Question 12-pin GPU power connector, of NVidia's own design? (Non-Standard!)

[VirtualLarry]

NVIDIA Confirms 12-pin GPU Power Connector

www.anandtech.com

Cue PicardWTF.gif

I mean, I get the idea behind it, and perhaps it's time, but ... I would much prefer this to be a product of the PCI-SIG, or whomever is in charge of STANDARDS in this area, rather than a (potentially patentable) Nvidia "invention". (That would see AMD having to pay licensing fees, should this catch one, and PSU makers switch to the NVidia standard.)

Yet another way that NVidia is "infecting" the "gaming" ecosystem with proprietariness.

Edit: The distinct possibility exists, should this "catch on", that there will then be "Nvidia" and "AMD" PSUs, much like "G-Sync" and "FreeSync" monitors. And this will likely inhibit Intel's 12VO PSU standard from catching on as well. (Or maybe, it will help? Combined 12VO and NVidia GPU PSU connector PSUs?)

 

[BFG10K]

This was discussed weeks ago in the Ampere thread.

 

[VirtualLarry]

Sorry for the re-post. Mods, please close, and clean the title.

 

[Martimus]

Sorry for the re-post. Mods, please close, and clean the title.

Why though? The Ampere thread is quite long, and this is a discretely seperate subject. I would prefer not to go through hundreds of posts on that thread to find information on this topic when i don't really care much about the rest of the stuff on that thread.

 

[Stuka87]

Its not a proprietary connector, its a Molex connector and is expected to become the standard power connector on 12VO PSU's.

Once its here, it will actually make choosing a PSU much easier. No more trying to figure out if it has dual 6pin, 6+2, single 8, or dual 8, etc. Just one connector (which is smaller than a single 8pin) for everything. You just worry about the PSU wattage, which we already have to be aware of.

 

[NTMBK]

This was discussed weeks ago in the Ampere thread.

I for one missed it- that thread is a bajillion pages long, and full of random tangents, so I haven't read every single page of it! I think it's worth having a separate thread for this, for people who haven't tracked every post in that thread.

If @Stuka87 is right and this is just a standard Molex connector that AMD can also start using, then this is a good move. Multiple power cables always felt like a hack.

 

[Stuka87]

I for one missed it- that thread is a bajillion pages long, and full of random tangents, so I haven't read every single page of it! I think it's worth having a separate thread for this, for people who haven't tracked every post in that thread.

If @Stuka87 is right and this is just a standard Molex connector that AMD can also start using, then this is a good move. Multiple power cables always felt like a hack.

Yeah, the majority of people won't even look at that monster thread.

The connector is Molex Micro-Fit 3.0 Connector. So its not proprietary, which is good for us. I expect we will start to see it in OEM machines starting this fall. But most likely be a while before we start to see 12VO mobo's and PSU's. LTT did a little preview using a PSU and motherboard that were sent by the manufacturers for testing that is worth watching, but no mention of the 12pin connector on that specific video. Though he does mention it in another video.

 

[MrTeal]

You know, looking at it again it's likely not a standard 12 pin Microfit. The plug and keying looks standard which will be really nice for cable modding, but the receptacle doesn't look to be 12 circuit.


There are only four through holes on the board for the connector, so it would seem that all the pins except possible sense lines are paralleled on the connector itself, rather than having individual pins into the PCB.

 

[Stuka87]

You know, looking at it again it's likely not a standard 12 pin Microfit. The plug and keying looks standard which will be really nice for cable modding, but the receptacle doesn't look to be 12 circuit.


There are only four through holes on the board for the connector, so it would seem that all the pins except possible sense lines are paralleled on the connector itself, rather than having individual pins into the PCB.

Thats because they are mounting it vertical. And its only a single voltage. So one 12V and ground per side. This is not uncommon for vertical mounts.

 

[MrTeal]

Thats because they are mounting it vertical. And its only a single voltage. So one 12V and ground per side. This is not uncommon for vertical mounts.

Well yeah, but it's not a standard Molex part. A 12 pin arranged like that is pretty uncommon, and likely why Nvidia is saying they custom designed the part.

 

[Stuka87]

Well yeah, but it's not a standard Molex part. A 12 pin arranged like that is pretty uncommon, and likely why Nvidia is saying they custom designed the part.

Molex will make the connector with any number of pins that the customer wants, it was designed to be configurable. That doesn't make it proprietary. Many companies work with Molex on new connector types, and it would not be surprising if nVidia was involved Micro-Fit 3.0 (along with other companies).

My point being, anybody can use this 12pin connector, and I would not be surprised if it makes its way into PCI-E or ATX standards.

 

[MrTeal]

Molex will make the connector with any number of pins that the customer wants, it was designed to be configurable. That doesn't make it proprietary. Many companies work with Molex on new connector types, and it would not be surprising if nVidia was involved Micro-Fit 3.0 (along with other companies).

My point being, anybody can use this 12pin connector, and I would not be surprised if it makes its way into PCI-E or ATX standards.

I don't believe I said it was proprietary. I said it's not a standard part and you can't buy one off the shelf, so Nvidia likely did create the specification for the footprint when they ordered it to minimize the board space. It's standard microfit keying so anyone can use an off the shelf receptacle to plug into it or could use a standard 12 pin header, but unless Nvidia and Molex agree to it it doesn't seem you can buy this header anywhere.

Regardless, going with this design versus 12 independent pins does have some thermal and current implications, especially if one of the pins is still a sense line.

 

[Stuka87]

I don't believe I said it was proprietary. I said it's not a standard part and you can't buy one off the shelf, so Nvidia likely did create the specification for the footprint when they ordered it to minimize the board space. It's standard microfit keying so anyone can use an off the shelf receptacle to plug into it or could use a standard 12 pin header, but unless Nvidia and Molex agree to it it doesn't seem you can buy this header anywhere.

Regardless, going with this design versus 12 independent pins does have some thermal and current implications, especially if one of the pins is still a sense line.

Yeah, and I think an AIB would go with a horizontal mount, like we typically see for 6 and 8 pin connectors. I think nVidia went with the vertical mount because of the oddball design of the RTX 3800 and 3900.

Besides the potential current limitations, there is also the physical limitations. With it on end, the chances of it being bent while somebody is plugging into it greatly increases.

 

[GodisanAtheist]

Wonder if stuff like this will be used to cripple the resale/used value on cards as well to ensure maximum sales on the new stuff. If the connector/pinout is truly proprietary, that makes the cost of picking up a used part the part + a new PSU that supports a potentially transitory PSU connector type.

Might not happen on this gen, but used NV sales are probably NV's biggest competitor at this point. Only makes business sense to clamp down on that market like game makers did with day 0 DLC and eventually digital downloads.

 

[A///]

I'd be surprised if 12VO makes it out of the OEM market into the DIY one.

 

[CP5670]

Some pictures of modular cables for this connector have been posted, but they are supposed to plug directly into the PSU. The PSU-side connectors are not standard and incompatible between brands or even individual product lines from the same brand. So each PSU company will need to sell an adapter for all of their PSUs, which many of them won't do. They would prefer to sell you a whole new PSU instead. Hopefully there will be some AIB cards that use the standard PCIE connectors.

 

[Muadib]

Some pictures of modular cables for this connector have been posted, but they are supposed to plug directly into the PSU. The PSU-side connectors are not standard and incompatible between brands or even individual product lines from the same brand. So each PSU company will need to sell an adapter for all of their PSUs, which many of them won't do. They would prefer to sell you a whole new PSU instead. Hopefully there will be some AIB cards that use the standard PCIE connectors.

This is how I took it as well. If Nvidia can't get all PSU companies to get on board, this new connector will fail. I for one already have a 850 PSU, and I'm not looking to buy another any time soon. I haven't bought an AMD video card in decades, but this could change that.

 

[EliteRetard]

I desperately hope the 12VO PSU "standard" dies in a fire right now.
I'd gladly take adapter cables for every GPU over that garbage standard.

10 pin + 6 pin, 8 + 6/8 EPS, plus additional connectors and massive PCB space for the second PSU on MOBO.
To replace a common standard 24pin that takes FAR less space and doesn't require a new PSU on the MOBO.

If we need a new standard MOBO connector, how about we create one that's not absolute trash.
If Nvidia can shrink 2x8 into a small 12 pin, why can't we shrink the 24 pin and eliminate all EPS?
Even if the connector stayed the same size physically, but just got rewired to better suit modern/future systems.
Swap several of the 3.3/5v for 12v pins, require a lower gauge/higher amp, and/or use additional/smaller pins.

It will always be far superior to let the PSU makers build the PSUs. We already have quality and efficiency.
It's absolutely stupid to force two separate companies to build half a PSU each, and that somehow it will be better?
MOBOs are already very complex and space constrained (especially MATX/ITX), this would make it so much worse.
I have so many more things to say about how absolutely, monumentally stupid 12VO is...but I'll stop here in this thread.

I'm fine with a new smaller and smarter GPU connection for top of the line cards. Especially if it's a free open standard.
Even if AMD/Nvidia had a different connector and required an adapter, it'd be tolerable. Especially with modular PSUs.
I do hope Nvidia will do the smart thing and make it a free standard, just to help themselves by making it easy to adopt.

 

[Markfw]

What happenes when you pay $1500 for a 3090, and it has one 12 pin connector, and NO PSU's have one ? Sorry, I may have missed the answer somewhere.

 

[Martimus]

What happenes when you pay $1500 for a 3090, and it has one 12 pin connector, and NO PSU's have one ? Sorry, I may have missed the answer somewhere.

There is an adapter that makes two 8-pin connectors into a single 12-pin connector.

 

[Markfw]

There is an adapter that makes two 8-pin connectors into a single 12-pin connector.

I comes with the card ?

 

[Martimus]

I comes with the card ?

Yes. That is what they said in the presentation today.

 

[Markfw]

Yes. That is what they said in the presentation today.

Since its 12 pins, it should use 2 6 pin GPU connectors.... But whatever....

 

[Bouowmx]

The 12-pin is just slightly bigger than a 8-pin. The 12-pin mounted vertically should just fit under one slot (less than 20 mm). For the liquid cooling enthusiasts.

 

[Hitman928]

Since its 12 pins, it should use 2 6 pin GPU connectors.... But whatever....

Power wise it's more than 2x6 pins can handle so it has to be 2x8 pin.