Tiido wrote on 2020-01-26, 04:44:

They should introduce 24V instead, less copper, less current, less waste...

More on board VRMs requiring differant case designs and ventilation.
Ohh and watts is watts no more how you do the PI math ;p

Most switching regulators / VRMs require no sort of design changes to accomodate 24V and they'll run with higher efficiency with lesser losses allowing higher power within same thermal budget among other things. PSUs will be able to produce higher powers with lesser components also, resistive losses are much lower with higher voltage, you get more bang for same amount of power.

Sure the PSU will have less components but the motherboard will have more. CPU, RAM, PCIe, Peripherals will now each need a VRM on the board, where most today have only cpu and ram. All it does it take them from the PSU and put them on the mainboard. More parts on the board to fail. Another thing is I do not believe it will actually save much as far as actual power usage. Current good VRM and PSU are 90% efficient. So even if you up it and maybe get 92% efficiency you only gain a few percent total on a system that draws maybe max 400-600 watts max. Electric ranges, forced air heaters, electric water heaters, electric cars.. that is where the real power goes.

All the major components already have their own power regulators and have had so since P4/DDR era, there is almost nothing on a modern motherboard that uses 3.3V or 5V directly, with only USB being the only thing keeping 5V around (though with USB-C it has changed, it can deliver tens of volts if needed/capable). The few components on the mobo that might need 3.3V or 5V need only tens of mA and can just use any of the really cheap linear regulators. Expansion cards are the only things where 3.3V matters, but most regulate down from 12V instead since all modern high speed parts run off 1.2/1.5/1.8 or 2.5V.
CPU, RAM and chipset componets all run off their dedicated regulators using the 12V input on a typical motherboard, all of those could use a higher voltage instead and run even cooler with lesser losses within the switching elements and inductors. It makes all the sense to use a higher voltage since it allows both power and monetary savings across the whole machine, it very much is a win-win situation.

Horun wrote on 2020-01-26, 04:27:

If that happens older ATX PSU prices may go way up....
Also that means motherboards will have many more VRM parts leading to more future failures....

added: Intel has not built a desktop motherboard for years so am curious of their motive. Wonder if board makers will ignore the new spec for desktop boards...

The motive is that if the industry adopts their idea, they get paid royalties on each unit sold. It was the same reason why IBM created Microchannel and Sony invented Betamax.

sliderider wrote on 2020-01-26, 07:57:

Horun wrote on 2020-01-26, 04:27:

If that happens older ATX PSU prices may go way up....
Also that means motherboards will have many more VRM parts leading to more future failures....

added: Intel has not built a desktop motherboard for years so am curious of their motive. Wonder if board makers will ignore the new spec for desktop boards...

The motive is that if the industry adopts their idea, they get paid royalties on each unit sold. It was the same reason why IBM created Microchannel and Sony invented Betamax.

It won't fly in court. I bet you can find any kind of stuff out there for years, that have been powered by any kind of voltage you can think of. 5, 12, 24 or whatever. And if you go as far as integrating the voltage convertion to the motherboard and get rid of the ATX power supply, then you actually have just moved the PSU onto the motherboard instead. ATX is just an industry standard, so why not just make an open standard and name it whatever, then Intel most certainly do not have any levarage in court. Even though it only delivers 12 and minus 12. Or 24, or whatever. They can not patent a voltage rating. Just call the open standard for OPS (Open Power Supply) and you are good, and Intel will have nothing to go for.

well yeah, there are several devices that consumers can use and buy that use 48V, but I meant wide spread adoption in high current power supplies.

though I think the safety limit is something like 60v currently.

Thankfully we already have a solution to this problem, at least for lowish draw situations - PicoPSU is already designed to take 12v and spit out ATX voltages. I'm sure if necessary a version for higher wattage versions would be made.

imi wrote on 2020-01-26, 12:37:

well yeah, there are several devices that consumers can use and buy that use 48V, but I meant wide spread adoption in high current power supplies.

though I think the safety limit is something like 60v currently.

Sometimes the safety limit is less, but it depends on working conditions. Dry skin vs wet skin, possible points of contact, puncture, etc.

Pico DC DC modules.

Going all 12V concept is very very good and higher efficiency, helps with cost somewhat, and forces motherboard makers to be more efficient using switching regulators as well. HP was doing this much earlier starting with zx00 and elite 8000. Even HP used 11V standby power instead of 5V.

The most of PSUs for last 8 years have gone to main 12V power supply only, and used sub DC-DC converters fed by single 12V PSU output to generate 5v and 3.3V. So this is easy step for PSU makers to go all 12V.

I was waiting for that to happen for newer computers in future when I happens to buy one.

Lower voltages 3.3V and 5V usually goes higher losses you get trying to generate voltages with one transformer (typical!) due to space limitations in a PSU chassis and cost, in older PSUs also harder to design regulating individually voltages so these had to go by group regulation monitoring one of voltage as fixed value while other voltages fructrates somewhat.

Notebooks have been doing this already. Either from power supply adapter of one voltage output or using battery's one voltage to feed DC-DC converters on the notebook's motherboard. Not in series of DC DC converters. There is no charging circuit on the notebook motherboard's. The battery pack has communication for status and charging circuit built in.

Cheers,

pentiumspeed wrote on 2020-01-26, 21:05:

Pico DC DC modules.

That's my thinking, too. For backwards compatibility, maybe have a mezzanine board that you can plug into the ATX connector with breakouts for drives and peripherals? They could be simple, compact, efficient buck converters.

Tiido wrote on 2020-01-26, 06:54:

All the major components already have their own power regulators and have had so since P4/DDR era, there is almost nothing on a modern motherboard that uses 3.3V or 5V directly, with only USB being the only thing keeping 5V around (though with USB-C it has changed, it can deliver tens of volts if needed/capable). The few components on the mobo that might need 3.3V or 5V need only tens of mA and can just use any of the really cheap linear regulators.

I guess you forgot about good PCIe video cards and most boards still have SATA connections in newer machines, they definately do not use an on board VRM and draw way more than "only tens of mA".

Be aware x570 boards require 2 12v eps connectors 1x8pin and 1x4pin (some need 2x8pin)
most psus only have 1 8pin

imi wrote on 2020-01-26, 12:37:

well yeah, there are several devices that consumers can use and buy that use 48V, but I meant wide spread adoption in high current power supplies.

though I think the safety limit is something like 60v currently.

Well, it varies depending who you ask, but 24V is the most conservative treshold I've heard/read about IIRC.

If you manage to pierce the skin, even 24V can be somewhat unpleasant. Happened to me a few times, those cable strands or soldered through hole leads can be pretty sharp 😀

Horun wrote on 2020-01-27, 05:52:

Tiido wrote on 2020-01-26, 06:54:

All the major components already have their own power regulators and have had so since P4/DDR era, there is almost nothing on a modern motherboard that uses 3.3V or 5V directly, with only USB being the only thing keeping 5V around (though with USB-C it has changed, it can deliver tens of volts if needed/capable). The few components on the mobo that might need 3.3V or 5V need only tens of mA and can just use any of the really cheap linear regulators.

I guess you forgot about good PCIe video cards and most boards still have SATA connections in newer machines, they definately do not use an on board VRM and draw way more than "only tens of mA".

PCIe power connectors deliver 12V, it's not the 'good' cards that use 3.3V but the low-end ones, and low-end discrete GPUs are a dying breed, as integrated GPUs are 'good enough' for pretty much any task that does not require a big monster with its own PCIe (12V) connector.

As for SATA, four letters: NVME. Optical drives are a thing of the past, HDDs have moved from desktop PCs to NAS solutions, and we're getting very close to price parity per GB between HDD and SSD. Once we hit that - probably within a year - even budget PCs will ship with SSDs, and those SSDs won't be connected via SATA but plugged straight into an M.2 slot.

Consider that a 250GB SSD (easily enough for modern OS and desktop workload) costs less than EUR 40 these days.

I wonder how difficult an adapter from this new standard to "old" ATX would be to make.

Actually I wondered this also for some while before hearing about this new standard for the older ATX standard where 5v is more relevant than 12V.
Would an adapter from 12v to 5v for socket A boards be costly to manufacture ?

Deksor wrote on 2020-01-27, 09:12:

I wonder how difficult an adapter from this new standard to "old" ATX would be to make.

As already mentioned they already exist: PicoPSU

Actually I wondered this also for some while before hearing about this new standard for the older ATX standard where 5v is more relevant than 12V.
Would an adapter from 12v to 5v for socket A boards be costly to manufacture ?

Probably not hugely so. It might however shift the balance on whether it's worth refurbishing old PSUs (replacing those caps) instead.

Horun wrote on 2020-01-27, 05:52:

I guess you forgot about good PCIe video cards and most boards still have SATA connections in newer machines, they definately do not use an on board VRM and draw way more than "only tens of mA".

I did talk about motherboard components not offboard things. Local voltage regulator on a would be 12V only motherboard will have no problem supplying the necessary currents on any of the 3.3V or 5V requiring devices, and it will produce much better output than whatever is in a power supply solely on the fact that there's no long cables with excess inductance in the way, line and load regulation will probably be an order of magnitude better. HDD power could also be provided by some cables that run next to the existing SATA connector, i.e combined SATA+Power cable with same connector on both ends (which would probably be more elegant for those that like tidy cabling).
Vast majority of PCI-E cards generate their own voltages from 12V, since most modern components run at voltages below 3.3V and again local regulator would have no problem providing that power.

dionb wrote on 2020-01-27, 09:24:

As already mentioned they already exist: PicoPSU […]
Show full quote

Deksor wrote on 2020-01-27, 09:12:

I wonder how difficult an adapter from this new standard to "old" ATX would be to make.

As already mentioned they already exist: PicoPSU

Actually I wondered this also for some while before hearing about this new standard for the older ATX standard where 5v is more relevant than 12V.
Would an adapter from 12v to 5v for socket A boards be costly to manufacture ?

Probably not hugely so. It might however shift the balance on whether it's worth refurbishing old PSUs (replacing those caps) instead.

Sure, but PicoPSU aren't that powerful ... I doubt it would work well alongside a Athlon XP 3200+ which can take just for itself up to 76W !
We'd need more powerful pico psus imo

I doubt it will be a problem for next 10+ years. Niche for old style PSUs will be too big to ignore by manufacturers. It's not even a big problem to acquire new AT PSU after 20+ years.