CPU to FAN1?

[Borja]

i read in another post that i can connect cpu fan to fan1-4 connectors. Never to fanA in a supermicro MB. Is this right?

Thanks.

 

[Bidule0hm]

Yes.

 

[Borja]

Yes.

it's rare because Fan1 is working at 1000rpm and fan3-4 and fanA is working at 2400rpm

 

[Bidule0hm]

I don't think I understand what you want to say, but the SM MBs (usually) have 2 fan channels: FAN1-4 and FAN.

FAN1-4 is controlled by the CPU temp and FANA is controlled by the PCH or system temp.

 

[Ericloewe]

it's rare because Fan1 is working at 1000rpm and fan3-4 and fanA is working at 2400rpm

They're different fans.

The closed-loop control controls the temperature, not fan speed.

Warning: Control Theory ahead

Fan speed is just a (vector of) state variable(s). System state is multiplied by the output matrix and the result is the temperature you want to control.
The system has quite a few state variables besides fan speeds, even in a simplistic approach: Two or more per fan, relating PWM input and "airflow" output, at least one for the system whose temperature is to be controlled, plus whatever the controller has.

Of course, engineering practice here is to say "Screw this, throw in a PID controller".
Fancy controller design relies on knowledge of a good model of the system, which is not practical in the case of PC fan control. So the default option is to throw in a PID, probably with small k_I and k_D, and then add a bit of adaptive control ("the fan crapped out, set them all to max and try to recover or compensate", though not very fancy, is technically adaptive control).

 

[Bidule0hm]

and then add a bit of adaptive control ("the fan crapped out, set them all to max and try to recover or compensate", though not very fancy, is technically adaptive control).

Yeah... let's call that something like stall recovery :D

I'm not even sure there's a PID controller, looks like it's just a P controller...

 

[Ericloewe]

Yeah... let's call that something like stall recovery :D

I'm not even sure there's a PID controller, looks like it's just a P controller...

I did say "small k_I and k_D". :D

 

[Bidule0hm]

Yeah, I think there are something like 0.001 then... :D

 

[Ericloewe]

Introduction to Control Theory:

1. Learn P.
2. Learn PID.
3. Learn all sorts of state-space control theory black magic.
4. Use P.
5. If 4 did not solve your problem, use PID.
6. Pretend 3 never happened.

 

[Bidule0hm]

Yep, right now I've done 1, 2, 5 but I still need to test the PI part of the controller (no need for D here and to tune the Ziegler-Nichols rule will be simple and good enough) before I make the full version and release the schematic and explanations :)