New GPU Power Connector Eliminates Cables, Delivers More Than 600W

Asus motherboard and RTX 4070 With Hidden Power Connector
(Image credit: Asus)

@momomo_us on X (Twitter) has unveiled official documentation on a new customized HPCE power connection standard aimed at delivering supplementary GPU power through the motherboard. The documentation reveals all the key specifications of the new connector, including its ability to deliver more than 600W of power, which is even more power than what the new 12VHPWR connector provides.

According to the documentation, the HPCE power connector behaves very similarly to a PCI Express connector. There are four groups of pins in total, with 16 pins dedicated to power delivery and another twelve dedicated to signaling or communication between the mainboard and the graphics card. The connector itself is roughly the size of a PCIe x1 connector.

When utilized on a motherboard, the HPCE power connector resides in line with the primary PCIe x16 slot and sits behind the x16 slot where the motherboard chipset heatsink usually resides. This allows the graphics card to be slotted into both the x16 connector and HPCE connector, making the installation process less complicated. The official documentation recommends not to install a locking mechanism on the HPCE connector itself, since the x16 slot's built-in locking mechanism is already adequate enough to hold the graphics card in place.

This new connector is stated to be capable of delivering over 600W to the graphics card. The documentation does not state an upper limit, but it stresses the necessity to deliver more than 600W of power to support future graphics cards. 600W is a lot of power, but current generation RTX 4090s are already capable of surpassing that mark with modified firmware. RTX 4090's designed for LN2 overclocking (like the HoF cards) are also capable of surpassing 600W, and as a result, come with dual 12VHPWR power connectors.

We first saw this connector at Computex, where several prototype motherboards and graphics cards were showcased featuring the hidden power connector. The goal of the new connector is to improve cable management and aesthetics by removing all supplementary power cables from the graphics card and providing power instead via the motherboard. This way the graphics card looks super clean without any power cables getting in the way, and cable management is simplified behind the motherboard tray.

An interesting detail about this new connector is that it is not an original design. Instead, it is a modification of the High Power Card Edge (HPCE) standard that is used heavily in the server industry. In theory, this means this new consumer-focused HPCE connector won't succumb to similar reliability issues as the new 12VHPWR connector, since the connector has already been field-tested in the server space.

We'll have to wait and see what companies, if any, adopt the standard for consumer devices. So far, everything has come via Asus, so it could end up being limited to its motherboards and future GPUs.

Aaron Klotz
Contributing Writer

Aaron Klotz is a contributing writer for Tom’s Hardware, covering news related to computer hardware such as CPUs, and graphics cards.

  • rluker5
    Maybe I'm just old fashioned in that I like copper lines that deliver a lot of power to have decent cross sectional area, but I don't like it. Will we have to plug an extra 4x 8 pins into the mobo and then have all of that EM pollution all up in there next to our ram traces?

    I'm probably just old fashioned in my thinking this will just make more problems. Just give me fat wires and big connectors.
    Reply
  • -Fran-
    So... It'll get the extra juice from the 24 pin? the EPS 8pin? Still requires plugging the PCIe 8pins or the new 12pin to the motherboard still?

    I would rather Intel get a new version of ATX and work with PCI-sig to get this on their documents so it's a proper standard instead.

    Personally, I don't mind the cables, but I can see why aesthetic enthusiasts would want this and all in all, I don't think it's a bad idea... Problem I'd have is the way they go about it.

    Regards.
    Reply
  • TJ Hooker
    Admin said:
    The specifications for a new HPCE GPU power connector have been unveiled, which allows a high-power graphics card to skip the cables and pull all the power from the motherboard.
    In theory, this means this new consumer-focused HPCE connector won't succumb to similar reliability issues as the new 12VHPWR connector, since the connector has already been field-tested in the server space.
    According to the images in the linked tweets, this design still involves the 12V-2X6 connector* (the updated 12VHPWR connector, tweaked so that the signal pins are set a bit further back), it just gets plugged into the motherboard rather than the card.

    So it doesn't eliminate any cables. Nor does it do anything to avoid the "reliability issues" of the 12VHPWR connector, other than what is already being accomplished by the transition to the (backwards compatible) 12V-2X6 connector. It would arguably hurt reliability, by introducing another point of failure. Except maybe if the motherboard header happened to be situated in such a way as to allow a better/less stressed connection compared to the graphics card power header, depending on your particular case layout.

    If the HPCE connector really is more robust, I guess there could also be advantage in using that where you're more likely to have more mating cycles (I'm assuming graphics cards get replaced/reconnected more often that motherboards), but other than that it does really seem like it's largest benefit is cable management/aesthetics.

    * Or 12V8Pin, presumably for cases where less power is needed.
    Reply
  • Kamen Rider Blade
    -Fran- said:
    Personally, I don't mind the cables, but I can see why aesthetic enthusiasts would want this and all in all, I don't think it's a bad idea... Problem I'd have is the way they go about it.
    You really don't want the "Aesthetic Enthusiasts" determine ANYTHING about PC Hardware.

    We're already inundated with RGB Rainbow-Puke on everything that serves no real functional purpose other than creating a gaudy Rainbow light show.
    Reply
  • -Fran-
    Kamen Rider Blade said:
    You really don't want the "Aesthetic Enthusiasts" determine ANYTHING about PC Hardware.

    We're already inundated with RGB Rainbow-Puke on everything that serves no real functional purpose other than creating a gaudy Rainbow light show.
    RGB can be turned off and having less cables is not a bad thing, overall. As I said, it'll depend on how they go about it.

    Regards.
    Reply
  • Evildead_666
    If it works for servers, why not ?
    Morherboards might need extra, or thicker layers, so it might add cost here and there.
    Motherboard, and likely gpu too
    Reply
  • hotaru251
    rluker5 said:
    Just give me fat wires and big connectors.
    and if they dont liek the look? do what EVGA did in past where they had a adapter that moved plug to a different location.
    Reply
  • tenow
    Awesome, but have they seen a Mini-ITX motherboard?
    Reply
  • InvalidError
    -Fran- said:
    I would rather Intel get a new version of ATX and work with PCI-sig to get this on their documents so it's a proper standard instead.
    Intel kind of missed the mark on that by giving its 12VO spec insufficient power on the main connector to ditch the EPS12 cable and continuing with 12V. It think it needed at least two extra 12V pins.

    I get a feeling there will be power-efficiency pressure for 24VO PCs 10 years from now, maybe even 48VO if GaN or something beyond that manages to close the cost-efficiency gap with MOS. We'll definitely need new standards designed around native DC power distribution voltage for efficiency reasons once that happens.
    Reply
  • Alvar "Miles" Udell
    It's something that should have been done ages ago, and especially when mutli-GPU consumer setups bit the dust, but the problem is cost and reliability. In server setups it's not an issue as they're built to a standard, consumer boards though, especially lower end ones and ones from certain OEMs, would you really trust, say, an ASRock A730 Mini-ITX to be built to handle a 600w GPU? I wouldn't.
    Reply