.
I'm not saying that a ticc ass wire wouldn't solve it, but it's also not the problem here.

Its mitigation, I agree. But it remains to be seen how all the other 5090's fare.

 

[BoggledBeagle]

Honestly I'd rather just have a motherboard power connector for the GPU.

Yeah, I already mentioned this. I think that Asus should state what are their plans with their design, if they want to plan to use it just for themselves, or make it available for everybody. I presume that they patented it.

 

[Bomby569]

Its mitigation, I agree. But it remains to be seen how all the other 5090's fare.

there, you can have solutions in either the cable, the psu or even the gpu, i agree.
But it still doesn't address the actual problem, I'm just saying not sure the best solution is mitigation and ignoring the actual cause, because the mitigation can fail and you end up in the same place.

 

[qxp]

Possible workaround for the melting problem, which might also explain why some people see the effect and some don't:

Possible workaround for NVidia connectors - use poor quality cables

It might be that the workaround to NVidia connector problems is exactly the opposite of what people are recommending, and short, high quality cables are bad for 4090/5090. As discussed in other threads, the problem with 4090/5090 happens because the wires are supposed to carry large current in...

www.techpowerup.com

It could be that short, high-quality cables are more likely to have uneven current distribution.

 

[Toss]

When a cable supposedly gets damaged with 5 mating cycles as opposed to 30 and is running on the edge even with a perfect installation, the bare minimum they should've done is put two of these for 600W and load balance the cables.

That's learning from your mistakes. Right now, they're making it worse every generation.

cable is getting more mating action than me

 

[Solaris17]

Yo; not really feeling the name calling and rude BS today. I didnt even bother finishing looking at all the reports. Point blank it wont be a delete or a LQ, I'll just send you on vacation. Talk about melted cables and how much nvidia sucks without attacking the person. If you cant do that then I dont want you here and you dont deserve to be.

 

[Toss]

Sounds almost as genius as buying that 14900K to disable your e cores for more performance.

We left no choice, there are no P only variants!

 

[LittleBro]

All the power supplies with cables which are sold are intended to build one computer

Where is this written? I've never seen such thing in manual of any of 5 PSUs I owned.