News ATX12VO spec: does it matter for us?

[DrMrLordX]

How Intel is changing the future of power supplies with its ATX12VO spec

Intel's new ATX12VO spec eliminates 3.3-volt and 5-volt rails and rejiggers the balance of power in future PCs.

www.pcworld.com

Looks like Intel has developed a new PSU spec for OEMs. Everything moves to the 12v rail. Hmmm.

 

[VirtualLarry]

Great for OEMs, perhaps, but the extra voltage converters / VRMs on motherboard, won't be welcomed by enthusiasts, due to losing even more floorplan real-estate on mobos, that could be used for useful features.]

Edit: And a really dislike having to pull power for SATA devices off of tiny little proprietary mobo connectors on OEM boxes, makes it hard to upgrade and add a drive.

 

[LukeSavenije]

Great for OEMs, perhaps, but the extra voltage converters / VRMs on motherboard, won't be welcomed by enthusiasts

in defense, you can exactly make the coil to the amount of power x drives or something else that would need 3.3v or 5v, making it more efficient

with that you do have some more room by eliminating the 24 pin for a 10 pin connector, so the difference won't be huge

 

[gorobei]

in the short term no, but eventually

you will lose some space on the mb when you move the 3.3/5 v components (some normal vrm size caps, a fairly large coil, and some transistors). you will lose even more space when mb makers have to put 5v sata power cable socket headers somewhere around the crowded edge of the board. one for each pair of sata data cable sockets. you will also need some cooling or a heatsink for the components. unless they go to a longer pcb something is going to get less space. or everyone will have to agree on how to mount important parts on the back of the mb and all the case makers agree to put a cutout there.

gordon talks about this on the pcworld podcast, it is an opportunity to ditch the atx layout for pc and do something that has better spacing for modern builds.

 

[Mr Evil]

...extra voltage converters / VRMs on motherboard...

Most things have already need to convert voltages. The CPU and GPU are obvious due to the masive size of the regulators, but all sorts of other things run on a voltage that isn't provided by the PSU, and even when one is provided by the PSU, they may not use it, e.g. SATA devices that convert 5V to 3.3V, even though SATA provides 3.3V directly. The use of 5V and other low voltages from the main PSU will only decrease in the future as reducing power consumption dictates that devices use lower and lower voltages, which can be most efficiently generated from a high voltage at the point of load, rather than trying to send them over long wires.

 

[VirtualLarry]

The other problem, with essentially moving portions of the existing ATX PSU onto the mobo, is points of failure. If most PSUs last 5 years, but mobos with solid caps last 10-15, that's a problem if they're moving some of the circuitry onto the mobo, that only last 5 years, and thus, the entire mobo might have to be replaced after 5 years, like the "bad old days".

 

[ch33zw1z]

Lenovo has been doing this for a while. 12v to the board, 4pin connectors for power to disks and cdroms

 

[mindless1]

The other problem, with essentially moving portions of the existing ATX PSU onto the mobo, is points of failure. If most PSUs last 5 years, but mobos with solid caps last 10-15, that's a problem if they're moving some of the circuitry onto the mobo, that only last 5 years, and thus, the entire mobo might have to be replaced after 5 years, like the "bad old days".

Most PSU only last 5 years (which might be a fair average, but is highly variable) for three main reasons:

1) Fan failure, so the same fan in new design PSU won't change PSU mortality from fan failure, given same thermal margins.

2) AC power surges. Again more or less same risk to PSU whether 12V or multi-rail.

3) Capacitor failure. This is primarily cost cutting measures by PSU manufacturers using electrolytic capacitors. PSU that have solid/polymer caps seldom have those fail within the life of the PSU, so with so many mobo caps now being this type, it should be expected that these more robost caps would be used for the subcircuits added to arrive at the lower voltage rails on the mobo.

I don't think it will take up much more mobo real estate, since there are already so many ICs needing lower voltage anyway with their own buck regulator circuits for that, so more of them just stepping down from 12V instead of 5V. It may raise the cost of the mobo some, but "should" lower the cost of the PSU at least as much, and reduce the required size of the PSU.

I see it as a step in the right direction. It's something I wanted to happen well over a decade ago, but concede that at the time it wouldn't have worked out well because most boards at that time used marginal quality electrolytic caps, and going even further back, less efficient transistors so they took up more space and needed more copper PCB area too.

It could pose a little more challenge for DIY builders to make sure they have adequate airflow in the associated mobo regions, yet not a lot because few other things use anywhere near as much power as a CPU or video card.

I just don't want them messing with the PSU rear case mount layout because I have a fair amount of cases that I'd like to reuse, but I could crank out some adapter plates for that if I had to.

 

[jonnyGURU]

As @ch33zw1z said, Lenovo, as well as HP and Dell, have been doing this for a while.

Intel is just trying to standardize things.

Hardly anything uses +3.3V from the PSU. Even RAM uses the +5VSB (which is now +12VSB, oddly enough). The big push back is going to be from mobo manufacturers that don't want to have to put MORE DC to DC on their motherboards... but they're already putting SOME DC to DC on their motherboards today. Essentially, the only thing they'd have to add is power for SATA drives, and if the majority of users are ok with using M.2, then we don't even need those.

 

[VirtualLarry]

Essentially, the only thing they'd have to add is power for SATA drives

Would we instead start to see SATA SSDs, that could take 12V from the SATA power plug? Or 12V / 5V, and convert from either/both? Is there enough room on a SATA SSD PCB to down-convert 12V to 5V or 3.3V? (*If they're already down-converting from 5V to 3.3V internally, would it be hard to accept 12V input as well, and down-convert that too?) I guess I don't have a good idea how much PCB space that it would take. (*I know that regular 3.5" HDDs take 12V for the spindle motor, and 5V for the electronics, but I don't know if they down-convert at all.)

Edit: I guess, lastly, if the last bastion of SATA drives, SATA SSDs, got to be 12V-input only, alongside 12V-only output ATX PSUs, couldn't they then simplify the SATA connector, to route both data AND 12V power from the mobo? Could they simply use the same singular SATA data cable, to then carry drive power as well as data signals, and skip requiring connecting the SATA power connector? (It might physically be present, for chassis/legacy reasons, but electrically unused?) Basically, rather than connect SATA data to mobo, and SATA power to PSU, connect PSU to mobo, and route 12V to the SATA data connector sockets, and then plug SATA data from mobo to drive, also carrying 12V power to the drive?

 

[Mr Evil]

...I guess I don't have a good idea how much PCB space that it would take...

Here's a photo of a DC-DC converter module that I have used before. It's capable of providing more than enough power for an SSD, is very cheap, reliable (note that it uses ceramic capacitors, not electrolytic), and would easily fit into a 2.5" drive.

 

[jonnyGURU]

Would we instead start to see SATA SSDs, that could take 12V from the SATA power plug? Or 12V / 5V, and convert from either/both? Is there enough room on a SATA SSD PCB to down-convert 12V to 5V or 3.3V? (*If they're already down-converting from 5V to 3.3V internally, would it be hard to accept 12V input as well, and down-convert that too?) I guess I don't have a good idea how much PCB space that it would take. (*I know that regular 3.5" HDDs take 12V for the spindle motor, and 5V for the electronics, but I don't know if they down-convert at all.)

Edit: I guess, lastly, if the last bastion of SATA drives, SATA SSDs, got to be 12V-input only, alongside 12V-only output ATX PSUs, couldn't they then simplify the SATA connector, to route both data AND 12V power from the mobo? Could they simply use the same singular SATA data cable, to then carry drive power as well as data signals, and skip requiring connecting the SATA power connector? (It might physically be present, for chassis/legacy reasons, but electrically unused?) Basically, rather than connect SATA data to mobo, and SATA power to PSU, connect PSU to mobo, and route 12V to the SATA data connector sockets, and then plug SATA data from mobo to drive, also carrying 12V power to the drive?

Intel is not going to turn over the entire hard drive industry. It's too wide spread. They are simply going to have to put DC buck converters on their motherboards and power connectors to deliver that power if they want the customer to be able to use HDD's and SSD's. And if they want the board to support more drives, they're going to have to include more powerful DC to DC converters and more connectors. That's all there is to it.

Remember: The SATA consortium (SATA IO), the PCIe consortium (PCI-SIG) and Intel are all separate entities. It's a miracle anything works at all.

Did you know that SATA DROPPED the +3.3V from it's spec way back in 2013? Then in 2016, they REPLACED the +3.3V with a PWDIS signal to put drives to sleep (ironically, they use a +3.3V signal, so a SATA connector that has +3.3V effectively keeps the drive from functioning at all!). Yet Intel NEVER REMOVED the +3.3V from THEIR spec. So right there you have CONFLICTING specs. Now they'll be removing it from the next revision of the ATX12VO spec (I have but can't share because of NDA), but the MAIN REASON they're doing it is not to align with PCI-SIG, but to make more space on the motherboards by not requiring it to be there!

 

[VirtualLarry]

It's a miracle anything works at all.

LOL! Too true! PCs are miracles of engineering. Literally.