Pure 12V PSU Standard, Named ATX12VO, Debuts Later This Year

[TheLostSwede]

We were talking footprint. Not warranty. If I buy the same components at twice the rate, I have twice the ecological footprint... makes sense? Or is my train of thought wrong over here?

Again, this is NOT going to happen over night. So in your case, you can most likely match the upgrade pattern you use to fit in a new PSU when it suits you.

 

[Valantar]

Fair enough for a SBC but what about an ATX sized board?
Let's say 8 USB ports on the IO and 4 internal each 900mA. Plus 7 PCIe slots with each 3A and two M.2s with 2.5A each. That's over 50 Watts for the 5V and 80 Watts for 3.3V supply.
Putting additional VRMs of that dimensions on an already feature packed PCB is tricky and costly.

Given that hardly any AICs these days use 3.3V there's no reason to spec out power delivery to deliver the maximum amount of power for each port like that. m.2 does need 3.3V, so that would need at least enough to populate all slots, but most PCIe cards will work just fine without the 3.3V rail there. The same goes for 5V, really - no need to spec for full utilization of every single port, as that's never going to happen in the real world.

A move to a pure 12V standard will also encourage AIC vendors to further embrace 12V input and on-board voltage conversion, which will again accelerate the long-awaited and too slow in coming demise of 3.3V.

The problem with USB these days is that with USB-C, at least some ports can (should?) be as much as 100W, but that's at 20V/5A. However, it's unlikely we'll see this on any desktop PC, but maybe these new power supplies will make 12V/5A more likely to appear than today.
However, long gone are the days of 500mA or 900mA USB connectors, as most boards deliver 1-2.4A per USB ports these days, simply because people expect to be able to charge their mobile devices from the ports on their PC.

That's the USB-PD charging standard, not the host device standard. Most (all?) host devices max out at 15W output, which is a relatively reasonable number (though it should be rather trivial for desktop ports to add a 12V3A mode to their current 5V3A mode). The USB-C/3.1G2 front panel connector is specced for 3A of power, so moving beyond that would require yet another standard before this has even seen any real adoption. That's a bad idea. But beyond that, we really don't need the ability to charge our laptops from our desktops. Sufficient power to run an external 3.5" HDD would be nice, though - but that can be done with 12V1-1.5A.

@Valantar That Murata part is insanely expensive though, so that's not likely to find its way into regular PCs as yet. Also, high Amps on the minor rails is a way that the PSU makers have managed to "uprate" their PSUs for years, since it's the combine Wattage that's put on the box at the end of the day...

Very expensive, yes, but it was also the literal first result that showed up on Mouser. I'm sure there are cheaper options out there for someone actually looking.

 

[iO]

Nvm. At the latest when DDR5 arrives there should be enough space on the PCB for any minor voltage rails as DDR5 has it's VRM moved on the module itself...

 

[DeeJay1001]

Pretty much all OEMs already do this but they use proprietary connectors. I have a dell 7020 from 2014 sitting here and it has proprietary keyed 8 pin connector in place of a 24 pin.
Setting a standard for OEMs to adopt is a good thing.

 

[londiste]

If someone didn't want to click to the source article and look for the link there, here is the Design Guide:

https://www.intel.com/content/dam/www/public/us/en/documents/guides/single-rail-power-supply-platform-atx12vo-design-guide.pdf

 

[Ferrum Master]

. Sufficient power to run an external 3.5" HDD would be nice, though - but that can be done with 12V1-1.5A.

Nope... some even modern drives during spin-up eat over 3A... not the greatest idea.

 

[Axaion]

relying on motherboard makers to make great VRMs to deal with this is going to be a disaster, at anything but high end

 

[wiyosaya]

Still only 2 12V line just like the current specs?? Then why bother to change it.

As I see it, this is more of an issue than meets the eye. Back in the day of slot-1 procs, I had a dual proc system. The power connectors from the supply to the MB literally (which were before PS 2 days) burned over time so that I did my own mod and soldered them. The connectors were these http://www.playtool.com/pages/psuconnectors/connectors.html#oldpc , and could only handle 5A per connector.

So with the current ATX power supply standard, the max current per connector is 6A and the power rating per connector is 72W per http://www.playtool.com/pages/psuconnectors/connectors.html#atxmain24
This means that the MB better consume 144W or less - at least without other auxiliary connectors. The additional CPU power connectors on today's MBs are there because of the power requirements of CPUs.

 

[laszlo]

...CPU VRMs deliver hundreds of amps. CPUs run at anywhere from 1V to 1.4V (-ish) while consuming ~50-300W.

sorry but a whole system which use 600w is using 50 amps so won't deliver hundreds

on topic atx psu is using much more components to obtain the other voltages than 12; is much easier and cheaper to obtain lower voltages from dc ;f.ex. from 12 v 10A - you can get 5 v 7A; 3.3 2A with minimal components and no capacitors; i left 1A out for heat and power loss...

in my opinion this approach will be the norm and i have nothing against it as it will lower overall costs and components usage

 

[iO]

If someone didn't want to click to the source article and look for the link there, here is the Design Guide:

https://www.intel.com/content/dam/www/public/us/en/documents/guides/single-rail-power-supply-platform-atx12vo-design-guide.pdf

Meh, so much for a 12V only PSU...

Edited because of brain fart

 

[Ferrum Master]

sorry but a whole system which use 600w is using 50 amps so won't deliver hundreds

on topic atx psu is using much more components to obtain the other voltages than 12; is much easier and cheaper to obtain lower voltages from dc ;f.ex. from 12 v 10A - you can get 5 v 7A; 3.3 2A with minimal components and no capacitors; i left 1A out for heat and power loss...

in my opinion this approach will be the norm and i have nothing against it as it will lower overall costs and components usage

CPU VRM does deliver hundreds. Because the voltage is 1.0-1.4V. At 100A it will be 100-140W...

Meh, so much for a 12V only PSU...
View attachment 143017

Looks like a derp...

The PSU doesn't have any sort of SATA power at all

 

[iO]

CPU VRM does deliver hundreds. Because the voltage is 1.0-1.4V. At 100A it will be 100-140W...



Looks like a derp...

The PSU doesn't have any sort of SATA power at all

View attachment 143019

View attachment 143020

Whoopsie, I'm just too dumb to read: "This connector is used on the Storage Device cable coming from the motherboard"
Those boards are certainly going to be interesting.

 

[agatong55]

Wouldn't something like this make it harder to overclock or would it make it easier?

 

[Arct1c0n]

Fark this shit, I donk't wanna PAY MORE for a more complicated and more likely to fail mobo instead of the PSU doing the conversion. Screw you Intel, this is another BTX standard crap idea that needs to die

 

[Valantar]

sorry but a whole system which use 600w is using 50 amps so won't deliver hundreds

...that's at 12V. Amperage is relative to voltage, after all. CPU VRMs deliver anything from ~.5V (idle) to ~1.5V (load). That translates to >100A for a high end CPU at stock clocks between the VRM and CPU, and several hundred for something really power hungry like a clocked-to-the-nines 9900KS. It's true that the amperage between the PSU and VRM is much lower, but again, that's because the voltage is higher. This is the whole point of moving the VRM as close as possible to the power sink, after all - the further away, the more overbuilt cabling/traces you will need.

Wouldn't something like this make it harder to overclock or would it make it easier?

It wouldn't change anything in that regard whatsoever. CPUs already run off the 12V (only) EPS connector. It will likely make better 12V PSUs cheaper though, so as such it might make overclocking ever so slightly "easier" to achieve - but that's such a small difference it likely won't be noticeable.

Nope... some even modern drives during spin-up eat over 3A... not the greatest idea.

Couldn't that be solved by spinning up the drive more slowly? I guess that would require some firmware trickery, but I can't imagine that it wouldn't be possible. But as I stated above, adding a 12V3A mode to current USB-C ports and front panel USB-C ports should be relatively trivial, and that would be sufficient for what you're describing. It's not like spin-up is a sustained state, and delivering the max spec for a few seconds should be fine for any properly made connector.

relying on motherboard makers to make great VRMs to deal with this is going to be a disaster, at anything but high end

That's what standards are for. The current ATX power supply standard specifies pretty much every single relevant specification for a PSU to be compliant. There's no reason whatsoever to expect a standard like this to be less strict. And enforcing minimum MTBFs or similar specs for the relevant components would be a simple addition to the spec.

Beyond that, why, exactly? Other than for overclocking, can you show me an example of a motherboard with a CPU VRM bad enough to actually be problematic?

 

[londiste]

If motherboard manufacturers fail at providing necessary outputs or the quality sucks, there will be 3rd party adapters/splitters that output the stuff that you require. I am pretty sure we have a couple guys in this thread who can easily create one and it would be pretty cheap as well.

 

[Ferrum Master]

Couldn't that be solved by spinning up the drive more slowly?

It is actually, there are features, kinda... but yet still some manage to make some drives... *cough Seagate cough* There were some forum reviewers a while a ago... imagine it as a problem power firing up a NAS with many discs... it can produce a horrible spike thus the supply could crap out.

I don't understand the worry about motherboard maker ability to make it work... laptops are doing it for years...

 

[Dredi]

Meh, so much for a 12V only PSU...
View attachment 143017

You didn’t really read it, did you?

 

[ppn]

I don't want any sata power coming from motherboard. It destroys the clean simplistic layout.

 

[Dredi]

I don't want any sata power coming from motherboard. It destroys the clean simplistic layout.

Then don’t use sata devices? They already get a cable from the motherboard, big deal. U2 is a better alternative anyway.

 

[candle_86]

I'm surprised this hasn't happened sooner tbh. With storage moving towards M.2, even things like SATA and Molex connectors aren't going to be widely used, apart from when people are using their PC's and keep done older parts.
It's also quite efficient to convert 12V DC to lower DC Voltages, especially compared to going from AC to DC. I've done some power conversion boards for a client and they were quite cheap to make and used very few components, yet can support quite high loads without producing much heat.

When m.2 can replace my 4x2tb drives affordably maybe but otherwise nope.

 

[Valantar]

When m.2 can replace my 4x2tb drives affordably maybe but otherwise nope.

It would be quite trivial to design HDDs to run off 12V only. And if not, this new standard includes provisions for standard SATA power off the motherboard.

 

[TheLostSwede]

When m.2 can replace my 4x2tb drives affordably maybe but otherwise nope.

I put all my mechanical drives in a NAS. I only have SSDs in my PC with two out of three being M.2. Maybe I'm ahead of the curve, but I expect M.2 to become the main drive standard within a couple of years for anything but mechanical drives.

It would be quite trivial to design HDDs to run off 12V only. And if not, this new standard includes provisions for standard SATA power off the motherboard.

Or to make a small adapter, much like what was available for older IDE drives to make them work with SATA connectors.
Or as I said elsewhere in the comments here, SATA backplanes will become standard in cases, so you simply connect the power and the SATA cables to the backplane and then slide in the drives, with no cables connecting to the drives.

 

[Valantar]

I put all my mechanical drives in a NAS. I only have SSDs in my PC with two out of three being M.2. Maybe I'm ahead of the curve, but I expect M.2 to become the main drive standard within a couple of years for anything but mechanical drives.


Or to make a small adapter, much like what was available for older IDE drives to make them work with SATA connectors.
Or as I said elsewhere in the comments here, SATA backplanes will become standard in cases, so you simply connect the power and the SATA cables to the backplane and then slide in the drives, with no cables connecting to the drives.

That is exactly the directions I see going forward too. m.2 taking over end-user PC storage entirely (no cabling, yay!), and mechanical drives/mass storage being outsourced to some sort of networked storage, whether that's an OTS NAS or some DIY solution. nGbE becoming cheaper should help this along quite a bit, as even 2.5GbE doesn't bottleneck a 3.5" HDD and can provide decent responsiveness off something with SSD caching. The only reason any of the PCs in my house have mechanical drives in them are a) that they're my HTPC/NAS combo that serves all our media, or b) that they're used for 4k video editing and need a lot of storage that would be significantly bottlenecked by the current GbE network. In the coming year I see both of those changing, as I'll be building a new bespoke (and tiny) HTPC and relegating the old one to pure NAS duties, and hopefully also moving to some form of nGbE as long as someone makes a damn switch that doesn't cost a fortune. After quite a few HDD failures over the years I really can't wait for all of those to be centralized into one easily managed and redundant storage setup where I can forget about their noise and reliability until one of them needs swapping.

/rant

 

[LabRat 891]

Mixed feelings.
The idea of mass-production minimalized-cost OEM boards being responsible for more power conversion on board, especially with the era of USB powering so much mobile consumer gadgetry, is worrisome.
That said, this could improve reliability(simplicity in single rail design), efficiency, and BOM cost of OEM PSUs, as well as amount of copper needed in a given system. All of which are important in some regard to mass production and marketing. Raising efficiency, even at other costs, looks good to regulatory commissions and investors.

From an enthusiasts' standpoint (since we know this will eventually trickle into 'our world'): I am intrigued at the idea of building very simple, high-current power supplies and running off common battery power sources with less need for complicated SMPS building. (Linear regulated supercap - lead acid/lifepo backup power?) IIRC Google runs mostly custom hardware with Common Rail DC power for simplicity in power backup solutions. (Though this may have changed since I read about this years ago).

I wouldn't mind seeing 2.5" and 3.5" drives be pushed to an all 12V design. I just don't see storage being pushed fully to NGFF anytime soon. Mechanical HDDs are still highly competitive for high capacity cost efficiency. As mentioned, backplane or power adapter boards can be produced inexpensively en masse. Ex. Look at automotive USB chargers. I've actually used a combo of a 12V wallwart and a car charger to power a SATA drive while connecting via eSATA - SATA (I don't recommend this BTW. The ripple and transient AC may kill your drive)