Friday, September 19th 2025
ASUS Proposes Reinforced PCIe Slot to Deliver 250 W of Power From the Motherboard
ASUS unveiled a prototype this week with a new method for graphics cards to draw power from the motherboard's PCIe expansion slot, proposing a solution that could eliminate the need for card-mounted power cables for many entry-level and mid-range GPUs. The concept focuses on beefing up the physical PCIe power path: ASUS engineers widened and thickened the contact pads at the front edge of the card's PCIe interface, enlarged the copper area inside the corresponding slot, and fused several 12-volt lanes into larger, lower-resistance conductors. To share the load, the traditional 8-pin connector is relocated onto the motherboard, allowing the slot and onboard connector to work together. Higher-conductivity materials and a larger terminal cross-section are used to increase the continuous current capacity without modifying the card's external form factor.
This idea could be a good match for GPUs in the 180-220 W range, where customers might appreciate a cable-free card. Examples could include RTX 5060 Ti-class parts or comparable AMD models. The company has not released full schematics, continuous current ratings, thermal validation, or long-term endurance data, and the headline 250 W claim leaves some questions unresolved. A conventional slot plus a motherboard auxiliary connector typically totals 225 W, where the 8-pin supplies 150 W and the PCIe connector gives 75 W. ASUS is already known for experimenting with alternative power delivery, with an example being the Graphics Card High-Power (GC-HPWR) connector. This is a higher-power design, rivaling NVIDIA's 12V-2x6V connector.
Sources:
IT Home, via ComputerBase
This idea could be a good match for GPUs in the 180-220 W range, where customers might appreciate a cable-free card. Examples could include RTX 5060 Ti-class parts or comparable AMD models. The company has not released full schematics, continuous current ratings, thermal validation, or long-term endurance data, and the headline 250 W claim leaves some questions unresolved. A conventional slot plus a motherboard auxiliary connector typically totals 225 W, where the 8-pin supplies 150 W and the PCIe connector gives 75 W. ASUS is already known for experimenting with alternative power delivery, with an example being the Graphics Card High-Power (GC-HPWR) connector. This is a higher-power design, rivaling NVIDIA's 12V-2x6V connector.
95 Comments on ASUS Proposes Reinforced PCIe Slot to Deliver 250 W of Power From the Motherboard
That'd be about as 'simple' of a solution as merely increasing voltage on the aux power input. I mean... 24VDC isn't uncommon, and 48VDC is quite common in Enterprise and Telecomm :laugh:
Actually, I think it'd be 'simpler' to up the DC voltage. The onboard power phases on a given card would just need to be 'wide range' for backwards compatibility.
First 'transitional' generation(s) would have implicitly monsterous power phases to support 50+Amps @ 12VDC but, once well-adopted could be as little as 12.5Amps@48VDC.
Dell can use this on prebuilt systems to lower costs while making replacing the GPU with a standard card more difficult, and Nvidia can increase the power consumption of their cards to 800W.
And yeah, I recall there being issues with PSU's being tripped from too much noise on PCI-e slot power.
Especially while trying to keep backwards compatability with stuff.
Which means 200W would probably cover 50% or more of all gamers gpu users.
Asus is not investing money on these experiments for fun, they do it because they know it can potentially be attractive to a majority of their buyer's base.
Not everybody has a 600W 5090.
GPUs can already get enough power and cleaner of disturbances/noise straight from the power supply.
I know it's old and dated now, but have a look at how the cylinder Mac Pro delivered power to its 2 GPUs, Apple used 2 wide, flat copper rails that were held to the GPU boards by screws at the power contact point. Very secure, with big enough conductors to handle the load. Considering we already have our screwdrivers out in order to install our GPUs, I wonder why something like that hasn't become a consideration over the plastic clip?
Bend connectors on asus mainboard articles for mainboard experiments speak a clear language for myself. Back to the future = BTF had too many bend FAN header pins. The lack of capability to not bend them back on a press booth or having at least 20 spare boards lying around for press pictures. Srcape PEG sockets which ruin the core functionality of plug in cards. A few examples for the lack of expertise in the mainboard segment. a 2000€ mainboard maybe feels like luxuary. I talk about the press pictures, tech press articles and generic impressions of the equivalent of hte hp microtowers and asus prime mainboards.
Most people remember old PCI, but in servers you had PCI-X everywhere. In this case it was to support 32/64b, but there is nothing which stands in the way to add extra powerlines inside PCI-e spec/slot at the very end.
It just requires the will for a change. Let's say from 2030 PCI-SIG will put HP-PCIe (HighPower) standard into practice. No need to worry for old GPUs as standard will be backward compatible for older hardware. Yes, SSI-EEB will be the new ATX in 2 years at the top end.
It's not really feasible long term to deliver a lot of power for future GPUs, without more power coming from the motherboard through slots. CPUs are already there - bigger and bigger and enormous sockets are purely for power delivery to handle 1kW+ of power. 75W through PCIe was fine with GTS8800 in 2008 or GTX580 in 2010. Time to admit we're at the wall what current PCIe slot can do for modern GPUs.
Oh and when they are at the prototyping stage let them fix 12HPWR-whatever power delivery standard to be actually safe. It's ridiculous that nGreedia doesn't allow partners to fix obvious faults with 12HPWR connector by splitting the load, proper fusing and current balancing.
Engineering management is near extinct.
Nowadays most corporations are run purely by MBA/marketing scammers and other psychopaths.
And this proposition is just another thing out of their anus just like that grossly overspecced 12VHPWR/what ever the name of the day for turd.Nothing simple in that.
It would need rerouting lots of stuff on mobo and where would you put things displaced by it?
Mobos are already running out of real estate courtesy of misfit for desk PC M.2 slots originally designed for slim, limited expandability mobile devices.
Simplest thing is delivering power directly as close as possible to where it's consumed.
That has also the least things which can be failure points.
Servers are again entirely different animal because their hardware is purposely designed for specific needs/functionality.
They go up to 1800W to the best of my knowledge, maybe more? As you can imagine though, it needs a lot of copper in a nice, thick, expensive PCB trace - so not something you're going to see on cheap motherboards.Maybe it does in a few scenarios but most GPUs recess the power connector, and if there's no room to fit the power connector, I've been using things like this for at least a decade:
There are plenty of solutions to this problem already, and Asus' idea is another solution that seems to have just as many downsides as it does benefits. For an ultra-cramped mITX motherboard, it means you've got to sacrifice precious real estate on the board for some space-inefficient fat power traces. No through-hole components can be used in any of the PCB layers these power traces take up, and the PCIe slot location on the very edge of an mITX board is likely to be so close to where the 8-pin power cable for it plugs in that the cable-avoidance savings are likely to be minimal, if not entirely pointless.
As for compatbility with PCIe riser cables that so many SFF cases rely on? Just LOL. I'm sure Asus will make and sell a 250W PCIe 5.0 riser for you, for $150 or something offensive like that!
On ATX and mATX boards where there's some decent distance between the PCIe slot and the edge of the board where the 8-pin power plugs in, you'd need those long, heavy-duty PCB traces, and if you're struggling to squeeze a power cable into a case big enough for an mATX or ATX board, something's gone very wrong already.
Nah, this is about vanity IMO. Pure, premium, glass-cased, show builds, to be sold at a vast premium for the deep-pocketed people who would rather spend $500 on ARGB than more performance or more silence.
Pulling 150W through PCIe x16 would require either going full 12VO or redesign of ATX power plugs.
Alternative is simply plug a second 8-pin EPS 12V just for PCI-e power.
In this case, one 8-pin EPS could only power three-four PCI-e x16 150W slots.
RIP WS boards with 6+ PCIe slots : Two EPS 8-pin for CPUs, and another two for PCIe slots, and what, like... 3-4 12VHP/12V2x6 for actual GPUs ?
Problem only get's worse if you want to scale power on PCIe ports even higher.
Those of us with more potent GPUs often choose to tune them for efficiency by undervolting them too, I run my 5070Ti at 250W, my 9070XT at 250W, and I ran my 7800XT and 4070S at 200W, all without really losing any performance.
You are talking about more efficient, yet all efficiency comes from drastically increased power consumption.
Efficiency would be having 150W GPU performing well...
Not having 800W GPU which requires 3 kg of aluminum inside chassis in order to work decently...
Mr Meeseeks was tasked with this.
Additional bonus: it would make external GPUs (oculink, etc) simpler.
Yes, full backward compatibility could be achieved at a low marginal cost. Proof: USB PD exists and works. But a few companies would have to invest some engineering talent and time in it. Certainly more than was invested in (all attempts at making) the 2x6+4 connector.
And maybe just maybe there is a reason or 2 why 75 though the PCI-e slot, interference be one i would of thought.