New Asus GPU 600W Motherboard power connector

[05032-Mendicant-Bias]

Summary

Asus is experimenting with a 600W power connector (Wccftech) mounted on the motherboard, replacement for both the standard 2x4 150W connector, and the 600W 12VHPWR connector that became notorious for melting if not inserted correctly (Gamer Nexus in depth reporting).

 

 

 

Quotes

Quote

The connector is rated to deliver up to 600W of power and it will use a special connector that is available on a select range of motherboards.

 

My thoughts

The regular PCI-E slot can provide 75W of power. From an engineering viewpoint, extending the PCB with special 600W traces makes sense. The GPU is supported more firmly, without a bulky connector at awkward angles trying to pry it off it's PCI-E slot.

 

Everything else is a mess:

• Nvidia is trying to force a new standard with the 12VHPWR connector, useful ONLY for their GPUs.
• Nvidia ships adapters from standard 8pin connector to whatever proprietary connector they make.
• The 12VHPWR connector is already unfit for duty, as some 4090 variant use two of them! 
• Asus is trying to force a new standard, with their non-standard PCI-E extension.
• 600W is a lot of power for a consumer GPU to draw.

Servers do the right thing, and mostly use standard mezzanine or fanless PCI-E GPUs with standard 8 pin connectors.

 

Manufacturers, including OEMs should just stop making non standard components and non standard form factors.

You want to manufacture a GPU that draws 675W? Put four 8 pin connectors on the side of the card!

 

You want a better standard? Negotiate with other manufacturers for a proper backward compatible extension to the standard.

 

Sources

https://wccftech.com/asus-shows-off-geforce-rtx-40-graphics-card-without-power-connectors-uses-proprietary-slot/

[Poinkachu]

Inbefore patented by ASUS somehow and nobody else wants to implement it due to the patent. ?

 

Then everybody else try to make their own version of it, which only compatible with their own mobo.

Then we're in hell. (as users)

[leadeater]

6 minutes ago, 05032-Mendicant-Bias said:

Servers do the right thing, and mostly use standard mezzanine or fanless PCI-E GPUs with standard 8 pin connectors.

Sadly this isn't the situation. When it comes to Nvidia GPUs the SXM slot is proprietary to Nvidia and they even go so far as to require server vendors to use board modules design and manufactured by them or to be manufactured to their design.

 

For PCIe power servers use vendor specific power delivery so the only standard part is the single device end with the PCIe power connector on it. You cannot even guarantee that the same power cable will work between different generations of servers from the same vendor.

 

Servers are a hotbed of proprietary designs and implementations and they only conform to industry standards and specifications where actually necessary.

[Spotty]

1 hour ago, 05032-Mendicant-Bias said:

replacement for both the standard 2x4 150W connector, and the 600W 12VHPWR connector that became notorious for melting if not inserted correctly
[snip]

Everything else is a mess:

• Nvidia is trying to force a new standard with the 12VHPWR connector, useful ONLY for their GPUs.
• Nvidia ships adapters from standard 8pin connector to whatever proprietary connector they make.
• The 12VHPWR connector is already unfit for duty, as some 4090 variant use two of them! 

Except this is not a replacement for either of those connectors. The motherboard needs a way to get the power to provide the graphics card. The back of the motherboard features a 12VHPWR connector as well as 3x 8pin PCIe connectors. All this does is move the connector to the motherboard instead of the graphics card.

 

Quote

The Z790 TUF Gaming connector-less design also hosts three 8-Pin connectors and a single 12VHPWR connector. These connectors feed power directly to the graphics card through the proprietary slot.

[05032-Mendicant-Bias]

1 hour ago, leadeater said:

Servers are a hotbed of proprietary designs and implementations and they only conform to industry standards and specifications where actually necessary.

You are right, custom server boards are more common than I'd like to admit.

30 minutes ago, Spotty said:

Except this is not a replacement for either of those connectors. The motherboard needs a way to get the power to provide the graphics card.

The Asus 600W is a replacement for 8pin 150W connectors and 600W 12VHPWR connectors. What matters is that the GPU no longer draw power from those connectors, but draw power directly from the non standard PCI-E extension on the motherboard.

As you say, The MoBo does have to draw that power from somewhere. A long time ago MoBo started using an additional 4pin 12V connector and now we have a dedicated 8pin 12V connector for the CPU. In the proposed Asus design, there are additional 8 pin connectors on the motherboard that feed the 600W PCI-E non standard extension for the GPU.

Which also competes with intel proposed 12V Only standard to replace the ATX connector. There is a single voltage to provide, and still, manufacturers are competing for their own flavour of the connector... I argue manufacturer should come together to ratify an extension of the ATX standard for 12V. Not just Intel, Dell, etc... Everyone has to be onboard so that we consumers are not stuck with power supplies that have a special cable for a single generation of Nvidia GPUs...

 

12 minutes ago, porina said:

Now I'm also wondering if 12VHPRW connectors finally do away with ATX 24 pin mobo connectors? 12VO is long overdue.

Logically, having a 12VO MoBo connector sized for low end low power system, with a number of identical 8pin 150W connectors that scale based on what CPU and GPU the system is supposed to support look simpler and smarter than the current 24pin ATX connector, the special 8pin CPU connector, and the mess of GPU power connectors. 12VO also specifies the MoBo to SATA drive connectors.

 

[porina]

I'm more interested in that they seem to have done away with the annoying release tab commonly found on x16 slots. Such a pain to reach those when it is obscured by a GPU, the CPU cooler, and maybe ram too.

 

It would give a clean look and do away with the extra cables to the card.

 

On the additional connector, this would bring with it a load of compatibility questions. Can you use a GPU with it, without a matching mobo? For example, by having regular power connectors on it in addition? If so, is the card generally mechanically compatible with other mobos?

 

This could make some sense for prebuilts, or Asus on Asus. Say you had one of these, and upgrade GPU in future, would it limit resale potential of the old GPU?

 

Now I'm also wondering if 12VHPRW connectors finally do away with ATX 24 pin mobo connectors? 12VO is long overdue.

[Levent]

I am all for reduced cables, I like it.

 

Now, PLEASE, DO THE DAMN FRONT PANEL IO.

[Spotty]

7 minutes ago, 05032-Mendicant-Bias said:

The Asus 600W is a replacement for 8pin 150W connectors and 600W 12VHPWR connectors.

How does this replace those standards? The motherboard uses the 8pin PCIe and 12VHPWR connectors to provide power to the graphics card. The connectors have been moved to the motherboard instead of on the graphics card.

 

5 minutes ago, 05032-Mendicant-Bias said:

In the proposed Asus design, there are additional 8 pin connectors on the motherboard that feed the 600W PCI-E non standard extension for the GPU.

There's a 12VHPWR connector on the motherboard which is to provide power to the card.

 

 

 

6 minutes ago, 05032-Mendicant-Bias said:

Which also competes with intel proposed 12V Only standard to replace the ATX connector. There is a single voltage to provide, and still, manufacturers are competing for their own flavour of the connector...

This doesn't compete with 12VO standard...? This motherboard works with a standard ATX power supply. There's still 5V, 3.3V, 5VSB.

 

 

12 minutes ago, 05032-Mendicant-Bias said:

There is a single voltage to provide, and still, manufacturers are competing for their own flavour of the connector... I argue manufacturer should come together to ratify an extension of the ATX standard for 12V. Not just Intel, Dell, etc... Everyone has to be onboard so that we consumers are not stuck with power supplies that have a special cable for a single generation of Nvidia GPUs...

12VHPWR is an industry standard connector. It's designed by PCI-SIG as part of PCIe 5 specifications and was also adopted in to Intel ATX design specifications.

It is not Nvidia exclusive. The only reason AMD is not using it is because they chose not to.

[Poinkachu]

18 minutes ago, porina said:

I'm more interested in that they seem to have done away with the annoying release tab commonly found on x16 slots. Such a pain to reach those when it is obscured by a GPU, the CPU cooler, and maybe ram too.

Well, if there is 2 things I want the most from ASUS motherboard :

1. M.2 Clip (No more tiny little bastard screw)

2. Their "detached" PCIE Release system on some of their Z700 boards (dunno what its actually called)

 

Probably pantented

[05032-Mendicant-Bias]

21 minutes ago, Spotty said:

It is not Nvidia exclusive. The only reason AMD is not using it is because they chose not to.

I think we are coming down to semantics. It's also the reason we have de-facto standards, that are not ratified but everyone uses.
I consider it a standard if the majority of the industry uses it.
I consider it proprietary if a minority of the industry uses it.

 

The fact neither AMD nor Intel are using the 12VHPWR connector, makes it proprietary in my book.
I love the Intel 12VO standard, but nobody uses it still, so it's proprietary in my book.

 

One of my favourite standards is the industrial automation connector. it specify the screw (M8 M12 M23, etc...), the number of pins (4, 5, 8, ---), and if it's female or male. Eveyone manufactures them and it makes it so easy to plug everything into everything. In industry you find it everywhere, from power cable, to photodetectors, to actuators, to high speed cameras.

  

21 minutes ago, Spotty said:

This doesn't compete with 12VO standard...?

Asus is suddenly plugging more power connectors into a motherboard, in a different way from what everyone else is doing,  at a time when manufacturers are trying to agree on a new way to power motherboards. The evergreen ATX, Intel 12VO, Dell 12 Volt Only (different from Intel 12 Volt Only), Asus 600W GPU extension, and so on, and so on... (evergreen xkcd comic)

[Poinkachu]

Honestly, at current rate of increase in power consumption.
Soon enough the standard will be like :

 

[porina]

1 hour ago, 05032-Mendicant-Bias said:

I think we are coming down to semantics. It's also the reason we have de-facto standards, that are not ratified but everyone uses.
I consider it a standard if the majority of the industry uses it.
I consider it proprietary if a minority of the industry uses it.

The main difference I see between a formal standard and proprietary one is that the former is controlled by a standards body, and the latter is controlled by company. Formal standards can be initiated by companies, but they give up sole control over it in the process of standardisation. They may still provide input to future versions. Popularity does not play a part in this.

 

By my interpretation, 12VHWPR would not fall under proprietary as it is published by PCI-SIG. I'm not sure about 12VO. ATX and derivatives were (are?) an Intel standard, but I don't know who runs it now. Is there an industry group? So I'm wondering if ATX is a de facto standard but is proprietary? Another example of this might be CUDA.

[DuckDodgers]

The old AGP standard had a "Pro" spec with extended power supply though the slot. Hardware vendors should all adopt similar extension to the PCIe and avoid proprietary lock-ins. 

[Ydfhlx]

4 hours ago, 05032-Mendicant-Bias said:

• Nvidia is trying to force a new standard with the 12VHPWR connector, useful ONLY for their GPUs.
• Nvidia ships adapters from standard 8pin connector to whatever proprietary connector they make.
• The 12VHPWR connector is already unfit for duty, as some 4090 variant use two of them! 

"ONLY" nvidia GPUs means like half of models and 80%+ market share. I'm not counting Intel, as right now they don't make a GPU that would need so much power.

 

And well yes, nvidia ships adapters, but I assume Asus also has traditional connectors on the GPU.

 

Also, if that connector is "already unfit", then so is this connector, as they deliver exactly same power.

[SorryBella]

4 hours ago, Poinkachu said:

Inbefore patented by ASUS

I mean, the precedent is there.

[HenrySalayne]

5 hours ago, 05032-Mendicant-Bias said:

Servers do the right thing, and mostly use standard mezzanine or fanless PCI-E GPUs with standard 8 pin connectors.

Standard? If I'm not mistaken these are often EPS 12V connectors because that's what you will find in a server. PCIe 8P is rather exotic. EPS 12V supports up to 336 W.

3 hours ago, 05032-Mendicant-Bias said:

The fact neither AMD nor Intel are using the 12VHPWR connector, makes it proprietary in my book.

That's very narrow-minded. Intel, Nvidia and AMD launched their current generation pretty much around the same time and only shortly after the 12VHPWR connector became a thing. It's not surprising that adoption was not uniform and mainly the manufacturer with the alleged 600W GPU opted for it immediately. It's very ikely we will see 12VHPWR on all next gen GPUs.

[jagdtigger]

4 hours ago, 05032-Mendicant-Bias said:

12VO

That is basically a cheat to get better efficiency ratings on PSUs by removing the less efficient parts fron it and move it to the MOBO so it doesnt count against their ratings......

[Doobeedoo]

Really routing everything on mobo vs cables would be good. Definitely cleaner and really more direct rigid connection. Make it a 1KW connector and done heh.

[cooky560]

I think this connector is a great idea, provided that ASUS provides some kind of adaptor for that port to work with non asus gpus, we'd also need a shift in where gpus put their power connectors, however if both of things come to pass, it's a great method to improve cable management. 

 

I'm sure I saw some gpus once with the power connectors on the motherboard side for good cable management, but I can't remember where or how long ago.

[Poinkachu]

1 hour ago, cooky560 said:

I think this connector is a great idea, provided that ASUS provides some kind of adaptor for that port to work with non asus gpus, we'd also need a shift in where gpus put their power connectors, however if both of things come to pass, it's a great method to improve cable management. 

 

I'm sure I saw some gpus once with the power connectors on the motherboard side for good cable management, but I can't remember where or how long ago.

Well, by my own logic : if non ASUS GPU doesn't implement the same thing, meaning the adapter will be cable based.

And since GPU sizes are different, the cable on the adapter will be long enough to cater to all (it will be either dangling under the GPU fans, or laying on the backplate).

And knowing ASUS... most likely they are going to price it premium style.

 

Quote

also need a shift in where gpus put their power connectors, however if both of things come to pass, it's a great method to improve cable management. 

This I agree.

IIRC, EVGA planned make their 4000 series connector on the tip of the GPU, facing the front of the case. But yeah... too bad they are not in the game now.

 

But yeah, I'm also curious about what kinda thermal the motherboard gonna exudes.

[Age of 9394]

Here is the YouTube video, if you know Chinese, or you can set the subtitle to English

[SansVarnic]

I have no issue with creating a new power delivery standard, the same has been around for some time. Granted it will be bit of a headache for early adopters until a new standard is finalized of course and hopefully it will be an actual improvement. 

[Quackers101]

sort of cool and why this wasn't a thing? and if its just a directly behind the motherboard to the card.
but can see loads of issues, can you trust every mobo to handle this correctly, universal design, where it gets the power from, etc.

[jagdtigger]

55 minutes ago, Quackers101 said:

why this wasn't a thing?

Because making traces thick enough for that sort of PD on a ~dozen or so layer board aint cheap.....

[Quackers101]

2 hours ago, jagdtigger said:

Because making traces thick enough for that sort of PD on a ~dozen or so layer board aint cheap.....

yeah and sounds complicated if doing 600w and if its not a "straight through" design with a complicated path. which is the cost concerns as with the other cost concerns for motherboards like power delivery for smaller devices, DDR 4-5, newer PCIe gen, nvme storage etc.