Question about protection circuitry in single-rail ATX power supplies

[Ranayna]

Hi,

in a discussion i recently had, the pros and cons of single-rail ATX power supplies came up.

Here is an example of such a supply:
https://seasonic.com/prime-ultra-titanium#specification

Compared to older ATX power supplies, this PSU has a single 12 Volt rail that can supply almost 1000 watts alone.
I see the use for new, extremely power hungry modern graphics cards, but i am unsure about the protection circuitry of such PSUs.
The linked supply claims to have overcurrent protection, among others

12 V is not only used on CPUs and GPUs, but also for SATA drives, some PSUs even have old style "Molex" or floppy connectors. If a PSU just has a single rail, are the cables with SATA connectors separately protected?
Those have relatively small diameter cables and the connectors are also not rated for 83 amps, so common sense says they need their own protection.
But i have read several reports of melted, or even burned, cables after short circuits on SATA cables, so if there is separate protection it does not seem to be adequate.

Does anyone know how those supplies are typically protected?

[magic]

I'm not familiar with those, but in older PSUs you would have no per-wire overcurrent protection either.

There would be a shunt resistor and current sense IC for a group of wires, for example:
- one for the main ATX plug and disk cables
- one for the CPU power connector
- one for a pair of GPU power connectors
- another for another GPU connectors

This was the case even for PSUs which had only one 12V rail (no separate transformer windings, rectifiers, regulators), the kind of design which was the norm. Sometimes these groups of outputs sharing common OC protection would be described as "rails" on the label. You typically had some 100~150W per such fictional "rail".

[strawberry]

never know whats inside that black box
cheaper/simpler is to make general protection rather than for each rail/connector
marginally better supplies are with secondary DC-DC converters for each rail/connector
performance and/or protection will most likely under-perform in real life setup
more in following  Gamers Nexus YouTube series (they miss some technical points but fair enough)

[Berni]

Most power supplies just connect all the 12V cables to the same spot on the PCB. So indeed there is no separate protection.

In a lot of cases a short circuit is severe enough to draw a huge amount of current from the rail and make the PSU shutdown anyway. At 12V to pull 85A you need 140mOhm while a 20 AWG wire has 33mOhm per meter. (but a worse contact at the short could definitely set the wires on fire)

Indeed with how powerful these CPUs are getting it would be welcome to provide at least a few zones of protection. Such as 30A for CPU, 20A for sata/molex, 50A for PCIe..etc. The issue is that this doesn't really add much value for the typical PC gamer while it does cost extra components to implement. Especially when most people just buy the cheapest PSU for the given number of watts.

[BeBuLamar]

1000W at 12V? It's almost 100A how big is the cable?

[Ranayna]

Those supplies obviously do not use a single cable
8 Pin PCI Express power connectors have 3 current carrying pairs (12 V and GND). The most power hungry graphics cards have three of these connectors.
For the CPU, at least 2 pairs are used (2pin EPS), 4 pairs are common (8 Pin EPS , and for the extremes some mainboards can use 8 (2x 8 pin EPS).
There are also 2 pairs on the main 24 pin ATX power connector, and each SATA and Molex cable strand has another one.

On good PSUs, those cables are 18 awg, so plenty enough, as long as the load is reasonably distributed.

But theres the rub, and the reason for my question: What happens on a short, or near short, on - for example - a SATA connector? The AWG 18 cable will definetely not be able to carry all the current the power supply can put out and melt or catch fire.
So i would have expected that those cables are protected according to their individual current capacity, but apparently for many PSUs they are not.

And enter crap like this: https://support.fractal-design.com/support/solutions/articles/152733-how-do-i-connect-the-cables-on-the-fan-controller-
Just look at those *tiny* cables on that SATA connector. If you are lucky that thin cable acts like a fuse, if not, it will burst into flame...

[LinuxHata]

I don't know whenever my input counts, but...

When using Intel 486DX4-100, with rather noname 200W AT PSU, I came across an older hard drive, which I've plugged in and pressed the "power" (AT supplies had mechanical power switch). As it turned around, hard drive had a short, so excess current went into cable, it got hot, but then protection turned on and psu switched off, but since it was mechanical power on, after shutdown, psu restarted, again catching a short and starting to melt the cable, all this happened in loop, for about 5 seconds, until I understood what was happening and unplugged the PSU. IDE cable was totally fried, and if not me, it would definitely catch the fire.

[Berni]

This is one improvement of modern PSUs. Once tripped into protection they only restart when you turn them off and on again. So this would not happen.

But as others have said that might not help when the PSU can push 80A trough a single cable, turning it into a load in the form of a 1000W heating element.

The connector system on PSUs is well thought out tho. The low wattage PSUs have less connectors with less pins while each connector has a defined limit of how much it should draw. This way you don't end up overloading the PSU by plugging in too power hungry hardware. If you try to then you run out of cables to do so (Unless you resort of hacky adapters). For example the 6 pin versus 8 pin PCIe power connectors only have GND as the 2 extra pins. No extra 12V lines, but the existing ones are designed to be thick enough to carry the 150W. At the same time the graphics card can use one of the ground pins to detect if a 6 pin was plugged in and limit the power draw down to 75W per plug. All this was thought up by Intel in the ATX specification.

Computers catching fire seams to be a pretty rare occurrence (but does happen). Perhaps things don't short out that often, or perhaps the cheap PVC insulated cables can melt fast enough to short together near the PSU and kill it before it does start to properly burn.