PCI-SIG prepares to replace controversial 12VHPWR with 12V-2x6 connector

[DuckDodgers]

 

 

Quotes

Quote

The design alterations to the new connector appear to be predominantly mechanical, including modifications such as the recessing of sense pins, aimed at ensuring a reliable contact prior to escalating power demands from the GPU. Encouragingly, the draft specification denotes that the 12V-2x6 connector will maintain backward compatibility with existing 12VHPWR connectors.

 

The proposed design introduces new power modes for sense pin detection. In addition to the established 450 and 600 Watt modes, the 12V-2x6 connector will feature 150 and 300 Watt modes. Although each pin is specified to accommodate a minimum of 9.2 Amps, the draft makes reference to two distinct classifications, H+ and H++, with no further elucidation regarding their distinctions.

 

Finally an admission that the root problem wan't the user but a botched design.

 

Source: https://www.igorslab.de/en/rest-in-peace-12vhpwr-connector-welcome-12v-2x6-connector/

[leadeater]

Wow who would have thought that open circuit defaulting to device protection is the proper and safety first design.

[jagdtigger]

Still say this connector is a solution seeking its problem to solve.  It just makes thing more confusing when a card doesnt perform as it should, not to mention the flawed design that allows it to melt like crazy. Just use the old connectors, or maybe manufacturers could switch to the EPS connectors thaat AFAIK can deliver more than the pcie variant....

[starsmine]

1 hour ago, DuckDodgers said:

 

Finally an admission that the root problem wan't the user but a botched design.

 

I am lost as to where that was admitted to.  Like I found NOTHING in terms of saying it wasn't user error. 

 

Sense pins look to be the same almost the same as the old ones except open open and adding in short, so the difference is the latch

OLD sense pins

Sense pins are now harder to make contact, you have to go deeper, its now 2mm rather than .45mm

[Doobeedoo]

While good though, I do like the idea of no cable but a slot power from mobo. Obviously needs to be standardized.

[dogwitch]

OG design was a safety hazard... nvidia test benches proved that in white paper.

[LogicalDrm]

@DuckDodgers

Please expand your own thoughts and add bit of summary on the subject. Your own contribution should be more than one sentence to get pass guidelines.

[Quackers101]

still find it comical that people said colored plastic was a "innovating"
still doesn't seem like the change I wanted to see and same little tab to hold it.

[vetali]

Would've liked to see a better locking mechanism that ensured a flush latch that didn't have play. There is mention of of axial load rating, but doesn't mention if its improved and it looks the same on the renderings.

[NumLock21]

I'm glad they're redesigning the connector, rather than sweep it under the rug, and keep on insisting it's User Error, as parroted off by so called "enthusiast" out there, with some sense of elitism.

[Distinctly Average]

I've always hated the additional cabling at the top of the card. Not only does it look untidy but the 180 degree change of direction required always seemed daft. When I was designing electronics that would have been thrown away at an early stage. It is doomed to fail. It would have made much more sense to at least have the sockets mid board and facing wither to the rear or straight down. I am also sure there are tried and tested connectors that would have done a far better job.

 

The amount of power required by modern GPUs seems excessive too. It feels like design has taken a wrong turn somewhere. Instead of optimising it feels like they simply thrown more cores at it. That and the weight load required to keep them cool and I wonder if it is time for the slot to have a re-design. Something that can better support the card as well as supply enough power without the need for cables on the card.

[starsmine]

42 minutes ago, NumLock21 said:

I'm glad they're redesigning the connector, rather than sweep it under the rug, and keep on insisting it's User Error, as parroted off by so called "enthusiast" out there, with some sense of elitism.

As mentioned above. 
Where and when have they said it was not a user error?

Making a revision does not magically mean that it's not user error, nor does it mean an admission of guilt.  

[Albal_156]

Seem to remember that the melting issues were user error combined with a design flaw in the connector itself.

[SkyHound0202]

At some point they should just adopt the 8-pin EPS12V (i.e., the CPU power connector rated for 336 watts maximum) for consumer graphic cards as well. It's already being used in server graphic accelerators (Tesla) and professional graphic cards (Quadro) anyway.

 

8-pin PCIe power connector (150 watts max) is nearly two decades old and won't cut it in the current hardware landscape.

[starsmine]

56 minutes ago, SkyHound0202 said:

8-pin PCIe power connector (150 watts max) is nearly two decades old and won't cut it in the current hardware landscape.

Trying to figure out why two decades old means anything here. wondering when the laws of electronics changed. V=IR has been constant I thought.

[Uttamattamakin]

As I wrote long long ago putting a lot of power through a small connector is like piping a lot of high pressure water through a hose.  As long as everything is connected right it can work but if ANYTHING is a little off there are PROBLEMS.  A lot of energy packed into a small place tends to either work perfectly or go terribly catastrophically wrong. 

IMHO we saw the real final solution for this at computex

This is what is needed.  Something like this.  Big THICC connectors that will channel as much energy into the card but not concentrate it.  All the talk about contacts and details of electrical engineering simply restate the basic physical properties of the situation in a specific form.  

The redesign will help but there is a fundamental physical flaw in putting so much load on one small connector. 

[Mark Kaine]

9 hours ago, starsmine said:

user error,

while its somewhat understandable that a random forum poster would parrot that nonsense,  this is exactly the reason i blocked GN, who said something like "this could be easily avoided by making a pin longer for better contact,  *therefore* it is "user error"..." parroting the exact same nonsense despite the huge contradiction

[Mark Kaine]

10 hours ago, Distinctly Average said:

I've always hated the additional cabling at the top of the card

the funny thing is nv cards have the option to have the power connector on the side (at the end of the card basically)  but AIBs somehow never choose that option (afaik there is *one* kingpin card and a few workstation cards that do tho)

[jagdtigger]

3 hours ago, Uttamattamakin said:

IMHO we saw the real final solution for this at computex

Not really a solution, piping 600W of power through the mobo will be quite expensive. Not to mention everyone will have its own version of it.....

[Franck]

4 hours ago, jagdtigger said:

Not really a solution, piping 600W of power through the mobo will be quite expensive. Not to mention everyone will have its own version of it.....

Or you put only what's needed at the different watt bracket and make a SKU for each extra motherboard configuration and when people upgrade video card they have to buy a new mobo too with more power delivery on the slot.

 

Motherboard manufacturer be like :

 

[leadeater]

7 hours ago, Mark Kaine said:

the funny thing is nv cards have the option to have the power connector on the side (at the end of the card basically)  but AIBs somehow never choose that option (afaik there is *one* kingpin card and a few workstation cards that do tho)

Nvidia "guidance" is Geforce cards do not have power connectors on the end because that makes them more easily used in server chassis, especially3U where a top connect results in the plug connector and cable protruding outside the chassis. Same deal for horizontally mounted GPUs in risers, top connectors often result in impossible to used.

 

The EULA for Geforce is "not used in the datacenter" and it's encouraged to not create designs that assist in being able to ignore that.

[starsmine]

6 hours ago, Mark Kaine said:

while its somewhat understandable that a random forum poster would parrot that nonsense,  this is exactly the reason i blocked GN, who said something like "this could be easily avoided by making a pin longer for better contact,  *therefore* it is "user error"..." parroting the exact same nonsense despite the huge contradiction

Why anyone calls it nonsense is nonsense. 

lets bring it to an extreme, negligent discharge is ALWAYS user error. some guns are easier to negligently discharge than others, but at the end of the day the one being negligent is the user. The fact that the manufacturer could make it harder to negligently discharge a gun, does not mean the discharge was not negligent on the user part. 

Every single case of melting 12vhpwr has been shown to be user error. 

Let's take a different perspective, the world will always invent a better idiot. 
That is not to say the users who made the errors in this case are idiots but more to say you can not manufacture something and not have a user find a way to error it.

 

The only way to say it's not a user error is for the user to have done everything reasonably right and it still failed. Not inserting the cable connector go into its receptacle all the way IS user error. Torquing it out IS user error. For example, no one plugs in to their outlets at home, have it half way hanging out and things "yea thats fine"

Making it harder for the user to error is fine, but its a stupid take to make the claim that a revision is saying it was user error. If making revisions is admitting to culpability, then no one would make revisions on products that have had users make errors because you are admitting culpability.

 

This revision by pushing the sense pins back 1.5mm increases the safety margin of how much a user can fuck up and it not go to shit. It does not say, oh it was not user error that caused it to go to shit before, it was our connector. 

[Brooksie359]

On 7/3/2023 at 3:47 AM, starsmine said:

I am lost as to where that was admitted to.  Like I found NOTHING in terms of saying it wasn't user error. 

 

Sense pins look to be the same almost the same as the old ones except open open and adding in short, so the difference is the latch

OLD sense pins

Sense pins are now harder to make contact, you have to go deeper, its now 2mm rather than .45mm

Generally things shouldn't be designed in a way where easy small mistakes can lead to your gpu setting on fire. Also not realizing you didn't plug something in all of the way is fairly easy to do and I feel like calling it user error makes it sound like people are making big mistakes which lead to the issue. Ideally if the thing wasn't plugged in all of the way the gpu wouldn't power on. 

[Brooksie359]

7 hours ago, Mark Kaine said:

while its somewhat understandable that a random forum poster would parrot that nonsense,  this is exactly the reason i blocked GN, who said something like "this could be easily avoided by making a pin longer for better contact,  *therefore* it is "user error"..." parroting the exact same nonsense despite the huge contradiction

I mean it was user error. I think a big point that GN was trying to make is that if you use the connector properly you shouldn't have any issues. If the cause of the issue was something that could happen when used properly then it would be a huge deal and nobody should buy the newer cards with such a connector. Now yes I do agree that user error does seem a bit loaded for a mistake anyone could easily make. Anyways I think the reason why GN called it user error was to make it clear that your PC won't start on fire if you plug the power connector in all of the way. 

[IkeaGnome]

1 hour ago, Brooksie359 said:

I mean it was user error. I think a big point that GN was trying to make is that if you use the connector properly you shouldn't have any issues. If the cause of the issue was something that could happen when used properly then it would be a huge deal and nobody should buy the newer cards with such a connector. Now yes I do agree that user error does seem a bit loaded for a mistake anyone could easily make. Anyways I think the reason why GN called it user error was to make it clear that your PC won't start on fire if you plug the power connector in all of the way. 

I think the term everyone is looking for would be along the lines of “manufacturer induced user error.”  
Yes, if it’s plugged all the way in properly there’s almost no chance of melting or catching fire, but the manufacturer made a connector that’s easy to not plug all the way in or stay all the way plugged in.(Poor latching mechanism)

 

Not everything has to be just one person’s fault. Not saying you’re saying it is. Just saying this to people here in general. It’s become habit for me to check the connector for my 4090 fairly regularly. I plugged mine in before I put the card in my computer and it did take much more force than 8 pin connectors, but there was a very audible and tactile click when it seated all the way. 
With how overly careful some people are with hardware I almost wonder if this is part of where the issue on the users side comes from. 
On the manufacturers side, it shouldn’t have been designed in a way that takes significantly more force than people are used to.