Thursday, March 3rd 2022
Intel ATX 3.0 16-pin Power Connector for PCIe Gen5 is Smart, Has Four Power-Delivery Variants
Intel is reforming the ATX specification in a big way with ATX 3.0, and power supplies shipping with new 16-pin power connectors for the next generation PCI-Express Gen 5 graphics cards. The 16-pin 12VHPWR (12-Volt High Power). The need for a new connector standard is arising from the growing power demand from graphics cards, as contemporary high-end GPUs crossing the 375 W barrier (two 8-pin PCIe inputs), and some of today's high-end cards even shipping with typical board power values of 450 W at stock frequencies. Custom-design cards based on these GPUs create cable-spaghetti in your case, as they come with three or more 8-pin connectors. To solve these problems and more, Intel innovated the 12VHPWR.
The 12VHPWR connector has 12 electrical pins and 4 side-band pins, for a total of 16 pins. The side-band pins enable low-fi communication between the power-supply and the graphics card, and two of these pins, labeled "SENSE0" and "SENSE1," let the graphics card know what kind of connector is plugged in, so it can accordingly adjust its power-management. There are four variants of the 16-pin connector based on the electrical capability offered by the PSU. These include 600 W, 450 W, 300 W, and 150 W. The arrangement of the two SENSE pins is how the graphics card tells the four apart.The 12VHPWR connector delivers power in two system states—Initial Permitted Power, and Maximum Sustain Power. The former defines a power-delivery limit when a system is powered up and awaiting boot (up to the stage of the graphics driver to get loaded). The latter is the limit allowed once the driver takes over (software configuration). The table above shows the various power configurations. Will this cause chaos? We don't think so. Any graphics card with a 16-pin connector will take in any connector, and if the connector doesn't meet its requirements, the GPU will default to the lowest power spec, and adjust its frequencies accordingly (performance will be affected). The GeForce RTX 3090 Ti Founders Edition is (or was supposed to be) the first graphics card with a 16-pin (12+4 pin) connector, including the side-band stub, even though the GPU doesn't support PCIe Gen 5.
Sources:
momomo_us, Wccftech, VideoCardz
The 12VHPWR connector has 12 electrical pins and 4 side-band pins, for a total of 16 pins. The side-band pins enable low-fi communication between the power-supply and the graphics card, and two of these pins, labeled "SENSE0" and "SENSE1," let the graphics card know what kind of connector is plugged in, so it can accordingly adjust its power-management. There are four variants of the 16-pin connector based on the electrical capability offered by the PSU. These include 600 W, 450 W, 300 W, and 150 W. The arrangement of the two SENSE pins is how the graphics card tells the four apart.The 12VHPWR connector delivers power in two system states—Initial Permitted Power, and Maximum Sustain Power. The former defines a power-delivery limit when a system is powered up and awaiting boot (up to the stage of the graphics driver to get loaded). The latter is the limit allowed once the driver takes over (software configuration). The table above shows the various power configurations. Will this cause chaos? We don't think so. Any graphics card with a 16-pin connector will take in any connector, and if the connector doesn't meet its requirements, the GPU will default to the lowest power spec, and adjust its frequencies accordingly (performance will be affected). The GeForce RTX 3090 Ti Founders Edition is (or was supposed to be) the first graphics card with a 16-pin (12+4 pin) connector, including the side-band stub, even though the GPU doesn't support PCIe Gen 5.
56 Comments on Intel ATX 3.0 16-pin Power Connector for PCIe Gen5 is Smart, Has Four Power-Delivery Variants
Other than that, this is kind of cool. But also too simple a mechanism for me to trust any custom cabling that isn't very high quality.
www.cnbc.com/2022/03/02/why-intels-ceo-was-a-guest-at-bidens-state-of-the-union-address.html
nom nom
Now, as for the subject at hand, it seems to be about that time to transition to a new cable for GPU's. Seeing as all but the lowest power cards need 2 cable connections already, it only makes sense to do away with the multiple 6-8 pin connectors and compress everything back down to 1 cable.
They will need three months to design a prototype.
Six months testing.
a) First production sample = at eight months time.
Internal testing two months.
Getting all certifications = six months.
Give or take = 14 months for the fast running teams.
This is the circular logic of business in late-stage capitalism: shareholders run at the first sign of danger, which sends stock markets plunging. Stock markets are seen as a measure of the stability of the economy, despite having near zero direct relation to the actual stability of jobs or incomes (beyond being able to run companies out of business for no reason other than fear). Governments don't want that, so they spend public money funding private profits to keep the shareholders from panicking. This makes shareholders accustomed to artificially inflated profit margins and wildly unrealistic costs of doing business, which further increases their anxiety for anything approaching an ordinarily stable business. So they lobby government for more support. And 'round and 'round it goes. Yeah, it's pretty messed up. I always put this down to essentially greed and fear in the face of competition: You have a gen 1 product that does X perf at Y watts. Then you have gen 2 that could do 1.3X perf at Y watts - but what if 30% isn't enough, or competitors deliver more? So you push power, maybe to 1.5Y, and get 1.8X perf instead. More efficient? In a way, yes. But also poorly tuned and sub-optimal, and you'd have gotten better efficiency if you weren't so dead set on pushing things as far as they would go. And for each generation as the pushing succeeds, the engineers start thinking "well, we could push it to where it is now, why not further?" And suddenly we go from the 30-40W CPUs of the early 2000s to 250+W ones today, and dozens-of-watts GPUs at the same point (or hundred-ish for what most people could get a decade ago) to 400W today and 600W tomorrow.
Heck, my 6900XT is definitely a product of this process - but then I keep it undervolted and underclocked to a far more efficient range, cutting nearly 150W off its power budget.
The government giving subsidies/tax breaks to private corporations is a neo-lib/con capitalist concept (Neo-Lib). The expectation being that these private entities are more efficient at meeting market demand than a government entity would be, and because of this, the economic impact in the long run is a net gain.
So again, the phrase you are looking for is corporate welfare, and both parties in the USA, which are full of neo-lib/cons, continue to be in favor of it.
I got some questions about this:
Why you call this an Smart power connector, as I know my VGA uses 8 pin connector capable of 150W + 75W PCI Express conection but my card and ALL consumes diferent power depending of his load so it's not fixed to 225W or other number, it varies, so I cannot understand the usefull of those new standard with 4 prefixed W.
Also I got other question ¿Is ATX 3.0 from Intel is compatible with ATX12VO Power Supplies or there are two diferent standards both from Intel for future PSU?
people should already be wary of shoddy cables. :)
Its far too easy to always point the finger elsewhere. The same goes for 'power usage' and how 'commerce makes it increase'.
We're the idiots buying everything.