Monday, May 17th 2021
Intel Encourages Adoption of ATX12VO Standard on Alder Lake-S Motherboards
The ATX12VO power standard is a new specification for desktop power supplies which boasts greatly increased efficiency over regular desktop power supplies. The new standard requires a compatible motherboard with a 10-pin power connector along with a compatible power supply which only features 12 V rails. The standard requires that any voltage conversion above or below 12 V must be performed directly on the motherboard which increases the complexity and cost for motherboard manufacturers. Intel is interested in promoting the standard with their upcoming 600-series motherboards for Alder Lake-S however most enthusiast boards are unlikely to feature the standard. The standard may find higher adoption with entry-level motherboards for system integrators and pre-built suppliers who need to meet strict government power efficiency regulations.
Sources:
VideoCardz, Hardware LUXX
124 Comments on Intel Encourages Adoption of ATX12VO Standard on Alder Lake-S Motherboards
The bulk of the power draw is from the 12V rail and there is no change to that apart from the connectors.
Most of those 650W PSU costs around $100 or less.
Anyway this is a cheap $45 400W unit, and the efficiency is no where near as bad as suggested.
www.techpowerup.com/review/kolink-classic-power-400-w/5.html
For PC PSUs it is the usually expensive high power units that have lower efficiency at idle.
www.techpowerup.com/review/be-quiet-dark-power-pro-1500-w/5.html
I'm sorry if I sound harsh, but this is another unnecessary idiotic step that Intel shouldn't bother. Power supply now is very efficient, with DC to DC converter, full bridge rectifier and even very good power filtering on each rail. I would like to see if all of that is left to motherboard manufacturer. Currently, there are many types of power delivery motherboards, whether it's single phase inductor or double phase double inductor, with single or double tunnel controlling. Many motherboards look decent but don't provide same efficiency, let alone adequate cooling. Not to mention how do you provide arbitrary "Standard" on every motherboard? Will the price not go up because of that certification? Kinda amuse Intel to take a winding road just to make the Elder lake look good.
Well, cheap units have been decent for a long time, but still I wonder why Linus saw massive gains in efficiency just from PSU change.
Better power management, why?
Motherboard idle power consumption can vary wildly especially on Intel motherboards, because different units enforce different Power Limits and some just choose to ignore it completely.
Asrock mobos tend to follow the Intel spec as Auto, while other makers often run the CPU at full boost with MCE as Auto.
Also something that people often overlook for idle power consumption is things like RGB LEDs actually adds up, turning off all of my RGB actually drops my idle power by up to 20W.
They are already doing 12V only PSUs for years.
Don't see the point here for DIY PCs
If doing DC - DC after AC - DC 12V is really that efficient and simple, just slap the same components into the PSU and reuse everything else.
Mid-range motherboards will never have good efficiency. Only high-end ones.Nah, copper is costly these days.
Also, let's keep in mind that people change motherboards way more often than PSUs.
You can keep a good PSU (like Corsair RM850x) for 10 years.
Chances are you're not going to keep the same motherboard for 10 years (especially if we're talking about Intel).
So the consumer in the end will have to pay more for upgrades... and if you care about efficiency, you're going to pay even more for high-end mobos, because I don't think mid-range ones (let alone low-end ones) will offer the same quality in DC-DC parts.
The card Linus used in his video was a Titan X, very similar to a 1080 Ti. The card pulls around 10Ws at idle. FYI an RX580 draw around 9W idle.
So, Linus' idle power consumption on his standard ATX system is actually quite inline with a 9900k on a "non-stock" behavior.
Well, if i9 is really all that conservative with power at idle, then I guess Linus's results could be skewed a lot by different motherboard. However, there's one thing. His rig still consumed less power with ATX12v0 PSU and that wattage was below anything you could achieve on stock part with normal power supply. So new standard definitely helps, just not as much as I thought it would.
An we have to remember that TPU reviewer likely adjusted voltage manually and stock 9900K can only boost to all core clock of 4.7GHz. With MCE Linus could have had closer to 5GHz or 5.1GHz speed with much higher voltages. Meaning that normal power supplies are even closer to ATX12v0 than I thought.
I'm really curious about this stuff, and it's a damn shame that Linus did such a shoddy introduction and didn't control variables better. He also got his hands on rare and likely impossible to buy product too. And now there's no accurate data about it. What a shame.
Those SATA power out are fed by the 6-pin 12V connector next to them, so you measure the difference in the input and output wattage you can get an idea of how efficient that is.
Compare that to a standard ATX PSU and you can get a better idea. (Not sure if that only feeds the SATA power or the whole motherboard though)
...but maybe I'm just a pessimist.
It will get new socket, likely DDR5, new architecture, likely node shrink and maybe big.LITTLE rip off technology. Once you look at everything, Intel will likely push power efficiency as a major feature. A new PSU standard (if it does what it claims to do) would be a nice addition. If Alder Lake doesn't suck and they put big and little cores on same chip, it can truly redefine desktops as we know. If it fails, then we will have some sort of bullshit like AMD FX with loads of Windows scheduler problems that weren't completely patched out until next major Windows release. I personally would like to see 8 big cores, 16 small cores and those 8 big cores hyperthreaded. If they can make it on new process node and not without some big sacrifices and at decent clock speed and turbo speed and still within reasonable TDP (let's say PL1 - 80 watts and PL2 - 150 watts, Tau 58 seconds), then it would be really interesting release. If they still can't get anything smaller than 14nm, then it will be yet another backported abortion. But if they can, that's going to be good. Meanwhile, AMD will be stuck with all even cores, but high core count. Alder Lake could be very interesting for multithreaded loads, but I wouldn't expect it to be much better at single threaded loads.
I see no reason to be so grim about them. They are very well aware that Rocket Lake is a stop gap and they are trying to remain respectable brands and reduce losses until they have something good. Creating a new architecture takes time and some planning for future. Rocket Lake likely was already engineered and functional some years prior.
Well of course, divide them into four, added some Crapxon polymer, Nikos D-PAK and Isat 200μH inductor is waaayyyy better than 24 pin motherboard :p
Motherboards already have tons of voltage conversion components on them; yeah they'll need a few more components for this to work but doing it in the PSU only creates a bunch of redundant componentry and complexity. If its done on the motherboard it will be simpler in terms of total design vs. having voltage conversion capability in two places needlessly. Then there is the efficiency and PF gains; efficiency and PF are bad at low loads on low / midrange PSUs because they don't have the high-end voltage conversion components that already exist on a motherboard to keep idle power consumption in check. Also the 24 pin connector is such a stupid bulky out dated design full of useless cables it really is time for it go.
BTW some old board, which had EPS connector, but 4 pin one actually managed to put it in different places other than that damned corner. Here are some examples:
You see, they could even put it where 24pin is now, I see no reason for EPS to bee in that cursed place. It's really not a rocket science to switch RAM position and rotate CPU socket. Really simple stuff, but it seems that nobody cares.