I like that one more. But as I said, why stop at 12 VDC? 24V? 48V? 120V? And reduce the number of wires need to carry the power to the motherboard?
48V would be my "let's just imagine every PC on the planet did not need to consider upgrades" pick. Lets you up PCIe slot power to 300W with no trace changes, slight efficiency gain in the primary (wall power to 48VDC) conversion, minor efficiency drop in secondary (48VDC to 1-1.5VDC) conversion for CPU & GPU. Already has partial report in some cases due to use of 48VDC distribution in datacentes & HPC. Big problem is a;ll that legacy hardware: either all components need to be double their SKU count, or all components need to be built with redundant power stages to handle both voltages. And if you try and maintain some cross-compatibility inevitably components will be fried by feeding them the wrong voltage.
New voltage new bus standard, which may or my not be backwards compatible. Consider when PCIe replaced, PCI and AGP, or PCI replaced ISA. Each time a new bus was released with zero backwards compatibility.
With both, there were transition periods where the ports coexisted. Boards with ISA and PCI, boards with PCI and AGP (which became the norm), a brief crop of boards with PCI, AGP and PCIe, some with PCI and PCIe, etc. With a new voltage rail, the only way for that to work would be effectively two PSUs and two parallel power delivery systems within one box, and always being one edge case away from bridging a 12V and 48V line. e.g. with a GPU using a wide-range VRM that can take 12V or 48V input, you'd need to reliably isolate the PCIe card-edge power lanes from the 'PCIe48V' plane or risk shunting 48V to the motherboard and releasing lots of magic blue smoke. Switching the delivery voltage really needs everything to be switched at once rather than incrementally.
