How does a power supply know how much power to deliver?

TheMohawkNinja

Reputable
Aug 20, 2014
24
0
4,520
Hello,

As I understand, while the PCI slots and 6-24 pin connectors all have a constantly stated amount of power that they deliver to a given component, I am also aware that things like CPUs and GPUs consume power at a rate proportional to their current computational load.

So, I am curious how that variable power usage actually works. Is a CPU/GPU always drawing a consistent amount of power, but internal hardware limits how much of that power actually reaches the processor, RAM, etc, or is there sort of a feedback loop whereby the CPU/GPU tells the power supply how much it needs, and the stated numbers for the connectors (like 75 watts for the PCI slot IIRC) is actually a maximum draw, and not what it constantly is drawing?
 
Solution
To explain in a way that you can understand.

The power rating on the side is the maximum that it is -rated- for. The maximum that it's components are capable of delivering without shorting out or burning up, and that the heatsinks and fans are capable of keeping within it's temperature limits.

Then, think of it as a pump, and all the pieces attached to it as restricting holes which limit the amount of power which can flow through them (like a pump moving water). The pump will push as much current out through those holes equal to the size of the holes. Lets say the PSU is a 500w 'pump'. If the size of the 'holes' is 400 watts, that's all the power that can be pushed through it. Now if you put bigger 'holes' equal to 600w the pump...

Rookie_MIB

Distinguished
To explain in a way that you can understand.

The power rating on the side is the maximum that it is -rated- for. The maximum that it's components are capable of delivering without shorting out or burning up, and that the heatsinks and fans are capable of keeping within it's temperature limits.

Then, think of it as a pump, and all the pieces attached to it as restricting holes which limit the amount of power which can flow through them (like a pump moving water). The pump will push as much current out through those holes equal to the size of the holes. Lets say the PSU is a 500w 'pump'. If the size of the 'holes' is 400 watts, that's all the power that can be pushed through it. Now if you put bigger 'holes' equal to 600w the pump will try to push that much power through them. However, the parts of the pump will overheat and either shut down (due to thermal protection) or burn up (too much heat) because the parts of the pump are only rated up to 500w.

As for variable loading, think of the CPU as a house, with the power draw either being like lights and the clock speed like AC. If the CPU is being loaded that means the whole house is lit and the AC is cranking. But, like a person who can turn off lights when they're not needed, and turn the AC down when it's not needed as much, the CPU can power gate and clock gate lots of 'rooms' in the 'house'. It can turn off sections that don't need lights (which saves power by being off) and turn the ac up or down which also affect power usage.

The GPU is the same way, although mostly it's power usage is clock-gated, but it can turn off processing units too (turns off the lights).

So - the less power it needs - in terms of what I described above, it's like a variable nozzle on a hose. It can go bigger or smaller and that's what limits how much the pump 'works'.

There are certain things which are still always on. IE, the 4gb of ram will always have power usage going on keeping it's contents refreshed. Certain other chips stay on (USB, audio, etc) but many things can go into a sleep mode reducing idle power usage.
 
Solution
Holly crap, I kinda like how wolf explains it, simple and something everybody use everyday.

OP, power delivery is not as sophisticated as you make it sounds, the only "intelligence," if we can call it that, is overload protection, because things may get too hot and plastic start melting and shorts ensue and that would be unsafe.