New ATX12VO Standard for PC Power Supplies, Good or Bad?

[MrSMT]

Hello everyone,

greetings from Germany! I'm new to the forum and I apologise if this isn't fitting here.

Intel announced a new ATX power supply standard called the ATX12VO and it’s supposed to make computers more power efficient by eliminating the 3.3v and 5v power rails form the power supply and just keeping the 12v, hence ATX-12Volt-Only. This new standard will effect only the power supply and motherboard, because PC components will always require 3.3v and 5v, the DC to DC 12v to 3.3v and 5v will take place on the motherboard instead of the power supply.

The most detailed news article about the standard I found was from GamersNexus and can be found here:
https://www.gamersnexus.net/guides/3568-intel-atx-12vo-spec-explained-what-manufacturers-think
The Design Guide from intel is also linked here:
https://www.intel.com/content/dam/www/public/us/en/documents/guides/single-rail-power-supply-platform-atx12vo-design-guide.pdf

Intel is calming that by moving from the mains to 3.3v and 5v conversion to a 12v to 3.3v and 5v DC to DC conversion the efficiency will be improved. This makes sense.
Now, I don’t have a lot of knowledge in electrical engineering, I’m just a med student. But from what I know, a lot of power supplies  already do this. There is even a power supply by Seasonic (Seasonic CONNECT) that does this DC to DC conversion on a separate board outside of the power supply housing. How is this different from doing the DC to DC conversion on the motherboard?

Some news outlets even claim that the improved efficiency is because of the extra wires being eliminated and thus minimising the loss over the cables. But this is not true in my eyes, because A) the losses over the cables are very small to begin with and B) the same cables will still be needed for the other component, they will just be connecting to the motherboard instead of the power supply.

I see this new standard more harmful than anything else. It will only make old hardware obsolete over time. Moving the DC to DC conversion to the motherboard will mean more heat around the CPU and other components, instead of it being contained in the power supply’s housing with its own fan and (heat) environment. Intel could have updated the existing ATX12V standard to discourage mains to 3.3v and 5v conversion and only allow the more efficient DC to DC conversion in the power supply. Why are all news sources praising this instead of criticising it? Am I completely wrong here?

[magic]

Buck converters are quite efficient and to make 1A at 5V they only need less than 0.5A at 12V. So the amount of current through PSU cables and connectors decreases and so does power loss. It's the same reason why power transmission is done at hundreds of kV over long distances and then stepped down to 230V.

There is now the added loss from the buck converter, but if the OEMs are doing it that way then it must be working out for them. They also save on the cost of cables and connectors.

So far they say it's an attempt at standardizing things that OEMs have already been doing. If this ever hits normal motherboards there will be some incompatibility, but frankly, an old PSU will run the new boards just fine with just an adapter and a new PSU will require something like a pico PSU to run old boards. The load on 5V and 3.3V rails in modern systems is minimal - mostly just USB and disks.

[ConKbot]

Yeah I'm not particularly buying it. Smells more of regulatory boondoggle than anything. You can bump up the PSU nameplate efficiency by relocating the buck converter board outside the PSU case. For the wire loss, If a PSU puts out 20A on a 5V rail, split between 4 ~15" 18awg wires, thats 0.002 ohms, or 0.8W of dissipation.  Thats 0.8% of the 100W 5v load, or 0.1% of the total load assuming a maxed out 600W power supply.  Meanwhile assuming a nice 92% efficient converter is used (higher can be had, bit this is a consumer product) 8.7w of dissipation (1.4% of the 600W max) just got removed from your efficiency figures.

One less cynical benefit is that the bucks can be sized appropriately, I.e. A 20A high power buck isn't sitting there in pulse skipping or DCM trying to maintain efficiency because the 5V rail is only loaded with a USB keyboard and mouse and thats about it. (M.2 SSD, no optical drive, etc.)

If Steve LTT compared a modern DCDC converter PSU to a 12VO system, I bet system-at-idle results would have been less impressive.

If PSU manufacturers were after chasing efficiency to the max, they would use 16AWG wire instead of 18 and reduce wire losses significantly. It fits in Molex mini-fits that get used for all the high current loads, so it absolutely is just their choice.

My money is on  PC OEMs need more efficient power supplies to help with efficiency phase (whatever) standards, didn't like the quote they got back from PSU OEMs, so they leaned on motherboard makers and Intel to help them out. Cheaper PSU, increase on motherboard BOM, but they can penny pinch the motherboard DCDC converters, as they aren't under the regulatory supervision. So motherboard/PSU combo are cheaper because of that. Plus if the PSU manufacturers have efficiency margin left over after the reallocation, they can budget cut that some as well, further decreasing PSU cost.

Spin it with some factually true, but practically irrelevant facts, hey presto, there's your press release.

Edit:just to clarify, not really for or against the idea of if, just not seeing it as some major boon to efficiency that the press releases would have you believe. When the solution was already had with separate DD/DC boards in the power supply.

[wraper]

Yeah I'm not particularly buying it. Smells more of regulatory boondoggle than anything. You can bump up the PSU nameplate efficiency by relocating the buck converter board outside the PSU case.

Motherboards need DC/DC converters anyway. There are not many things left on motherboards which natively need 5V or 3.3V. It's better to place adequate buck converters on motherboard rather in PSU for maximum possible current draw. Why make more expensive and less efficient PSU with many voltages when when the most optimal way if to convert voltages on motherboard according to actual needs.

[tszaboo]

The thing is, the most current and power in a PC is not used on 12V, 5V or 3.3V. It is about 1.2V, used by the CPU and the GPU. In fact, nothing really uses 5V and 3.3V, except maybe the motors in a HDD or the flash memory of the BIOS, and a few odds and ends.  It is just something that stayed with us, because in the 80s, there were a bunch of logic chips working with these voltages.
LTT tested, and the idle power usage of the PC went down drastically. The concept of ATX12VO is very good. To bad the actual implementation is pretty retarded, with 4 different connector on the MB, 2 for graphics. And they kept the old sata connector, but now it has only 12V, so there is now a 15 pin connector for 2 wires. Or you need to plug in the drive to the MB for powering it. Like, how much effort would be to build in a 12V to 5V buck on a small PCB, that is plugged into the drive itself?

[NiHaoMike]

One trend I have started to notice is that PSU manufacturers are starting to cut down on 5V and 3.3V capacity. A mere 15A is not uncommon, which is cutting things way too close for systems that have a lot of high current USB devices connected such as a bunch of smartphones for app development.

[Red Squirrel]

I would love this standard to come to the consumer market, but from my understanding it's only going to be for OEMs.  So it's just as good as being proprietary, since it will be hard to get the parts (motherboards and PSUs) that work with it.   

If you could home build machines with this standard it would be great for home/SOHO server rooms, you can just run everything at 12v and skip the PSUs for each machine and just have a bank of 13.5v PSUs that float a bank of batteries, like telcos do with 48v.  Depends if the standard has enough tolerance and would allow for 13.5v though.   Doing this would skip a conversion stage while providing full online power protection.

[NiHaoMike]

If you could home build machines with this standard it would be great for home/SOHO server rooms, you can just run everything at 12v and skip the PSUs for each machine and just have a bank of 13.5v PSUs that float a bank of batteries, like telcos do with 48v.  Depends if the standard has enough tolerance and would allow for 13.5v though.   Doing this would skip a conversion stage while providing full online power protection.

3S lithium ion would be a better match.

[filssavi]

The difference between ATX and ATX12VO is just the placement of the DC/DC converter, that’s it. Virtually all modern 80+ bronze and better PSUs already do DC/DC for 5V and 3.3V as it is much more efficient than using mag/amps.

Now this means that 12VO main contribution is is to cut down conduction losses on cables and connectors. It is true that you are moving heat to the motherboard, however those rails pull so little power that the dissipation of a modern high frequency buck converter will be less than a W, that is nothing compared to the core voltage regulator alone (that will dissipate from 5 to 30/40 W depending on cpu model, load, eventual overclock etc.)

As for the e-waste perspective, again I don’t see how it makes a difference, on OEM systems (for which this standard is targeted) you will have a PSU included regardless (Have you ever seen a dell workstation with no PSU included?). While in consumer land first I don’t see this ever catching on unless forced by law, as I don’t think mobo manufacturers too kin on having to do customer support for the transition (with hordes of angry people having bought a 12VO psu and a normal mobo or vice versa)

[magic]

Here's the California regulations they mentioned.

https://govt.westlaw.com/calregs/Document/IEEDE2D64EF7B4F168C0E85379828A8C2?transitionType=Default&contextData=%28sc.Default%29

It's not just PSU efficiency (table V-9) but also annual consumption (not sure under what load conditions - table V-7) and even extra allowances for various optional components (table V-8).
For instance, RAM expansion is supposed to use 4kWhr + 0.15 times capacity in gigabytes per year. They thought of everything

[Refrigerator]

Here's the California regulations they mentioned.

https://govt.westlaw.com/calregs/Document/IEEDE2D64EF7B4F168C0E85379828A8C2?transitionType=Default&contextData=%28sc.Default%29

It's not just PSU efficiency (table V-9) but also annual consumption (not sure under what load conditions - table V-7) and even extra allowances for various optional components (table V-8).
For instance, RAM expansion is supposed to use 4kWhr + 0.15 times capacity in gigabytes per year. They thought of everything

Does it say anything about RGB? 

[LeonR]

I wonder why haven't the standard upped the ante to something like 24-48V.

I just hope we don't have to rely on motherboards for powering up HDDs and the alikes.

[PKTKS]

I wonder why haven't the standard upped the ante to something like 24-48V.

I just hope we don't have to rely on motherboards for powering up HDDs and the alikes.

Simple facts of power ratings....

IL  = ( Vin - Vout ) Ton  /  L 

The larger the difference Vin-Vout  the higher IL current
which leads to much larger inductors and power losses

The 12V rails main power is 90% focused on VRM for
CPU and modern GPU so doubling from 12 to 24 will
just double the current on those buck inductors and so forth.

In ideal terms downsizing to 5V would be better
but 12V is still a very good figure of power

Paul

[LeonR]

I wonder why haven't the standard upped the ante to something like 24-48V.

I just hope we don't have to rely on motherboards for powering up HDDs and the alikes.

Simple facts of power ratings....

IL  = ( Vin - Vout ) Ton  /  L 

The larger the difference Vin-Vout  the higher IL current
which leads to much larger inductors and power losses

The 12V rails main power is 90% focused on VRM for
CPU and modern GPU so doubling from 12 to 24 will
just double the current on those buck inductors and so forth.

In ideal terms downsizing to 5V would be better
but 12V is still a very good figure of power

Paul

Modern CPUs work at 0.8-1.4V, using up to 150A. Not sure if having the PSU delivering CPU power at low voltages would help at all.

[PKTKS]

Modern CPUs work at 0.8-1.4V, using up to 150A. Not sure if having the PSU delivering CPU power at low voltages would help at all.

The above part is correct...

But I am not sure about your background so I will
answer for the sake of completion...

The required operating voltage for semiconductor
is not chosen randomly. It is imposed by the underlying
manufacturer process.

The 80s used 12V the 90s VLSI 5V the ultra high scale
integration (nanometers) downsized that to 2V. .1V and
it just got lower every year.

The power increases with the number of devices available
within the wafer.. so the current required growth was
exponential higher.

The voltage below 1V threshold is now required for those
common mobos and cards on the market and the above
VRM relation 12V Vin with roughly 1V Vout is the norm
for CPUs and GPUs

So if you double  Vin you also double the VRM inductors
amperage during Ton. You need more than twice as bigger
wiring and space for them...

1V today is not by choice - it is required cf manufacturer process.

My really big hope was and still is that market folks be
clever enough to stop using SINGLE ATX PSUs, instead
start using one PSU for each CPU and one PSU for each GPU

Multi PSU would solve a rather complex problem for integrators

Paul

[Scratch.HTF]

Here's what I think about 12V only ATX:

Advantages:
Reduced wiring (especially useful for small form factor units)
Higher efficiency through a single rail with use of local regulators as necessary
Easier to design and troubleshoot (faults can be localised in most cases)
Lower cost (should be)
Power supplies are easy to repurpose (should be) just like most server units with a single high output 12V rail

Disadvantages:
Reliability (possibily)
Board space for local regulators (not particularly a problem with modern high efficiency devices, especially in vertical mount modules)