Hardware manufacturers (dell, hp, etc) already do strange proprietary methods of 12VO.
Yeah, and they all suck. They lock you into having to throw away the entire machine and buy a new one when the power supply does fail, and it will. I've already started seeing cries for help about it, and the only options are to buy another used one and hope it works, or do component level repair on the existing unit, assuming it's not a charcoal briquette. The latter is usually far out of the realm of experience of the average computer user, and it's definitely not something I'd recommend to the novice, dealing with high voltage switching and all. Even back before 12VO was a thing, OEM machines with proprietary PSU form factors were a big problem.
They also lock you into a path of no upgrades. Dell, HP, etc. usually spec a power supply that's on the knife edge of working and not, meaning even adding a PCIe video card that doesn't need external PCIe power connectors may cause the system to become unstable and fail. This is a VERY common issue on SFF machines. But on mid towers it's even worse. The OEM will spec multiple proprietary power supplies depending on the system configuration. So if you want to add a higher powered video card or additional drives, you often have to find the part number for the higher wattage PSU, source it and install it. Then possibly replace the logic board with a different model that has the required PCIe power connector if it's a pass-through and not coming off the PSU. Same for more hard drives or fans.
The other option if you can fit a normal ATX PSU is to rewire the harness, but this is a monumental pain in the ass to do, and voids any warranty on the ATX PSU.
I think it would be nice if they standardized. ATX12V 24pin can continue for all i care, although I cringe every time I see those 24pins and know inside that I use about half of them to their actual capabilities (and 1 of them NOT AT ALL!).
It is true that -12v is rarely used anymore. The last slot is was present on was PCI and is mostly just used for biasing. Some of the ATX power supplies used by OEMs shortly before switching over to proprietary 12VO omitted -12v entirely.
But the -12v rail only takes a couple of components inside the PSU. It's usually an LM7912 (a negative linear regulator) and one or two capacitors. I dread the day when this rail goes away because it means powering older machines will be much more difficult. Creating a negative power rail from a positive rail is a lot more difficult than just having a center tapped transformer and a negative linear regulator inside the power supply. There are buck regulators and negative charge pumps that can do this, but the output isn't as clean than a simple linear regulator.
DanNeely didn't mention anything low power in the post you quoted, so this was confusing to me.
It was obvious to me what he was talking about with the multiple secondary power rails. Most of them are low current, low voltage rails, like I was talking about with audio and the Ethernet PHY. Not everyone does board level repair though so I shouldn't have assumed everyone knew the difference.
The buck converter we use on our 12 drive SAS/SATA backplane is 12A (20A MAX). It comes in under $2 at high enough quantities. Add in all the inductors and capacitors around that, it's probably going to be under $10. Many motherboards don't need 12 drives. So I think your $20-$30 estimate is too high.
Regardless this cost would presumably (although maybe slowly?) be subtracted from the PSU which no longer needs to do this job.
I'm going on the assumption of 100% replicating the missing 5v and 3.3v rails from the power supply to the motherboard, not just the bare minimum. I don't need manufacturers deciding what I do and don't need to be doing with my system. If I lose the 20A 5v and 20A 3.3v rails from my power supply, I expect them to be provided on the motherboard with no reductions.
Motherboards already do run +12V to fans, so I think there is no difference here.
Oh there is quite a difference. In the past with cooling only fans, it wasn't an issue. But with the rise of RGB fans, some of them can pull over an amp a piece and cause damage. This is one reason fan controllers have become a thing. I in no way would want a Corsair 1000D case full of fans being routed through the motherboard. You're talking about current levels equivalent to having a second high powered video card in the system.
If you can't trust a motherboard maker to make CPU regulators then I think you are simply buying the wrong motherboards. Cheap motherboards will have cheap parts no matter what you do. It's the same with cheap PSUs.
When reputable manufacturers can't even get CPU vcore right, for damn sure I'm not going to trust them to do other high power rails correctly. And this is not a recent phenomenon, I had Asus, Gigabyte and MSI boards back in the Socket A era that couldn't get it right either. I remember having to physically modify CPU power VRM circuitry on even supposedly "good" or "high end boards" because they were running vcore so out of spec. Many times this is on purpose to go along with them running the CPU out of spec to try and push their motherboard higher up the benchmark charts by squeezing just a bit more out of the CPU by running it faster. This leads to hotter running parts that wear out faster. This is still a thing today. Just look at all of the boards that ignore turbo boost guidelines that cause problems, and Intels B560 that came out that attempted to put a lid on it, only for people to complain about mysterious performance loss, when in fact it was the motherboard manufacturers being caught in their own lies.
I had one Asus board that was pushing 2.0v to an Athlon that was only supposed to be getting 1.65v, it was no wonder why the CPU had temperatures in the stratosphere. I think thermal junction of an Athlon back then was 62-65C and I couldn't get it under 75-80C at idle, which initially lead to the investigation.
Assuming that a name brand board is getting VRMs right is disingenuous and naive.