• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

Intel Introduces New ATX PSU Specifications

Hi,
Wonder if msi installed one of their aio's that produces goop inside
Not sure I'd try a new psu of theirs after that plus them scalping their own gpu's during mining spike.

Afterburner is all msi is good for.
Even that has been troublesome, to me they are in hot water like Gigabyte
 
It was never abandoned. It's just really an OEM-only thing.
Of course when I say it I mean "abandoned" by most of PC gamers and PC mainboards.
 
Good timing. I was thinking of getting a new psu now just have to wait for some reviews.
Don't hold your breath anytime soon. The specifications are very much for the future. For one thing, this version has a INSANE transient power delivery requirements of 300% TDP <100μs, and assigns a max 600W PCIe AIC card power...Basically calling out NVidia for their stupid high power requirement and transient burst.
The new requirements are more than stringent, straight up craziness in many areas. I don't think even AX1600i can fulfill the new spec... it will take many years before an ATX3.0 PSU becomes affordable.

For those interested, I didn't find pdf file for it.
there is an analysis of the specification in Chinese by fcpowerup: https://www.fcpowerup.com/intel-atx-3-0/
igorslab must have something on this as well.
 
Sometimes you just have to "force" change.
A lot has changed in the cable and connector world for current transfer and connectivity. The problem with the current SATA power connectors? many of em. For once, you are not even using its voltage range by pinout anymore. This connector remains mostly to supply 12V today with 5V optional for the drives that need it (mechanical mostly).

I do agree this connector sucks btw, as much as I think the very old 4PIN one do. Im personally a fan of Molex' new Micro-fit 3.0 series, which can carry a lot of amperage and use small connectors. While im not a huge fan of the fact Molex will probably make money off of selling and lisencing it, any improvement is welcomed

ATX and MATX may not see too much from this, but MITX and some custom small form factor specific boards might really, really do.
How do you force change in a non cohesive industry? What if the motherboard makers and PSU makers say no?
The cost, without any direct user benefit (kind of like RGB LEDs) woll go up when it comes to motherboards and the PSU companies will have to offer cheaper PSUs, as they can't charge the same for a 12VO PSU, since it lacks about a quarter of the parts. Admittedly the 3.3 and 5V conversion parts are pretty much just a small PCB soldered to the power supplies these days, but still. This doesn't even take the case manufacturers into account and they've invested in their tooling that should last for a certain amount of units.

Maybe you don't remember that Intel tried BTX and it failed miserably, largely due to it being designed for OEM/ODM only and the fact that the CPU coolers weighed in at around a 1kg each at the time.

With regards to the SATA connector, you got that backwards, most spinning rust needs 12V whereas most SSDs only need 5V. The 3.3V addition is something of a mystery, as almost nothing used it.

Oh, the Molex power connector for drives was shite, but it was carried over from the AT standard and maybe even longer than that, but that would be before my time.

Some mini-ITX boards have been 12VO for years, but I have never seen matching PSU's, as they used a 4-pin connector, but also often an external connector that was wide range input.

It's not just about the space though, current connectors are wasteful for no good reason. They don't offer or need to offer a meaningful increase in current capacity, are bulky and require a lot of materials that could easily be reduced, not only without loosing functionality but actually increasing it.
But it is an industry wide standard that is going to take a lot of time and convincing to change and that's the core problem here.
It's very hard to make everyone agree on standards, even more so when it would require so many companies to change at a single point in time.
Motherboards should really have been improved a long time ago, but alas here we are using a standard from 1995...

27 years and the graphics card is still upside down. :p:D This industry may be hard to change, but atleast Intel is making an effort, the same cannot be said of others since it would 'cost too much'. If there was to be large changes, there cant be backward compatibility, otherwise that reliance will not let the industry move forward. Adoption to new standards and designs wont pay off immediately and manufacturers will scream 'we profits now dammit!" and remain stuck with legacy and left in the dust.
It's not that efforts aren't being made, but the hard thing is to get an industry wide consensus.

Imagine if we change from ATX to say CTX (since BTX is taken), how many years do you recon it'll take for that to become the industry norm?
How many companies are there out there that want to have a say? In the bad old days, Intel could force new standards, which isn't remotely as easy for them today.
Just to get the major components manufacturers and OEM/ODMs to agree would take years. Just look at Dell, they can't even follow the standards that are out there.
There are going to have to be some serious compelling reasons to move to new standards.

Just to be clear, I'm not against change here, but it's going to take a serious amount of consolidated effort to make such a change and from what I can tell, there's no industry wide will to push for such a change and until that happens, we're stuff with an old, but still workable standard.

Personally I believe the shift will come when PCB traces have to change, due to high-speed interfaces no longer being able to use the ATX form factor. That might very well happen around PCIe 6.0 or whatever comes next after that.
 
Last edited:
$400 to $800 power supply?
Extremely doubtful they will cost that much.

Shouldn't even have to use them but cool
Things move forward. You can alternatively just use what you have. It should still work. Or buy a new PSU that can deliver 600W on a single connector. There really isn't much options for an inbetween.

Of course when I say it I mean "abandoned" by most of PC gamers and PC mainboards.
Most mainboards by far are OEM, but I do understand what you are trying to say. It is indeed very close to abandoned in the custom PC space.
 
It's not that efforts aren't being made, but the hard thing is to get an industry wide consensus.
I will doubt about the efforts, but there has to products, not just R&D still on paper, for the industry to even begin to make a wide consensus. Intels first PSU standard was announced 2 -3 years ago? And we're just now seeing a possible adopted change with one connector, maybe two? Any kind of jump to new forms and standards will have to a collaborative effort, This new PSU standard just might be that push to get the ball rolling. I dont expect much advancement for atleast another 3-5 years, and even then, its a foot note when it comes to priority.

Laptops wont change, thats a dead horse, no expectations there.

Things move forward. You can alternatively just use what you have. It should still work. Or buy a new PSU that can deliver 600W on a single connector. There really isn't much options for an inbetween.
I expect there will be cable adapters to the new connector.
 
I will doubt about the efforts, but there has to products, not just R&D still on paper, for the industry to even begin to make a wide consensus. Intels first PSU standard was announced 2 -3 years ago? And we're just now seeing a possible adopted change with one connector, maybe two? Any kind of jump to new forms and standards will have to a collaborative effort, This new PSU standard just might be that push to get the ball rolling. I dont expect much advancement for atleast another 3-5 years, and even then, its a foot note when it comes to priority.
Another issue people also forget about is patents. Just look at PSUs, there's a reason why not everyone can use a certain size fans in them. Believe it or not, but 135 mm fans are patented, can't remember by which company right now, but it's just plain daft.
A lot of other things are "owned" by someone that wants to see royalty payments that their competitors aren't willing to abide with and so the industry is at a standstill...
True innovation died a long time ago.
 
Don't hold your breath anytime soon. The specifications are very much for the future. For one thing, this version has a INSANE transient power delivery requirements of 300% TDP <100μs, and assigns a max 600W PCIe AIC card power...Basically calling out NVidia for their stupid high power requirement and transient burst.
The new requirements are more than stringent, straight up craziness in many areas. I don't think even AX1600i can fulfill the new spec... it will take many years before an ATX3.0 PSU becomes affordable.

For those interested, I didn't find pdf file for it.
there is an analysis of the specification in Chinese by fcpowerup: https://www.fcpowerup.com/intel-atx-3-0/
igorslab must have something on this as well.
See the vid card in my sig? I’m a patient man ;)
 
Sometimes you just have to "force" change.
A lot has changed in the cable and connector world for current transfer and connectivity. The problem with the current SATA power connectors? many of em. For once, you are not even using its voltage range by pinout anymore. This connector remains mostly to supply 12V today with 5V optional for the drives that need it (mechanical mostly).

I do agree this connector sucks btw, as much as I think the very old 4PIN one do. Im personally a fan of Molex' new Micro-fit 3.0 series, which can carry a lot of amperage and use small connectors. While im not a huge fan of the fact Molex will probably make money off of selling and lisencing it, any improvement is welcomed

ATX and MATX may not see too much from this, but MITX and some custom small form factor specific boards might really, really do.
Assuming I googled the right cable, that cable looks horrible compared to existing SATA power connectors, looks way too bulky and reminds me of the cables I have to exert a lot of effort to connect or disconnect to/from motherboard, 24 pins and 8 pins, over the years those cables have given me cuts, and strained muscles with how tight they are, also SATA power connectors I have found reasonably easy to use and do fit snugly whilst been easy to push in and out (never had one come loose etc. during decades), SATA data cables on the other hand....

Ok just got myself up to speed on how SATA power will work, looks a complete nightmare and a mess to be honest.

Obvious questions.

1 - What happens if you expand SATA connectivity via a PCIe SATA add-on card which has no power ports on it?
2 - Can PCIe add-on cards have power ports on them?
3 - Can power splitters be used like today so e.g. make one power port power up 6 drives?
4 - What is the actual benefit of this system other than to make the 24 pin connector smaller?

I expect this will kill the availability of 8 SATA port boards as now each SATA port is more expensive to implement plus boards become more expensive again as we now have to pay for the power infrastructure every time we buy a motherboard instead of buying a PSU once a decade.
 
Last edited:
Assuming I googled the right cable, that cable looks horrible compared to existing SATA power connectors, looks way too bulky and reminds me of the cables I have to exert a lot of effort to connect or disconnect to/from motherboard, 24 pins and 8 pins, over the years those cables have given me cuts, and strained muscles with how tight they are, also SATA power connectors I have found reasonably easy to use and do fit snugly whilst been easy to push in and out (never had one come loose etc. during decades), SATA data cables on the other hand....
Microfit 3.0 are very small, about half of current minifit size in volume. Similar current capability. They have 18AWG crimp options, as well as 20-24 and 26-30. quite versatile. I've been working with this series for about 3 years now and have been enjoying their integration process at what I do.

1 - What happens if you expand SATA connectivity via a PCIe SATA add-on card which has no power ports on it?
You have to provide power to 2.5" SATA drives. Its simply a part of their spec. Their data goes through the designated SATA cable.

2 - Can PCIe add-on cards have power ports on them?
Technically yes, but practically you don't want to chain power for these many connections. Its already AC to DC12, then to motherboard, then to card, then to drives. Your slot PCIe power is already quite limited in previous generation ATX standards.

3 - Can power splitters be used like today so e.g. make one power port power up 6 drives?
It can, but the current balance laws dictate that the root will be the hottest part, and the root has to carry the current equal to the amount of all its clients. You do not want to chain sensitive NAND containing drives in a serial power tree. Its a recipe for disaster.

4 - What is the actual benefit of this system other than to make the 24 pin connector smaller?
Size is most of what plays part here, reduction of plastic use. Other than that - not too much externally. Internally some improvements can be done with how you deliver power, but those were done partially without this new spec. The new standard makes a lot more sense in the way 12V+ and 12V- works. Current 8PIN PCIe is quite bizarre in the way it works, with the 2 additional pins being 12V- or GND if we wanna be PC about it. That won't magically make the 12V+ cables able to carry higher currents. So a better use is to do 4+4, that makes your effective current capability higher. Its how EPS 8PIN for CPUs work, and with its current blade-like contacts at the end of cables is already able to handle a lot more current than what people think it can. quality PSU and modern motherboard can easily use 10A per pin on this cable per polarity, that means 480W if you want to keep it safe.

Being able to move from 2 or even 3 large connectors on a card is good news. If a single one that's taking a third of the space is capable of giving you 6 18AWG wires for each polarity, that means you can do a lot with it, and use the same connector for pretty much all external-power cards. no messing about with different cable amount.
 
Last edited:
Yes, INTEL introduces something new which does not cancel, previews ATX specifications.
Within 2022, there is active PSU production of ATX 2.4 and of ATX 2.53 and the market will not change any sooner than 2032.
Yes PSU development this is a Giant this running at turtle max speed, and this is great as warranty to backwards compatibility.
 
Technically yes, but practically you don't want to chain power for these many connections. Its already AC to DC12, then to motherboard, then to card, then to drives. Your slot PCIe power is already quite limited in previous generation ATX standards.

75W is not that limited, for 3.5'' HDDs it won't power a lot of them but 2.5'' or SSDs would be fine.

It can, but the current balance laws dictate that the root will be the hottest part, and the root has to carry the current equal to the amount of all its clients. You do not want to chain sensitive NAND containing drives in a serial power tree. Its a recipe for disaster.

By serial you mean a long cable with several power connectors right? That's parallel and what's already done with current power supplies.

Size is most of what plays part here, reduction of plastic use. Other than that - not too much externally. Internally some improvements can be done with how you deliver power, but those were done partially without this new spec. The new standard makes a lot more sense in the way 12V+ and 12V- works. Current 8PIN PCIe is quite bizarre in the way it works, with the 2 additional pins being 12V- or GND if we wanna be PC about it. That won't magically make the 12V+ cables able to carry higher currents. So a better use is to do 4+4, that makes your effective current capability higher. Its how EPS 8PIN for CPUs work, and with its current blade-like contacts at the end of cables is already able to handle a lot more current than what people think it can. quality PSU and modern motherboard can easily use 10A per pin on this cable per polarity, that means 480W if you want to keep it safe.

Being able to move from 2 or even 3 large connectors on a card is good news. If a single one that's taking a third of the space is capable of giving you 6 18AWG wires for each polarity, that means you can do a lot with it, and use the same connector for pretty much all external-power cards. no messing about with different cable amount.

Hmm the 2 extra pins in PCIe power connectors are not simply 12V or GND, they're sense pins to "detect" what's connected to them (if you plug a 6 pins to an 8 pin connector the device will know because it will see the sense 4 is missing for example). PCIe made the standard power connectors 75w/150w for 6/8 pins, not really that the connectors couldn't carry a lot more
 
Hmm the 2 extra pins in PCIe power connectors are not simply 12V or GND, they're sense pins to "detect" what's connected to them (if you plug a 6 pins to an 8 pin connector the device will know because it will see the sense 4 is missing for example). PCIe made the standard power connectors 75w/150w for 6/8 pins, not really that the connectors couldn't carry a lot more
its the same as before but a different twist for higher power; it seems 8(6+2) might be going the way of the berg connector (and others of the past):

Capture2.PNG

The new 12VHPWR connector has 4 sideband signals defined that communicate between the Power Supply and the PCIe* Card. Two of these sideband signals are required with the connector and two of them are optional from the Power Supply.
The four sideband signals are:
• SENSE0
• SENSE1
• CARD_PWR_STABLE
• CARD_CBL_PRES#
For the most up to date and detailed description, refer to Section 5.3 – Optional Sideband Signal in the PCI Express* Card Electromechanical Specification, Revision 5.0. That document is available from www.pcisig.com.

Sense1 / Sense0 (Required)
These two sideband signals provide important information from the PSU to the Add-in Card and are therefore required from the power supply. These signals allow the PSU to tell the Add-in Card how much power the Add-in Card can use during both Initial Power Up and Maximum Permitted Power. In the PCI CEM Gen 5 spec these levels are called “Connector Initial Permitted Power” and “Connector Maximum Permitted Power” because they refer to the power limits that can be consumed thorough the Auxiliary Power Connector. The Maximum Permitted Power is the maximum supported by this auxiliary power connector alone. It does not correspond to power consumed through motherboard card edge connector.

The PSU providing the 12VHPWR auxiliary power connector must short the appropriate SENSE signals to ground or leave them open (high impedance) to indicate the power limits associated with the power supply. These SENSE signals must not change state while the Add-in Card is operational. Support of the SENSE0/SENSE1 sideband signals is required for a PSU.

Support for the SENSE0/SENSE1 sideband signals does not rely on any of the other optional sideband signals defined for the 12VHPWR connector.

pretty much the same as before. but then it gets really crazy:

CARD_PWR_STABLE (Optional)
This optional sideband signal functions as a “Power Good” indicator from the Add-in Card to the PSU. When this signal is asserted, the Add-in Card is indicating that local power rails on the Add-in Card are within their operating limits.This signal can provide a fault detection from the Add-in Cards to the PSU, which can allow the PSU a protection opportunity.

If this signal is used by the PSU, the signal will come from the Add-in Card as an open collector / open-drain fashion. In the Add-in Card, this signal will be tied to a 100kΩ pull-down resistor to ground. When implemented, the signal must be tied to a 4.7kΩ pull-up resistor to +3.3V at the power supply. When the PSU or system monitors the state of the CARD_PWR_STABLE signal it must be done with a high impedance 3.3V logic compatible device input.
and it goes on . .fortunately intel isn't as bad as pci-e sig hiding docs behind paywall.

section 3 explains alot.
 

Attachments

Microfit 3.0 are very small, about half of current minifit size in volume. Similar current capability. They have 18AWG crimp options, as well as 20-24 and 26-30. quite versatile. I've been working with this series for about 3 years now and have been enjoying their integration process at what I do.


You have to provide power to 2.5" SATA drives. Its simply a part of their spec. Their data goes through the designated SATA cable.


Technically yes, but practically you don't want to chain power for these many connections. Its already AC to DC12, then to motherboard, then to card, then to drives. Your slot PCIe power is already quite limited in previous generation ATX standards.


It can, but the current balance laws dictate that the root will be the hottest part, and the root has to carry the current equal to the amount of all its clients. You do not want to chain sensitive NAND containing drives in a serial power tree. Its a recipe for disaster.


Size is most of what plays part here, reduction of plastic use. Other than that - not too much externally. Internally some improvements can be done with how you deliver power, but those were done partially without this new spec. The new standard makes a lot more sense in the way 12V+ and 12V- works. Current 8PIN PCIe is quite bizarre in the way it works, with the 2 additional pins being 12V- or GND if we wanna be PC about it. That won't magically make the 12V+ cables able to carry higher currents. So a better use is to do 4+4, that makes your effective current capability higher. Its how EPS 8PIN for CPUs work, and with its current blade-like contacts at the end of cables is already able to handle a lot more current than what people think it can. quality PSU and modern motherboard can easily use 10A per pin on this cable per polarity, that means 480W if you want to keep it safe.

Being able to move from 2 or even 3 large connectors on a card is good news. If a single one that's taking a third of the space is capable of giving you 6 18AWG wires for each polarity, that means you can do a lot with it, and use the same connector for pretty much all external-power cards. no messing about with different cable amount.
Thanks for the answers, I perhaps should have mentioned the questions were all related to 3.5 inch spindle usage. On existing ATX. the 75 PCIE power doesnt matter for PCIE SATA add-on card as drives powered directly from PSU.
 
Hmm the 2 extra pins in PCIe power connectors are not simply 12V or GND, they're sense pins to "detect" what's connected to them (if you plug a 6 pins to an 8 pin connector the device will know because it will see the sense 4 is missing for example). PCIe made the standard power connectors 75w/150w for 6/8 pins, not really that the connectors couldn't carry a lot more
Yes, there's sense there, but common thinking is that somehow 8PIN has higher current capability than the older 6PIN pcie standard, which is of course incorrect since you still have the same amount of 12V+ wires. It is common practice to dummy-chain GND into those two from the same connector.

Personally im not a huge fan of this kind of setup. Sense could be done entirely on the card's side for each and every pin, including reverse polarity protection. So hopefully just more spots for symmetrical connectivity and higher current per connector volume.
 
Yes, there's sense there, but common thinking is that somehow 8PIN has higher current capability than the older 6PIN pcie standard, which is of course incorrect since you still have the same amount of 12V+ wires. It is common practice to dummy-chain GND into those two from the same connector.

Personally im not a huge fan of this kind of setup. Sense could be done entirely on the card's side for each and every pin, including reverse polarity protection. So hopefully just more spots for symmetrical connectivity and higher current per connector volume.

Yes but actually no. They (usually) have the same number of 12V lines but the 6 pin is supposed to only have 2 as opposed to 3 and the missing sense indicates the card can only pull 75W instead of the 150W. It's a bad and outdated spec that should have been updated a long time ago. Just simple elementary school math shows that connector/cable current limit wasn't the issue since 3*12V lines are delivering the same power as 2*12V lines

1648215618806.png

 
Back
Top