Intel releases the ATX 3.0 power supply spec

P

polonyc2

Fully [H]
Joined
Oct 25, 2004
Messages
25,829
Intel has published the final ATX 3.0 power supply specification, and it’s the biggest PSU spec change in almost 20 years...PC hardware has changed a lot since then, particularly the power requirements of GPUs...the new standard formally introduces PCIe 5.0 graphics card support with a new 12-pin 12VHPWR connector that can deliver up to 600W...the fact that cards can receive up to 600W from a single connector doesn't bode well for future GPU power requirements...it's not clear if next gen GPUs can receive the full 600W from current power supplies via 8-pin to 12-pin adapters

In addition to PCIe 5.0 support, there's the new ATX12VO 2.0 spec which introduces some other features aimed at providing better reliability...ATX 3.0 PSUs will introduce a second certification standard by Cybernetics alongside the existing 80 Plus ratings...Cybernetics certifications aren’t just for efficiency...there’s also a noise level certification and a related chassis soundproofing certification that will definitely assist users that want a quiet build

ATX 3.0 PSUs will come to market throughout 2022...

https://www.pcgamer.com/intel-releases-the-atx-30-power-supply-spec/
 
I'm not going any bigger than my 750 Watt psu and backup. I suppose it just to reduce clutter for power connectors along with other reason only Johnny Guru PSU could answer.
 
Last edited:
Can you get these new PS's and install them in older 4th or 5th Gen Intel machines?
 
jordan12 said:
Can you get these new PS's and install them in older 4th or 5th Gen Intel machines?
Click to expand...
You can use an ATX 2.0 power supply on an ATX 3.0 motherboard with a power adapter, but I'm not sure if it goes the other way. Complications arise due to the new ATX12VO specification and its removal of the 3.3V and 5V rails.
 
Armenius said:
You can use an ATX 2.0 power supply on an ATX 3.0 motherboard with a power adapter, but I'm not sure if it goes the other way. Complications arise due to the new ATX12VO specification and its removal of the 3.3V and 5V rails.
Click to expand...
It seems like ATX 3.0 and ATX12VO are separate things. I can't find any info if ATX 3.0 changes any of the standard motherboard connectors or if it removes any voltage rails.
 
Armenius said:
You can use an ATX 2.0 power supply on an ATX 3.0 motherboard with a power adapter, but I'm not sure if it goes the other way. Complications arise due to the new ATX12VO specification and its removal of the 3.3V and 5V rails.
Click to expand...
I think they're orthogonal to each other. 12VO is a branch of the 2.x and 3.0 standards. ATX 3.0 adds the new 600W GPU connector, and will be available with both classic 24pin and 12VO mobo interfaces.
 
For anyone curious about the Cybenetics standards; they test at a lot more load points than 80+ so they're not directly comparable but assuming the numbers are for 115V not 230V:
A+ is higher than anything in 80+, A roughly corresponds to 80+ Titanium, B is midway between 80+ Gold and Platinum, C to 80+ silver, with D being somewhere between Bronze and Silver. If ETA is evaluated at 230 instead, shift everything up 1 level, so A+ ~= Titanium, etc

Eta (power efficiency)
Code: 
ETA Level    AVG Efficiency    PF        5VSB Efficiency    Vampire Power
A+           94% to 97%        >=0.985   >79%               <0.10W
A            91% to 94%        >=0.98    >77%               <0.15W
B            98% to 91%        >=0.97    >75%               <0.20W  [I think this should read 88, and is a typo on Tom's part -  Dan]
C            85% to 88%        >=0.96    >73%               <0.23W
D            82% to 88%        >=0.95    >71%               <0.25W



They also have a noise standard, lambda:

Code: 
LAMBDA Level     Requirements
A++              < 20 dB(A)
A+               20 dB(A) to 25 dB(A)
A                25 dB(A) to 30 dB(A)
B                30 dB(A) to 35 dB(A)
C                35 dB(A) to 40 dB(A)
D                40 dB(A) to 45 dB(A)
E                > 45 dB(A)

https://www.tomshardware.com/news/cybenetics-eta-lambda-certifications-psu,34298.html
 
Last edited:
Armenius said:
You can use an ATX 2.0 power supply on an ATX 3.0 motherboard with a power adapter, but I'm not sure if it goes the other way. Complications arise due to the new ATX12VO specification and its removal of the 3.3V and 5V rails.
Click to expand...
ATX 3.0 and ATX12VO2.0 are different things. The ATX spec is designed for the consumer space where the 12VO spec is designed for OEM’s and SI’s who have to comply with various regulations that the standard ATX spec struggles to meet. Like it will but at 2-3x the price.
 
so what are PSU recommended requirements going to be for next-gen GPU's?...I've always gone over the recommended requirements with my previous builds- currently am using an 850w Corsair PSU with my 5800X + RTX 3080
 
polonyc2 said:
the fact that cards can receive up to 600W from a single connector doesn't bode well for future GPU power requirements
Click to expand...

Rumor has it that halo 4090 is 600 watts, with halo 79XX 800+. Not stock parts, probably watercooled, OEM parts, but still. That's board power only, not cooling, which isn't going to take much, but it's going to start adding up.

I checked and didn't see anything about it, but I wonder how long it'll be before we see either multiple-power cable PSUs, or daisy-chained PSUs with a discrete GPU power supply.
 
That image with the 12-pin 12VHPWR connector scares the shit outta me. Those is some thin-ass wires.

I smell smoke
 
Nebulous said:
That image with the 12-pin 12VHPWR connector scares the shit outta me. Those is some thin-ass wires.

I smell smoke
Click to expand...

It should be good, but flat cables will have better thermals than a round bundle and is probably advisable if you're snaking the wires around instead of running directly through the case.

It's using 16AWG wire, which is good for 18A @ 90C in raceway (which an cable managed system with wiring behind the mobo tray might qualify as) or 24A in an open environment in a 30C environment. a 12 wire bundle gets a 50% derating so that's a 9/12A limit; which is in line with the ~9A limit in the spec. The 4 sense/control pins use much thinner wire, but shouldn't carry any meaningful current levels.

As bad as the cooling is on some of the OEM systems Gamers Nexus has reviewed recently has been though, I'd feel a lot more comfortable if the 600W plug used thicker 14 AWG wires or if there was monitoring for wire overheat though.

https://www.lapptannehill.com/resources/technical-information/ampacity-chart

Edit: on second thought those same hidesously bad thermals would probably for the RTX4090 down to a 200W self protection mode, at which point the wiring would also remain at a safe temperature.
 
Last edited:
Although I am all for reducing cable clutter and making things easier/less complicated etc, but can someone please tell me exactly why Intel gets to dictate the specs/designs for these new connectors & cables ect.... shouldn't some governing body be involved too, like the PCIe-sig and/or IEEE ???
 
bonehead123 said:
Although I am all for reducing cable clutter and making things easier/less complicated etc, but can someone please tell me exactly why Intel gets to dictate the specs/designs for these new connectors & cables ect.... shouldn't some governing body be involved too, like the PCIe-sig and/or IEEE ???
Click to expand...
Intel worked with other companies involved, but at the end of the day ATX is an Intel spec dating back to the Pentium era in the mid 90s; so when new versions come out they still have Intel's name on top.
 
bonehead123 said:
Although I am all for reducing cable clutter and making things easier/less complicated etc, but can someone please tell me exactly why Intel gets to dictate the specs/designs for these new connectors & cables ect.... shouldn't some governing body be involved too, like the PCIe-sig and/or IEEE ???
Click to expand...
Mostly because there’s no money in it and there’s no demand to do so. Intel does a fair amount of work on it maintaining compatibility and keeping current with laws and regulations, all while providing features that manufactures and customers want.

Intel invented it and I’m sure if a known standards body came along and wanted to take it from them they might turn it over but nobody’s wanted too and there isn’t really a benefit in doing so.
 
bonehead123 said:
Although I am all for reducing cable clutter and making things easier/less complicated etc, but can someone please tell me exactly why Intel gets to dictate the specs/designs for these new connectors & cables ect.... shouldn't some governing body be involved too, like the PCIe-sig and/or IEEE ???
Click to expand...
ATX was created by Intel.
 
Lakados said:
The BTX spec actually proposed some good things and even now I think it deserves a revisit.
BTX paired with 12VO 2.0 could do some excellent things with thermals and case designs.
Click to expand...

While the return of high TDP CPUs does reverse one of the trends that stood in the way of BTX adoption; in other ways it's become badly dated.

Merging most of the old north bridge functions into the CPU and the remainder with the southbridge functions in a 1 piece chipset means that the component layout optimizations are no longer valid.

Replacing the conventional CPU cooler with a ducted 120mm front panel fan probably is an upgrade over the basic top down coolers that Intel and AMD bundle; but it's also a setup that only really works with full width cases not slim ones that OEMs and some mITX fans like. It's also mostly irrelevenant on the enthusiast side where 140mm towers and water blocks rein. Requiring good intake airflow from the front would be an issue with a lot of newer case designs; I'm fine with putting pressure on designers here but it would be an issue.

The front mounted cooler also defacto requires the mobo to run the entire front to back length of the case; with the result that picoBTX (closest equivalent to mITX) mobos are just as long as full size BTX and significnatly larger than mITX as a result. From the other direction there's no natural way to design a case to support both normal BTX boards an an extra long variant that would be the equivalent to eATX.

Putting the memory slots above the CPU instead of in front of it also effectively precludes layouts with more than 4 of them; which is an issue for high end CPUs these days.

In theory there probably is value in trying to design a new modern layout to replace ATX. But Intel only made incremental changes with ATX 12VO because there wasn't interest/support from PC OEMs for more radical changes. Instead we just got a standardized version of what they were already doing as proprietary designs anyway.
 
DanNeely said:
While the return of high TDP CPUs does reverse one of the trends that stood in the way of BTX adoption; in other ways it's become badly dated.

Merging most of the old north bridge functions into the CPU and the remainder with the southbridge functions in a 1 piece chipset means that the component layout optimizations are no longer valid.

Replacing the conventional CPU cooler with a ducted 120mm front panel fan probably is an upgrade over the basic top down coolers that Intel and AMD bundle; but it's also a setup that only really works with full width cases not slim ones that OEMs and some mITX fans like. It's also mostly irrelevenant on the enthusiast side where 140mm towers and water blocks rein. Requiring good intake airflow from the front would be an issue with a lot of newer case designs; I'm fine with putting pressure on designers here but it would be an issue.

The front mounted cooler also defacto requires the mobo to run the entire front to back length of the case; with the result that picoBTX (closest equivalent to mITX) mobos are just as long as full size BTX and significnatly larger than mITX as a result. From the other direction there's no natural way to design a case to support both normal BTX boards an an extra long variant that would be the equivalent to eATX.

Putting the memory slots above the CPU instead of in front of it also effectively precludes layouts with more than 4 of them; which is an issue for high end CPUs these days.

In theory there probably is value in trying to design a new modern layout to replace ATX. But Intel only made incremental changes with ATX 12VO because there wasn't interest/support from PC OEMs for more radical changes. Instead we just got a standardized version of what they were already doing as proprietary designs anyway.
Click to expand...
Well when you put it like that yeah…. But I do think with the way CPU and more specifically GPU’s are trending some changes to the format may be required sooner than not.
 
Lakados said:
Well when you put it like that yeah…. But I do think with the way CPU and more specifically GPU’s are trending some changes to the format may be required sooner than not.
Click to expand...
The last 15+ years have demonstrated no one cares enough about desktops anymore to do anything beyond the bare minimum needed to keep things shuffling along. The failure of BTX and the rise of laptops for all sucked the life out of the consumer/enthusiast platform.

Even little things that could've been done easily in a forward/backward compatable manor with a cheap plug adapter like downsizing the 24 pin ATX connector never happened. It could've dropped to 12/14 pins in the mid/late 2000's after the CPU and other mobo components all started running off of 12V, or to 10/12 pins in the early/mid 2010's as PCIe displaced PCI and most expansion cards began running on 12V as well removing the last major consumer of 3.3V power. Instead nothing until a half dozenish years after OEMs started rolling their own 12V only designs was the 12VO standard created; and it's strugging to overcome a chicken and egg problem in the DIY/System builder space.
 
You must log in or register to reply here.
Back
Top