How on earth does this power supply work?

[T3sl4co1l]

Weird, I've not taken apart one which didn't have one!

12V supplies make sense both from being more power-dense, and for being more consistent with mobile platforms -- laptops generally have a master 14-20V rail from which the rest are supplied (including battery management and ideal-diode circuits).  Wouldn't mind 24V really, you get even more power through less wire.

Tim

[dom0]

I have to say as an amateur radio operator I have a slight bias toward Anderson Powerpole connectors. Reliable, configurable, and the connector size is fairly small despite 30A per pin rating! Other nice thing is the gender-less design. I'd like to see those used in PCs too.

I think we can do better:



Should be good for a couple hundred amps :-)

[DanielS]

They also need to come up with a higher current standard connector.

Not really: the high-current version of MiniFitJr can already handle 13A per pin, so the 6-pin connector could handle GPUs up to about 450W if the GPU measured wiring losses to detect the use of high-current connectors and heavier gauge wires, then set power limits accordingly.

On the other hand, if Nvidia's last few generations are any indication, the trend is for GPUs to get considerably more power-efficient with a heavy focus on making more efficient use of their limited memory bandwidth. When the 14/16nm chips come out, I would not be surprised if even high-end single-GPU cards ended up well under 150W: the 980Ti in stock reference configuration uses only 180W and that's still a 28nm part. The days of high-end (single) GPUs being power hogs are coming to an end.

[mariush]

Most modern "high-end" ATX supplies use DC-DC converters to produce the 3.3V and 5V outputs. Eventually I think we'll get to a point where even this will be dropped and PC PSUs will be 12V-only.
Personally, I can't wait. This sure will take some of the rat's nest out of computer cases!
They also need to come up with a higher current standard connector. It's getting a little out of hand with some graphics cards having 12 or more pin 12V power connectors to supply the necessary current, each with pin having its own 18AWG wire. IMHO they should pick a new standard connector with a much greater current rating for the motherboard, graphics card, and any other high-current hardware devices.

The USB 3 standard introduces 20v as a valid voltage for usb devices. A usb device can choose between 5v, 12v and 20v and by default everything defaults to 5v. AFAIK there's few controllers which actually support negociating for higher than 5v voltages and probably no motherboard that would negociate to 12v for usb 3.

There are already some motherboards that have a standard 2.5 barrel jack that allows the motherboard and everything to be powered from an 18-19v laptop adapter (the mb has a bunch of sata power connectors on it and the board has dc-dc converters to produce 3.3v and 5v) but such motherboards still have the atx connectors as an option.

I would love to see power supplies drop 3.3v and -12v and offer just 5v (because dropping 5v stand by would be too big of a change) and 12v (for compatibility with existing video cards and sata drives and fans) and 20v (for cpu and mb and future video cards). 

[amyk]

Can't say I've ever seen that in any ATX power supply I've ever taken apart.

Most affordable ATX power supplies (as in sub-$80 or so) are group-regulated affairs in either flyback or half-bridge configurations. The oldest ones regulate the 5V output while the not so old ones regulate based on the 3.3V sense line. In modern Haswell-ready models though, 12V is the new king.

I do not remember seeing magnetic regulation in any of the PSUs I have opened either. The only reason I knew about them is because I skimmed a chapter about them in a switching supply design book many years ago.

http://danyk.cz/s_atx_en.html

29/44 of those ATX PSUs there use a magamp to regulate the 3.3V, the other 15 use either a discrete linear regulator or a separate coil on the transformer (more expensive).