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

[BoggledBeagle]

Mine ran fine for 1.5 years then melted...

Probably made from mud metal gradually losing grip on the pins.

 

[MxPhenom 216]

Do we know how many of these occurrence are happening from the splitter cable coming with the cards, vs ones people are using that came with their 3.0/3.1 PSUs? And then if there is a rate of failure difference between first gen of the cable vs newer 12v2x6?

 

[jnv11]

OC3D published an article about how 12V-2x6/12VHPWR cables wear out after too much reuse and how Asus's ROG Astral RTX 5090 LC (liquid cooled) detected the issue: https://overclock3d.net/reviews/gpu...ur-bacon-we-had-12vhpwr-12v-2x6-cable-issues/

Dark Side of Gaming found that its RTX 4090 and its bundled 4 x 8-pin PCI Express Graphics to 12VHPWR adapter melted after 2 years of continuous use despite showing no instability issues as seen in https://www.dsogaming.com/articles/...-rtx-4090-after-two-years-of-continuous-work/ .

 

[_roman_]

I expect more from a brand with proper quality management - One more reason to not buy corsair: https://www.corsair.com/br/en/explo...nits/will-my-12vhpwr-cable-work-with-12v-2x6/

OC3D published a ...

8.84 A = green
It's still over 1 Amps unbalanced per pin

We have a serious problem when 9.46A from 8.84A makes it red. 1 Amp difference is not that much from going green to red when comparing upper with lower picture.

Anyway if we talk about only 0.6Amps surplus for going bad - have fun ASUS - Nvidia - Corsair and those brands who sold us these products. I wish them lots of lawsuits.

how 12V-2x6/12VHPWR cables wear out after too much

RMA always
How can a cable which is not more than 2 years on the market most likely wear out?
Whataboutism: Any USB-C smartphone cable has more plug and unplug and bend cycles.

Why are there no one time fuses on each wire of those cables with certain time characeristics? Or inside the psu?

Those psu companies were lazy. think about some proper circuit which only allow a short burst of overcurrent than bring a saftey feature which reset itself. In the old days we had one time fuses with longer time delay. happy time with changing fuses. Some people than looked why the fuse got triggered: https://en.wikipedia.org/wiki/Resettable_fuse

 

[jnv11]

I expect more from a brand with proper quality management - One more reason to not buy corsair: https://www.corsair.com/br/en/explo...nits/will-my-12vhpwr-cable-work-with-12v-2x6/



8.84 A = green
It's still over 1 Amps unbalanced per pin

View attachment 384890

We have a serious problem when 9.46A from 8.84A makes it red. 1 Amp difference is not that much from going green to red when comparing upper with lower picture.

Anyway if we talk about only 0.6Amps surplus for going bad - have fun ASUS - Nvidia - Corsair and those brands who sold us these products. I wish them lots of lawsuits.

View attachment 384891

OC3D said that it reused the Corsair cables many times, and replacing the cable resolved the current imbalance. This suggests that the cables suffered a lot of wear and tear. Normal users don't reuse and abuse their cables like video card reviewers do. OC3D does admit that it reuses cables well beyond their designed reuse counts. 12VHPWR and 12V-2x6 are not designed to be reused anywhere near what OC3D was demanding out of its cables. The higher density of 12VHPWR/12V-2x6 has probably forced manufacturers to have less metal in their sockets and plugs, decreasing their reliability. I believe that the fault lays at the 12VHPWR and 12V-2x6 standards.

OC3D has changed its test procedures to consider 12V-2x6/12VHPWR cables as consumables afterwards to prevent these current imbalances afterwards.

As for the 2 years old adapter cable, I think that what happened is a current imbalance that sat undetected for 2 years melting it.

As for USB-C, it carries far fewer amps, so it can tolerate much more abuse with much less metal than 12V-2x6/12VHPWR.

 

[R-T-B]

Uh huh.

We do a lil hyperbole.

Do we know how many of these occurrence are happening from the splitter cable coming with the cards, vs ones people are using that came with their 3.0/3.1 PSUs? And then if there is a rate of failure difference between first gen of the cable vs newer 12v2x6?

Seriously, we need real stats, not this crowdsourced "pictures from somewhere" garbage.

How can a cable which is not more than 2 years on the market most likely wear out?
Whataboutism: Any USB-C smartphone cable has more plug and unplug and bend cycles.

I mean, this connector has a 50 plug/unplug spec and the other thousands. Like it or not, its not a USB-C connector and never pretended to be.

That rating may have been a big red flag though...

 

[wNotyarD]

I mean, this connector has a 50 plug/unplug spec and the other thousands. Like it or not, its not a USB-C connector and never pretended to be.

That rating may have been a big red flag though...

In comparison, standard Molex Mini-Fit is rated for 75 cycles with tin contacts, and 100 with gold contacts.

 

[phanbuey]

But ok - let's say the contact is bad... or cable goes bad, or i get some crumbs in there... shouldn't the card just shut off?

If I cut a mobo power cable the computer doesn't turn on - why are these able to turn on and then melt?

 

[R-T-B]

If I cut a mobo power cable the computer doesn't turn on - why are these able to turn on and then melt?

I take it you've never seen a melted ATX connector.

There is a distinct difference between cutting a cable and damaging one, also.

 

[sethmatrix7]

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.

So you’re saying that der8auers own RTX 5090 FE setup is improper? That’s the most ridiculous thing I’ve read all day.

 

[ShrimpBrime]

But ok - let's say the contact is bad... or cable goes bad, or i get some crumbs in there... shouldn't the card just shut off?

If I cut a mobo power cable the computer doesn't turn on - why are these able to turn on and then melt?

I think not. It's busy rendering.

When my 110v cable melted running 290x x2 cards benching TimeSpy benchmark, the system was running as the cord was melting.

PSU did not care. I had to physically risk my life and rip the cord out of the PSU as my garage filled with o-zone and smoke.

On that note, what is OCP? Useless then?

 

[Bomby569]

But ok - let's say the contact is bad... or cable goes bad, or i get some crumbs in there... shouldn't the card just shut off?

If I cut a mobo power cable the computer doesn't turn on - why are these able to turn on and then melt?

the wires carry the power, but it's the plastic that melts. There is no plastic melting sensor and the current can flow with the plastic burning

 

[Zyll Goliat]

Guys I found solution for the future Nvidia GPU's

 

[JAS_design]

In comparison, standard Molex Mini-Fit is rated for 75 cycles with tin contacts, and 100 with gold contacts.

With Molex MicroFit (not sure about Amphenol Minitek) there are different terminals rated for different durability cycles. The standard dimpled terminal (lubricated or unlubricated) is rated for 30 cycles at a maximum 8.5A (Link), the unlubricated reduced mating force spring terminal is rated for 40 cycles at max. 5.0A (Link), and the lubricated reduced mating force spring terminal is rated for 250 cycles at 5.0A (Link).

Leaf-spring terminals require less insertion force and have more wiggle room than dimpled terminals, and are therefore more user-friendly, but the tradeoff is that they're usually rated for lower current.

 

[phanbuey]

the wires carry the power, but it's the plastic that melts. There is no plastic melting sensor and the current can flow with the plastic burning

No but there is "I just sent all my amps across this pin" sensor. Cables won't melt if you don't overload them. 3090TI had them that's why those never melted.

 

[jnv11]

I personally think that it might be time to reject the 12VHPWR/12V-2x6 connector as being too dense to be safe for the amps that it plans to feed. The quick and easy fix would be to make EPS12V connectors the high power connectors and to ban the 12VHPWR and its derivatives.

Here are the constraints that I know of:

• Jon Gerow showed that safely implementing a full wattage native 12VHPWR socket that implements all of the optional pins on a PSU would drive up circuit board costs too much due to requiring more layers to safely handle that much current in such high density as seen in http://www.jongerow.com/12VHPWR/index.html . Since no other part of the PSU requires that high quality of a PCB for safety, that would make PSUs much more expensive for little gain. 12V-2x6 sockets have the same problem. EPS12V seems to be the safe density for power plugs on PSUs.
• The discussion at https://www.reddit.com/r/hardware/comments/hrvxd8 seems to show that moving to 48V is not ideal for generating the required voltages for CPUs and GPUs due to lower DC to DC conversion efficiencies, so moving to 48V would make more waste heat on motherboards and graphics cards.
• The old PCI Express Graphics (PEG) 6 and 8 pin power connectors were severely downrated to tolerate 20-gauge input wires probably because some low quality power supplies back when those connectors were designed had 20-gauge wires feeding their Molex connectors, making the PEG connectors way too low density because they can officially supply up to 75 watts and 150 watts. (Many video cards regularly violate these maximums, unfortunately.)
• 12VHPWR and 12V-2x6 have some smarts with sense pins and pins to tell the graphics cards how much power they can draw and whether the card is seated properly for 12V-2x6. Other standards except for the main motherboard connector have no communication whatsoever between the PSU and the powered devices.
• EPS12V maxes out at 300 watts per connector as seen in https://support.exxactcorp.com/hc/e...PCIe-8-pin-vs-EPS-12V-8-pin-power-connections .
• EPS12V and 8-pin PEG connectors are differently keyed, so users who are paying attention should notice that they don't fit together.

While moving to EPS12V would be the easy fix, there could be some opportunity to solve more problems if data pins were added to the EPS12V connector. Since the phenomenons of imbalanced and damaged connectors has become reality, maybe running a data bus like the PMBus as seen in https://en.wikipedia.org/wiki/Power_Management_Bus could help with diagnostics. If the GPU, PSU, and motherboard could share detailed power diagnostics data through the connector, that could could allow the OS, the baseband management controller (usually found on workstations and servers), or the UEFI firmware to be alerted of power problems such as imbalances and power usage limit breaches so they can send warnings to the user or syslog server so that IT management can proactively step in to deal with the power problem before it causes failures or hazards.

 

[R-T-B]

On that note, what is OCP? Useless then?

OCP is only for the DC side.

 

[ShrimpBrime]

OCP is only for the DC side.

Both sides isn't it? Should protect on the AC input as well.

 

[R-T-B]

Both sides isn't it? Should protect on the AC input as well.

No, I'm not aware of any designs that check the AC side like that. Thats because you should never be able to exceed the AC current limit with its DC output before DC OCP kicks in. You can easily use too thin an AC cable and burn up though, especially on 120v. I did it once too in my mining days lol.

 

[Frick]

Both sides isn't it? Should protect on the AC input as well.

How would that work? Your PSU is entirely unaware of your circuit breakers.

No, I'm not aware of any designs that check the AC side like that. Thats because you should never be able to exceed the AC current limit with its DC output before DC OCP kicks in. You can easily use too thin an AC cable and burn up though, especially on 120v. I did it once too in my mining days lol.

Can you even design a PSU that will power down if the AC cable melts? Now that I think about it probably yes with like thermistors...

 

[R-T-B]

Can you even design a PSU that will power down if the AC cable melts? Now that I think about it probably yes with like thermistors...

With enough engineering anything is possible. It'd be stupid to do though, cost vs benefit wise.

 

[ShrimpBrime]

No, I'm not aware of any designs that check the AC side like that. Thats because you should never be able to exceed the AC current limit with its DC output before DC OCP kicks in. You can easily use too thin an AC cable and burn up though, especially on 120v. I did it once too in my mining days lol.

How would that work? Your PSU is entirely unaware of your circuit breakers.


Can you even design a PSU that will power down if the AC cable melts? Now that I think about it probably yes with like thermistors...

So I dont know, I asked Google AI, which I will copy paste here. Not trying to duke it out, but genuinely thought OCP was on both sides.

Here's that copy paste.

Yes, a Power Supply Unit (PSU) typically has overcurrent protection on the AC side, usually implemented through a fuse or internal current limiting circuitry to protect the power supply from damage caused by excessive current draw on the input side; this is sometimes referred to as "Overpower Protection" (OPP) and is separate from the overcurrent protection on the DC output rails.
Key points about PSU overcurrent protection on the AC side:
Function:
If the current drawn from the AC power source exceeds a preset limit, the protection mechanism will activate, often by shutting down the PSU to prevent damage.
Implementation:
This protection is usually achieved with a fuse on the AC input line, or through internal circuitry that monitors the current and activates a current limiting mechanism when necessary.

 

[eidairaman1]

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.

You might say the same for BFG, Visiontek, XFX even

 

[R-T-B]

Not sure I'd trust ChatGPT or whatever on this, but will admit a fuse might be present (haven't seen one in any PSU review yet, but have not seen them all either).

Mind you if you blow a fuse, your PSU is basically dead. But I guess the wire would be ok.

 

[eidairaman1]

So you’re saying that der8auers own RTX 5090 FE setup is improper? That’s the most ridiculous thing I’ve read all day.

He keeps on trying to deflect the problem, nv needs to just own up to this and do a recall