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

[TheDeeGee]

Only board partners can save the 5090, but i have a feeling NVIDIA has their hands tied pretty well when it comes to designing the PCB.

 

[Toss]

Only board partners can save the 5090, but i have a feeling NVIDIA has their hands tied pretty well when it comes to designing the PCB.

ofc they have hands tied. They would make 4x 8 pin version and win the race instead
It happened since 4090, 3090 is last gen when they could make connector they want.
If only NVIDIA can dismiss the rule of mandatory 12 pin for GPU

 

[TheDeeGee]

ofc they have hands tied. They would make 4x 8 pin version and win the race instead
It happened since 4090, 3090 is last gen when they could make connector they want.
If only NVIDIA can dismiss the rule of mandatory 12 pin for GPU

I mean, are AIBs even allowed to add load balancing to 12V-2x6, since the connectors they use do allow for that, as all 16 pins have individual solder connections on the board, unlike the FE which is just a 12V and Ground.

 

[Toss]

I mean, are AIBs even allowed to add load balancing to 12V-2x6, since the connectors they use do allow for that, as all 16 pins have individual solder connections on the board, unlike the FE which is just a 12V and Ground.

looks like not. ASUS tried to make shunts to give current info per pin before to comply with designed rule of NVIDIA.
NVIDIA is only to blame here with their stupid rules which make their hands tied.
they made it by design to be unsable and burn as soon as waranity ends. It's planned obsolence by them designed ! Only one mishap that it doesn't even last waranity period, that's what was miscalculated by Nvidia

 

[sethmatrix7]

I think this image needs a new version for the current times

 

[wNotyarD]

As far as I'm aware, AIBs can use a second 12V-2x6 connector. GALAX at least used two with their 4090 HOF and 3090Ti HOF cards. How much they did load balancing, though, I don't know.

 

[Toss]

As far as I'm aware, AIBs can use a second 12V-2x6 connector. GALAX at least used two with their 4090 HOF card. How much they did load balancing, though, I don't know.

I can bet EVGA left because of those rules forced by NVIDIA. They checked that and understood they can't be like shit to their customers

 

[TheDeeGee]

I can bet EVGA left because of those rules forced by NVIDIA. They checked that and understood they can't be like shit to their customers

This may as well have been the case, they didn't want to deal with a lawsuit of a house burning down with possible deaths as a result.

 

[Dredi]

How much they did load balancing, though, I don't know.

At least the one on the right has six shunts, so safe to assume some balancing for pairs of conductors.

 

[Toss]

At least the one on the right has six shunts, so safe to assume some balancing for pairs of conductors.

if they are free do do it, they can literally make 2x connector versions for extra 50$ and win the market. I would instabuy even
MSI Ventus with 2x connectors.

 

[JAS_design]

Also, I found this particular interesting bit on the CEM-5 connector specs:
View attachment 384704

So much for our fellow here saying the connector is good for a single cycle only. But it certainly doesn't look reliable for reviewers' usage at face value.

Because I use crimp terminals on my own projects, I'm always very careful about following the datasheets to the letter.

The requirement to derate when using a connector with a greater number of circuits is common. And crimp terminals of this sort are almost always rated for no more than 20-50 cycles, with 30 being the average in my experience.

It's far more than one use and throw away, but for a reviewer who plugs and unplugs hundreds or even thousands of times, it's definitely not rated for that. That doesn't mean it won't continue to work, it just means there's no guarantee it will. If a reviewer starts to see power or temperature anomalies after a while, something like this could explain it.

MiniFit Jr used on the older 6 and 8-pin connectors is also only rated for 30-50 cycles, but the pins are already running so far below max specification that they're more resilient to the abuse. If the MiniFit Jr was pushing 9.5A+ through every pin, it would struggle as well.

 

[TheDeeGee]

At least the one on the right has six shunts, so safe to assume some balancing for pairs of conductors.

Probably 2x3 per connector like the 3090 had.

 

[taka]

EVGA last unreleased card followed Nvidia design related to power.

Only two shunt resistors visible.

 

[ir_cow]

Step aside, we have real tests now

 

[Dredi]

Step aside, we have real tests now

Too bad that he is measuring current by using shunt resistors, which will actually even out the load distribution somewhat.

 

[_roman_]

#438 / #439

I think he is the dr ... ... guy from igorslab who does his third party psu tests.

I disagree with using shunt resistors.

There are current clamp meters which do not need a shunt resistor. I think it is a wrong measurement setup. Because the shunt resistor - is a serial resistor which makes the measurement scenario not valid anymore. I think there are also clamp meters for oscilloscopes. Those current clamp meters use some special sensor (hall effekt in german) which is better suited in my point of view.

I'm curious how others see the measurement setup of the #438 video.

--

Igor also has again some thesis, which I may read another time and may comment.

Groundhog Day: Der 12V2X6, schmelzende Kontakte und Schieflasten - Was wir wissen und was nicht | Seite 3 | igor´sLAB

Eigentlich hatte ich mir vorgenommen, nichts mehr zum 12V2X6-Connector (ehemals 12VHPWR) zu schreiben. Doch nach den zahlreichen Meldungen und Diskussionen rund um das Thema habe ich mich dann doch…

www.igorslab.de

 

[dirtyferret]

Step aside, we have real tests now

saw that video earlier this morning, a follow up to yesterday's video (below)

 

[Frick]

I think he is the dr ... ... guy from igorslab who does his third party psu tests.

He wrote a lot for TPU too!

 

[LittleBro]

Because someone complained that the card spikes to 900 watts.

Spikes are not the problem here. We're talking constant high current per pin, not miliseconds of high current.

Anyway, new cases of 5090 and 5080:

RTX Cablegate update: First RTX 5080 cable melts, more 5090 cases, MODDIY changes its upgrade recommendation page - VideoCardz.com

Blackwell and Ada are hot right now Daisy chain of expansion cables are not a good idea for RTX 5090 FE Redditor RevolutionaryPea7570 claims that his RTX 5090 Founders Edition experienced a cable burn. While the card appears to be unaffected, the same cannot be said for the PSU cables, which...

videocardz.com

Btw, RX 9070 event planned on 28th if February 2025.

 

[Degreco]

He wrote a lot for TPU too!

Guys, this is Aris, the guy behind Cybenetics (their CEO). Chances are good, they certified the PSU you are using right in this moment (if you use a known brand).

After jonnyguru going corporate (to Corsair) this man is the actual Yoda in PSU knowledge. I would take him seriously and would be interested to see his take on the current situation when he gets his hands on a 5090(FE).

 

[ir_cow]

#438 / #439

I think he is the dr ... ... guy from igorslab who does his third party psu tests.

He is the Cybernetics guy and did TPU reviews as well for a number of years.

 

[Frick]

Guys, this is Aris, the guy behind Cybenetics (their CEO). Chances are good, they certified the PSU you are using right in this moment (if you use known brand).

Yeah I know, I was just keeping it short.

 

[dirtyferret]

I think he is the dr ... ... guy from igorslab who does his third party psu tests.

correct, they seem to repurpose his HB reviews there

After jonnyguru going corporate (to Corsair) this man is the actual Yoda in PSU knowledge.

Actually Aris stated Jon is the reason he got into PSU testing

He is the Cybernetics guy and did TPU reviews as well for a number of years.

along with Kit Guru, PCgamer, Toms, Igor Lab (as roman stated) among others

 

[bitsandboots]

Only board partners can save the 5090, but i have a feeling NVIDIA has their hands tied pretty well when it comes to designing the PCB.

EVGA shaking its head and saying "told you so" to other partners and all users.

 

[Knight47]

#438 / #439

I think he is the dr ... ... guy from igorslab who does his third party psu tests.

I disagree with using shunt resistors.

There are current clamp meters which do not need a shunt resistor. I think it is a wrong measurement setup. Because the shunt resistor - is a serial resistor which makes the measurement scenario not valid anymore. I think there are also clamp meters for oscilloscopes. Those current clamp meters use some special sensor (hall effekt in german) which is better suited in my point of view.

I'm curious how others see the measurement setup of the #438 video.

--

Igor also has again some thesis, which I may read another time and may comment.

Groundhog Day: Der 12V2X6, schmelzende Kontakte und Schieflasten - Was wir wissen und was nicht | Seite 3 | igor´sLAB

Eigentlich hatte ich mir vorgenommen, nichts mehr zum 12V2X6-Connector (ehemals 12VHPWR) zu schreiben. Doch nach den zahlreichen Meldungen und Diskussionen rund um das Thema habe ich mich dann doch…

www.igorslab.de

Clamp meter must be very expensive, since Aris doesn't even have one and Igor only uses two for 6 wires.