How Intel is Changing the Future of Power Supplies With Its ATX12VO Spec (pcworld.com) 188
"We don't often talk about power supplies, but Intel's new ATX12VO spec — that's an 'O' for 'Oscar,' not a zero — will start appearing soon in pre-built PCs from OEMs and system integrators, and it represents a major change in PSU design," reports PC World.
"The ATX12VO spec removes voltage rails from the power supply, all in a bid to improve efficiency standards on the PC and meet stringent government regulations. But while the spec essentially removes +3.3-volt, +5-volt and -12-volt and +5-volt standby power from the PSU, they aren't going away — they're just moving to the motherboard. That's the other big change..."
Long-time Slashdot reader UnknowingFool explains: If those voltages are required, the motherboard is expected to supply it by converting 12V power. Removing those two rails will shrink the connector from 24 pin to 10 pin.
One of the main reasons is increased flexibility for future needs. That dependence on these separate rails has been reduced over time but current power supplies must dedicate a fixed amount to those rails. The example given is that a current 600W PS may only need 15% (90W) of total power dedicated to those rails but a decade ago may require 25% (150W). Switching to the MB supplying the power means more 12V power.
A benefit is power supplies will be cheaper to make and also more power efficient. With two fewer rails, there is fewer connectors and parts and power supply makers can focus on efficiency of just one rail.
"The ATX12VO spec removes voltage rails from the power supply, all in a bid to improve efficiency standards on the PC and meet stringent government regulations. But while the spec essentially removes +3.3-volt, +5-volt and -12-volt and +5-volt standby power from the PSU, they aren't going away — they're just moving to the motherboard. That's the other big change..."
Long-time Slashdot reader UnknowingFool explains: If those voltages are required, the motherboard is expected to supply it by converting 12V power. Removing those two rails will shrink the connector from 24 pin to 10 pin.
One of the main reasons is increased flexibility for future needs. That dependence on these separate rails has been reduced over time but current power supplies must dedicate a fixed amount to those rails. The example given is that a current 600W PS may only need 15% (90W) of total power dedicated to those rails but a decade ago may require 25% (150W). Switching to the MB supplying the power means more 12V power.
A benefit is power supplies will be cheaper to make and also more power efficient. With two fewer rails, there is fewer connectors and parts and power supply makers can focus on efficiency of just one rail.
Who wrote this? (Score:2)
With two fewer rails, there is fewer connectors and parts and power supply makers can focus on efficiency of just one rail.
I hope it was the person describing the difference and not someone from PC World.
Re: (Score:2)
Re: (Score:3)
Am I being cynical that I'm expecting the price the end user pays for power supplies to say the same (or effectively even go up as higher power PSUs will be the norm at the price of the older PSUs at the same power rating) while the cost of a motherboard will climb 20% for the extra components?
Re: (Score:3)
Now that we have very fast power transistors, it is easy to make switch mode power supplies at point of need. This results in more accurate delivery of the voltage required and less noise on the power lines.
Motherboards need to meet very demanding spec, and have to be designed and made by competent people, or they just won't work.
PSUs are not particularly critical, and, after this change, will be able to
Re: Who wrote this? (Score:4, Informative)
It supplies only 12V power but it has a hub that connects to SATA, 24 pin, etc. Really it moves the modularity from separate cables to a hub with modular cables.
Re: (Score:2)
Any voltage conversion is going to generate heat. Move it from the PSU to the component will reduce heat generated by the PSU, and generate it at the component. That won't generally be a problem for your domestic desktop, but laptops and servers are whole different ballgame.
On the positive side, the mainboard power traces can be thinner because they're all 12V until they reach the component.
Re: (Score:2)
Laptops and servers are already a whole different ballgame.
Re: (Score:2)
Moving the problem does not get rid of the problem (Score:5, Insightful)
So, they're talking about how much more efficient it will be without the rails on the power supply - the one thing in your PC that is completely specialized in power conversions.
Now, instead, they're putting it on the motherboard, a highly complicated component that is very unlikely to be as efficient at doing the power conversion.
On top of that, the already often overcrowded motherboards have to find space for even more components and connectors? Wonderful.
This is a good idea why, again?
Re:Moving the problem does not get rid of the prob (Score:5, Interesting)
Re: (Score:2)
That's not how it works. Intel gets a cut for the Intel components, the rest is fixed priced regardless of how much you sell the board for.
Re: (Score:2)
You'be been astroturfing this thread pretty hard so far. It's quite obvious in this case that intel stands to benefit from this, as its board partners will be willing to pay more as they can charge more for additional components and benefit from reduced MTBF rate forcing customers to replace motherboards more often instead of a cheap PSU.
Re: Moving the problem does not get rid of the pro (Score:2)
What was the "problem", exactly?
Re: (Score:2)
A specification that over designed powersupplies to cater for lowest common denominators leading to massively inefficient power conversion at the 5V and 3.3V rails in addition to extra hardware and cables connecting the two all presenting a non-zero cost.
Given that most devices in computers already provide a level of voltage conversion to suit themselves it doesn't make any sense for the PSU to dedicate 5V and 3.3V to the motherboard.
Re: (Score:2)
The problem was that Intel did not get a "cut" of the proceeds from Power Supply failures and replacements. By moving more of the power conversion onto the motherboard, Intel now gets a cut because the motherboard will have to be replaced more often.
Sorry but that's just horseshit. There's no inherent reliability issue with voltage regulation that would make any difference at all to Intel's petty cash figures. The most unreliable part of the existing PSU is the voltage conversion from line voltage to 12V and that stays there. The second most heavily loaded power regulation system in a PC is on the GPU and the 3rd is already on the motherboard where votlage regulators already have to provide 100s of watts at tight tolerance to the CPU. By comparison the 5V and 3.3V are trivial to design, trivially loaded, and won't make a lick of difference in any reliability.
Re: (Score:2)
Note that Intel did the same thing with their last "Power Supply Improvement" which moved the power regulation and filtering components from the Power Supply onto the Motherboard, and they made this design change for exactly the same reason: to ensure that they got a "cut" of the lucrative Power Conversion failure market.
What's your view on COVID-19 being a conspiracy to rid the world of the Iranian rulers, and the CIA taking down the world trade centre on purpose to justify a war on terror?
Re: (Score:2)
Wasn't that 2000s, back when there was a specific factory that had some kind of a production problem that was supplying a lot of OEMs?
Leaky capacitors after a few months on wide variety of computer hardware was one of the things that jump started the marketing buzz of printing "japanese capacitors" on the box back then.
Re: (Score:2)
The production problem was a stolen proprietary formulation for electrolyte. The problem is that the documents contained a poison pill that would result in inferior failure prone caps if it was followed exactly...
Re: (Score:3)
Here [independent.co.uk] is an article, though it says the formula was mis-copied rather than a poison pill. A few more details here [theguardian.com]. I'm sure that at the time I read about the error being due to a pioson pill, but a lot of the articles seem to have gone to the bit bucket.
Re: (Score:2)
Re: (Score:2)
And how exactly is moving all of that to the motherboard any more efficient than having the power supply take care of power balance? Why can the PSUs not be designed to properly balance power to the different rails?
Re: (Score:3)
And how exactly is moving all of that to the motherboard any more efficient than having the power supply take care of power balance? Why can the PSUs not be designed to properly balance power to the different rails?
What you are asking is that the PSU be far more intelligent. It will require circuitry, maybe specialized chips. Current PSUs only supply power and a fixed max amount of power. That's all they are asked to do. In the future the PSU will do exactly the same but with fewer rails. Asking the PSU to regulate power to different rails is something that MBs do now especially with VRMs.
Re: (Score:2)
There is no reason the PS can't be built to provide it's entire capacity to the 12V raul internally and use it to power supplies for 5.5v and 3.3v internally. Those switching supplies would have a lot more room and so better cooling. The result is a PS that balances the power itself, no intelligence required abd the world doesn't have to change around them. For bonus points, that could be a module swappable by someone of modest skill.
Of course, with all of that, you still need to have an always-on 3.3v supp
Re: (Score:2)
Re: (Score:2)
Technically, that's more efficient, but the losses over the size of even a large server are miniscule.
Re: (Score:3)
PSUs can't balance rails dynamically. There's fixed limits for everything, and presently the rails provided to the motherboard are massively overspec'd for what the board requires thanks a lot to legacy requirements. Then there's the wiring issue which at present is a clusterfuck of voltage rails, connectors are larger and more expensive, and if the motherboard took care of the requirements for power it can be designed to meet it's requirement, not some silly older standard leading to lower loads on 5V and
Re:Moving the problem does not get rid of the prob (Score:5, Insightful)
To be fair, the motherboards are already in the business of having to step up/down voltage. Processors haven't run at 3.3V for a long time. PCIe cards when running for real are fed 12V (max 75W through motherboard, the rest comes in directly at 12V), and receive 3.3V for standby power (no more than 10W/adapter). USB does use 5V, but also may have to switch to 20V for USB-PD depending on the circumstances.
In short, there is very low power demand for any voltage other than 12V and plenty of voltage regulation already required on the board because they pretty much stopped adding to the ATX spec after 3.3V and started moving voltage regulation downstream. It's a redundant situation now with overly complex PSUs that don't save the board designers from having to deal with the problems anyway. Sure, some stuff may move onboard, but the PSUs had to oversize the low voltage rails 'just in case', which the motherboard is in a better position to know how much they need to size for their embedded components, so they may be able to go with a more appropriately sized solution depending on how much expansion they support and what components they chose for their board.
Re: Moving the problem does not get rid of the pro (Score:2)
CPU and RAM can operates at voltages as low as ~0.4V and more typical 1.2V because high speed switching can be faster and less noisy at very low voltage.
Re: (Score:2)
Problem: "My PC is occasionally hanging with mysterious device errors."
Solution: "Try swapping the PSU. It might not be delivering voltage within spec."
Future
Problem: "My PC is occasionally hanging with mysterious device errors."
Solution: "Buy a new motherboard. It might not be delivering voltage within spec. This will require you to remove the CPU and RAM, unplug all your drives, remove any add-on cards you might have such as a GPU. You'll need to scrape the thermal paste off the CPU and clean
Re: (Score:2)
So, they're talking about how much more efficient it will be without the rails on the power supply - the one thing in your PC that is completely specialized in power conversions.
Not really. Much (if not most) of the power conversion in a PC is already happening on motherboards, where the adjustable 1.X volt source that is needed by the CPU is created by a DC-DC converter. This is just expanding that concept to cover all the miscellaneous voltages needed by the system.
Re: (Score:2)
Re: (Score:3)
So, they're talking about how much more efficient it will be without the rails on the power supply - the one thing in your PC that is completely specialized in power conversions.
There's nothing specialised about power regulation in the PSU. In fact the single most specialised rail regulation is already on the motherboard, Vcore, the one that converts 12V at hundreds of W for a processor which presents incredibly transients loads to the regulators. The other specialised conversion is on the GPU which is the same. A shitty ATX powersupply by comparison is nothing more than utter garbage and many on the market could be designed by 1st year engineering students (though there are some g
Re: (Score:2)
Thanks for the informative response. It clears up quite a few things.
Re: (Score:2)
Sweetie, the motherboard already has a number of voltage regulators. That's how, for instance, it uses a 12V supply to power the CPU. Voltage regulators are not particularly complicated compared to most other things on a motherboard, and running a 12V supply would be more efficient than running 3/5V from the PSU because resistive losses are proportional to current.
Re: (Score:2)
It IS more efficient.
Firstly you only need one high power AC to DC converter. They tend not to be very efficient at very low power (sleep mode) so by putting it all on one converter it improves efficiency a bit by raising the minimum load.
Secondly it moves the regulation to where the power is used, reducing losses in the cables and traces. Motherboards already take 12V and regulate it down to voltages they need because hundreds of watts over 50cm of cable at 3.3V will require extremely thick wires and have
Why 12V? (Score:2)
If you are going to redesign the spec so deeply then why not set the DC voltage at 20V. This would be the highest voltage also seen in USB wall-warts for high speed charging. Lots of low-power PCs could then skip the internal power supply altogether without needing a separate spec.
Re: (Score:2)
Re: (Score:2)
But you are already going to put the step down circuitry on the motherboard by design, in this new spec. So why not up the voltage in, allow 100W in via standard commodity USB-C supplies with standard connectors which will drop in price a lot with economies of scale?
Re: (Score:3)
Stepping down requires components for it, and costs efficiency. If you can use 12V for some things, then you can skip having to stepdown that part. Your input is 12V, and you have stepdown stages for 5V and 3.3V. If you use 20V then you have to stepdown everything, and need an extra 12V stepdown bit.
Re: (Score:2)
Efficiencies of 99% are possible and available.
https://www.allaboutcircuits.c... [allaboutcircuits.com]
If you can source 100W and deliver 99W via a cheap commodity 20V, 5A supply then that is far superior to sourcing 60W and delivering 59.4W via a 12V, 5A supply. notice, that USB-C connectors (which are commodity, cheap and understood by consumers) are small and upping the current in them is unlikely in the future, so 5A is where it is at.
The difference in power makes a qualitative difference in actual computing. 60W PCs are rare
Re: (Score:2)
A lot of people have ancient tech in their heads where losses on buck converters were 5% or even more. Today, 98% efficiency is common:
https://www.digikey.com/en/art... [digikey.com]
and 99% efficiency is very possible. Conversion efficiency is an important design consideration but it is no longer the determining factor for choosing the operating parameters for the spec.
Re: (Score:2)
Re: Why 12V? (Score:2)
With modern chips you could negotiate the voltage between 2.5 and 25V, which would cover any usage scenario and eliminate waste. The problem is that your wires size according to voltage. Higher voltages require less thick wiring per watt but also more waste when conversion is necessary. Most of the power hungry and electrically noisy components are already at 12V (video cards, hard drives, fans) so you really don't want those lines passing through the motherboard
Re: (Score:2)
I am not following the logic here. You get 20V, 5A in via a USB-C port. You have buck converters in a dedicated area of the motherboard. And then you run it wherever it is needed. What's the problem?
And yes, the whole point is to run power via thinner lines and smaller commodity connectors (namely USB-C). The whole point is to enable laptops, dumb terminals, TV boxes, regular PCs, TVs, and monitors to use the same power supplies and connectors and take advantage of the economies of scale for commodity PSUs.
Re: (Score:2)
Re: (Score:2)
If you are going to redesign the spec so deeply then why not set the DC voltage at 20V.
They aren't redesigning the spec that deeply which is precisely why they standardised on the existing most highly loaded rail required in a PC. Moving 5.5V and 3.3V regulation to the motherboard is minor compared to changing the entire power distribution system within a PC. You'd be looking at a redesign of every single PCI device on the market, and redesigning regulators for CPUs, and GPUs, both of which have built an industry around standardised power MOSFETs finely tuned to 12V > ~1.2V conversion. Red
Re: Why 12V? (Score:2)
Re: (Score:3)
Re: (Score:2)
The higher voltage only runs from the PSU feed to the onboard buck converters. 20V or 48V aren't a problem for board design. Even higher voltages are seen on many PCBs without problems.
Preexisting standards and 12V is enough. (Score:5, Insightful)
The higher voltage only runs from the PSU feed to the onboard buck converters. 20V or 48V aren't a problem for board design. Even higher voltages are seen on many PCBs without problems.
Actually, 48V is an issue. You need more clearance. You need higher voltage input down-converters near the loads - and dissipate more power right there, next to the real load. To avoid resistive losses (and provide capacitance and ground plane continuity for high-frequency signal lines) the raw power is distributed in full-board planes rather than fat paths to the downstream regulators, so ALL the through-holes need a little more clearance, and it goes on.
About a decade ago, working for a networking equipment maker and WITH the guys designing the blades (about the size of a laptop each and pulling a couple hundred watts apice), I got some education in this. We had redundant 48V raw power - from the "battery room" busses in big centers, which might be actual giant batteries - to each board. We combined and down-converted them to about 12V for the trip across the board to the local regulators. Go much lower (even if you did make the board more expensive by using thicker copper in the power planes) and you lose a lot of power heating the copper. Go much higher and you get into non-standard components in the last-stage regulators, making things pricey and limiting vendor redundancy, as well as having to start using up real-estate and thickening the (already MANY layer) boards.
The sweet spot was probably pretty coarse - I'd guess about 10V to 20V. But due to historical standards (ultimately derived from lead-acid batteries) a lot of stuff has been built for standard nominal voltages of 12, 24, and 48 volts. So 12V it was.
I suspect that, given the prevalence of 12V in existing peripherals, off-the-shelf down-converter designs, and the fact that the upgrade is incremental, Intel is giving its customers a break on converting to the new standard with minor tweaking, rather than having to revamp the designs end-to-end in a single spasm.
Also: I haven't looked at the standard yet. But it does have downconverters all over. So if it can accept raw power from about 11 to 15V and shut down gracefully to very low draw (i.e. by disabling the final regulators) when told to, it can be powered directly by a solar panel / lead-acid battery / dump-load battery controller system.
(Go a little further on the low end and add aspike arrester and maybe a supercap and a FET ideal-diode (to cover a few seconds of cranking) externally and you might use the same board in the severe environment of a (non-EV) automotive application.)
By that logic, it ony makes sense, ... (Score:2)
... to move the rest of the PSU onto the mainboard too. Probably along with the next Intel "standard" where the rails move into the CPU as a chiplet! :P
And then the chiplet can expand until it contains all peripherals and former cards. And since it will be quite small and the primary component left we could add 5G functionality and call it a "phone mainboard".
Of course with all the lock-down of monolithic integrated ... final ... "solutions" ... for the freedom question.
Am I the only one who wants MORE modu
More points of failure on the motherboard (Score:4, Insightful)
Additionally there's this: I used to work for Intel. I've seen some really shitty onboard switching regulator circuits. A shitty switching regulator can cause all sorts of strange problems. I really think they might be fucking up with this move. PC power supplies are a mature technology and the only real differencees between one brand/model and another is the attention to detail and quality of components used. Why shake this up now? Like I said above: perhaps more planned obsolescence?
Re: (Score:2)
Re: (Score:3)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:3)
its about Intel trying to force the standard away from their competitors
Errr Intel's competitors will be completely unaffected by this change as Intel's competitors as well as Intel themselves already run on common designs with their own dedicated voltage regulation all over the board. . Intel's partners will adopt it as will the industry given it's a completely open standard, and their competitors will just shrug as those same partners of Intel are also partners of the competitor and they are ultimately re
Re: (Score:2)
Re: (Score:3)
Well with your knowledge you should also understand that now all powersupply regulation fails equally. The most sensitive and stressed out components in a computer PSU are and will remain in the future the 240V to 12V conversion, done at high power. The 5V and 3.3V systems are comparatively weak, far simpler, and far more reliable. You typically never see a PSU fail in the 5V or 3.3V rails for this reason and the load on them is so low that PSUs run passively when their 12V rails aren't heavily loaded.
Compa
Re: (Score:2)
I'm having a difficult time finding statistics on causes of computer failure, but this 2016 paper gives an indication that agrees with my initial suspicions: Alas, the slashfilter rejects the URL as too long. Look for "Analyzing the Reliability of a Personal Computer System using Fault Tree Analysis" on researchgate.net.
The primary cause of computer failure is hard drive failure, with power supply failure almost as common. PS failure is 2 to 3 times more common than motherboard failure.
Also, a poor power su
Re: (Score:2)
Re: (Score:3)
WTF modded troll? Fuck off mods, -1 isn't for "wrong" or "I disagree".
For what it's worth, I think you're wrong.
In my opinion as someone who has worked in electronics for over 35 years, and who has worked with electronics longer than that, this is a move backwards.
Yes and? Thing is this doesn't materially change that. By far the most power hungry components (CPU, RAM and GPU if you have one) already use 12V for the main power and have local switching to drop down to much lower, usually variable voltages. 1
Only makes sense with buck converter modules (Score:2)
Flaw #1... (Score:3)
Re: (Score:2)
Re: (Score:2)
GROAN... Not another power supply and form factor (Score:2)
End result (Score:2)
Power supplies will be more expensive and it will be a pain to find the right one. Thanks Intel.
Makes sense for small form factors (Score:2)
This makes sense for small form factors (ITX, Intel's NUC, and similar), where there is little internal components, except maybe a GPU. Motherboards already have voltage converters for CPU and RAM, and storage in these devices usually come in the NGFF m.2 SSDs.
However for a full tower PC / Workstation with multiple full sized HDDs, usually needing specialized cards, and multiple fan controllers, 12V only would not be enough. The motherboards are already fully cramped even in eATX configurations. In that cas
PSU change for dependency (Score:2)
Redesigned - it makes the PSU unusable and useless without the mass of a co-motherboard. It avoids high voltage to escape export restrictions as a munition.
Good idea but... (Score:2)
1.) Too bad they did not go to the other well stablished standard: 48Volts. plenty of componets to handle that, inherited from telco and telco datacenter tech (and soon automotive too).
2.) I hope that Mobo vendors bring back daughter cards for power regulation functions (at least in desktops, servers, workstations and all in ones).
Patents? (Score:2)
What part of their shiny new spec will they claim a patent on, to keep AMD from using the same power supplies?
12V? (Score:5, Funny)
12v is not efficient (Score:2)
Still too many Amps through the cables, video cards need multiple parallel cables, high end cpus also. Make it 24v at least.
Re: (Score:3)
12V is perfectly fine. The reason we have so many cables is due to the choice of connector on the far end and the way we use them. Let's talk about your video cards:
Did you know you can run high end video card just fine with only one of the two connectors without any worrying temperature rise?
Did you know of the 8 parallel cables in the PCI-e power connector, 2 don't carry any current what so ever?
Let's talk about high end CPUs:
Did you know that nearly no motherboards on the market actually need more than 1
PSU's 4, motherboards : 0 (Score:3)
In the last ten years I've had 4 PSUs burn out and need to be replaced, total cost US$30-50.
In that same time I have had 0 motherboards fail, Total cost US$0. (have replaced 1 MBs for upgrades but not failures)
Time it has taken to replace dead PSU's: 15-20 minutes each (most of that spent just moving stuff on/around computer to get to it).
Time to replace a MB, ~1-2 hours including loading all the drivers that Windwos needs and possibly having to reactivate Windwos. (Thank Gods I use Linux so driver support is much easier to setup and no activation BS.)
I'll take a pass on MB's with an integrated PSU for as long as I can, much easier, cheaper, and faster to have the PSU as a separate module.
Re: Now when voltage regulators fail (Score:4, Insightful)
Yup, that's the idea behind the whole history of "improvements" to existing tech that works - kick ball into others' court.
Or as we like to say: invent a problem for the solution.
Re:Now when voltage regulators fail (Score:5, Interesting)
Every board already has plenty of voltage regulators. The ATX rails are already inadequate to cover the gamut of voltages needed in a modern system.
This would improve PSU reliability and motherboards would probably be about the same given they already have this sort of responsibility.
All the high power expansion options are 12V or 20V (USB-PD). The processor is already not running at any voltage supplied directly by the PSU and thus drives some heavy duty voltage regulation on the motherboard. Everything else that could conceivably map more directly to non-12V rails is very low power consumption and so there isn't much problem doing that regulation on the motherboard, and it opens things up for motherboard vendors to more appropriately size the voltage regulation they require than the PSU vendors could guess at. The PSU could be in a full tower with tons of slots that need 80W of standby power with a motherboard that sucks down 3.3V or it could be in a mini-itx with one slot and a power sipping motherboard. The general single-dimension 'wattage' number doesn't give them a lot of room to engineer to different load characteristics that the motherboard is in a better position to be aware of based on its design.
Re: (Score:3)
Re: (Score:2)
Okay, what about all the SSDs and hard drives that run on 5v power?
The standard HDD power cable already supplies 12V and 5V. I would guess that HDDs can step down the voltage easily if needed. And SSDs consume even less power.
Re:Now when voltage regulators fail (Score:4, Insightful)
Storage is doing no such thing for anything other than single disk systems. HDDs still reign supreme in cost per GB, SATA SSDs will be more popular than M.2 based systems for a while to come, and that completely ignores servers, workstations or other high performance systems where the desire directly to not be forced into the M.2 standard led to the development of U.2 which provides the PCI-e lanes over a cable to a device that remains powered separately.
Re: (Score:2)
Re: (Score:3)
SATA power connectors include 5v and 3.3v as well, though I don't think many devices use 3.3v, only relying on 5v and 12v that is available from molex compatible power adapters.
Re:Now when voltage regulators fail (Score:5, Informative)
Not too long ago I read about one USB drive (WD?) that used the 3.3V connector to tell if you had taken it outside of the USB enclosure to use in your computer. IIRC, it bricked the HD, or perhaps just refused to work.
OK, Googling, it seems to have been WD reds with a white label where the drive used the 3.3V line as a reset and wouldn't spin up if put in a desktop.
This mentions it about halfway down the page, https://linustechtips.com/main... [linustechtips.com] and a Reddit thread, https://www.reddit.com/r/DataH... [reddit.com]
I guess they sold the drives cheap in a USB container and didn't want you to use them in your desktop. Fix was to disable the 3.3V line.
Re:Now when voltage regulators fail (Score:4, Informative)
Not too long ago I read about one USB drive (WD?) that used the 3.3V connector to tell if you had taken it outside of the USB enclosure to use in your computer. IIRC, it bricked the HD, or perhaps just refused to work.
The 3.3v pin is part of a change in the SATA spec [westerndigital.com] to allow remote rebooting of the drive, it's not a "tamper detector", as evidence by the WD document I am referencing here literally tells you how to get around it. Newer PSUs that meet this spec will power the drive directly with no problem, and using a Molex -> SATA power adapter the drive will work fine, too.
Re: (Score:2)
Storage is doing no such thing for anything other than single disk systems.
I think he means for most consumer PCs.
HDDs still reign supreme in cost per GB, SATA SSDs will be more popular than M.2 based systems for a while to come, and that completely ignores servers, workstations or other high performance systems where the desire directly to not be forced into the M.2 standard led to the development of U.2 which provides the PCI-e lanes over a cable to a device that remains powered separately.
True but many rack servers use FlexATX and this standard seems destined for consumer desktops.
HEAT on the motherboard, not next to the P. S. fan (Score:2, Troll)
Intel insufficient management [slashdot.org].
For many years, IMO, Intel has been having the same problems as Boeing [slashdot.org]: A social breakdown inside the company that means people just go to work. There is insufficient valuable, logical interaction and cooperation.
Management doesn't have technical knowledge, and doesn't want any. Management therefore doesn't listen to technically-knowledgeable employees.
Rick Schumann has good ideas. Geoskd agrees. (Score:2, Troll)
1) "... this is a move backwards. Power supply circuitry fails more often than the electronics it supplies, and being able to swap out a bolt-in unit like a PC supply makes dealing with that point-of-failure relatively cheap and easy compared to having to replace an entire motherboard"
2) "... in the long run there'll be more motherboard failures requiring replacement that otherwise wouldn't have to happen."
3) "I used to work for Intel. I've seen some really [r
Re: (Score:2)
1) "... this is a move backwards. Power supply circuitry fails more often than the electronics it supplies, and being able to swap out a bolt-in unit like a PC supply makes dealing with that point-of-failure relatively cheap and easy compared to having to replace an entire motherboard"
This is already a silly point. The two most heavily stressed powersupply circuits will always remain the 110-240V conversion to 12V, and the second being the very heavily loaded Vcore regulation that already exists on the motherboard. This change won't see any relevant change in motherboard reliability.
What he's seen on Intel boards is completely irrelevant since most people don't use Intel reference boards.
Re: (Score:3, Informative)
Now the HEAT from voltage conversion will on the motherboard, and not next to the P. S. fan!!!
Fortunately converting from 12V to 5V or 3.3V is far more efficient than doing so from 240V. As for the fan, I'm not sure what you're concerned about. Typical well made 600-800W PSUs do not even run the fan until you start applying serious load to them. That "serious" load I'm talking about is 100% on the 12V rail as it is. Passively cooling the 5V and 3.3V regulators is so simple it is already done right next to existing 12V heatsources.
For an example of this look no further than your own motherboard, whic
List of problems taken from the PC World article. (Score:5, Informative)
1) "Because the multi-rail power supply is sending very low current to all voltage rails, efficiency is just 50 percent to 60 percent."
That is not explained, and, to me, does not seem to be correct.
2) "On the downside, you're adding features to the motherboard, which means most cost and more demand on the limited real estate of the board. And, of course, these circuits will need to be kept cool, making ventilation more of an issue."
It seems to me that "more demand on the limited real estate of the board" is a HUGE difficulty.
3) "These smaller components are also more susceptible to damage from 'bad power' "
4) "Still, our motherboard sources said, moving both the rails and the power connectors to the motherboard means a greater burden of components, a larger PCB, and more PCB layers, meaning more complexity and more cost."
5) "One vendor noted that controlling noise on the PCB could be a challenge, let alone performance. The first ATX12VO-compatible motherboard will likely be expensive as a result,
6) "One tricky part could be hooking up any SATA-powered drives, such as hard drives or 2.5-inch SSDs. Today, you'd plug those directly into the PSU. On an ATX12VO build, you'd first plug a power cable into the motherboard and then into the drive. The spec allows for up to six power connectors, but it's up to the motherboard vendor to determine how many power connectors there are. These same SATA power connectors would be used to power up your drives as well as your AIO/CLC cooler or RGB LEDs."
I've had problems with there being an insufficient number of power connectors.
My reactions:
1) The Single Rail Power Supply ATX12VO Design Guide [intel.com] says what Intel wants, but does not help people understand the underlying details, or the reason Intel wants those specifications. (68 page PDF file, July 2019, Revision 001, Document number 613768)
Insufficient explanations cause problems for Intel. Consider the damage to Intel's reputation because of the criticism in this Slashdot discussion, for example.
2) That Intel PDF file of ATX12VO specifications does not allow copying the text!
3) The web page that links to the PDF file is badly-designed. It looks okay with one browser I tried, but not others.
4) If Intel were well-managed, Intel would have used a different name. The Slashdot story mentions that ATX12VO can be confused with ATX12V0.
5) Much more could be said, but that's enough for now.
That is some of the support for my saying that Intel is insufficiently managed. There are many other issues.
Re: (Score:2)
Of course. For the manufacturesrs.
Re:Now when voltage regulators fail (Score:5, Informative)
The other rails are a throwback to the 1980s and 1990s when TTL logic ran at 5v and later 3.3v. Today's logic don't run on 3.3v* or 5v or 12v; components run on less than 2v (sometimes less than 1v) and are highly sensitive to crosstalk and power droops.
It is simply not practical to build a PSU that supplies hundreds of amps on tiny, sometimes variable, voltages, on isolated rails with shielded cables for every component. As a result, devices today treat the PSU as a sort of mains power and implement their own converters and regulators to provide each component what it needs.
Motherboards have VRMs, GPUs have VRMs, drives have VRMs... a modern PC is full of them. So why bother with an assortment of legacy PSU voltages if they're just going to get reconverted anyway?
--
(*This is a bit of a simplification; there are plenty of native 3.3v components in a modern PC but they tend to have trivial power requirements)
It's been this way for a while now. (Score:2)
You're replacing your motherboard and not your power supply. This is a win?
For more than a decade switching power supplies have been SO good that:
- you lose less power running a moderate voltage (say 12V) across the board's power planes and down-converting it to the desired voltage (3.2V, 1.8V or whatever) near the big-load chips
- than you do running the low voltage - at correspondingly higher current - through the copper planes of the board.
It's similar to a scaled-down version of high-volt
Re: (Score:2)
Intel makes motherboards, so for them, yes. For you, No.
Re: Given it’s Intel (Score:2)
Plenty of power supplies come with microprocessors and even ARM chips. There is no way to communicate with them though.
Re: (Score:2)
Luckily BMCs are notoriously brilliantly engineered, not known for security flaws; and definitely don't have terrifyingly comprehensive access to the system at a very low level; so I'm sure this is fine.
Re: (Score:2)
This is intel, they'll surely find some way to cut corners that exposes user data through the microcontroller on the PS.
Re: (Score:3)
One thing is for sure, it's probably part of their plan to get to 6 Ghz.
They need to get more power to the chip somehow.
And probably some heat pipes under the IHS and using the sidepane of the computer with a hugeass heatsink/cooler