how to merge two PWM source into one based on max of them

[memberOfEarthSince1970]

Dear Internet,

 

Wasted a lot of time for finding a solution but ended there.

 

Case: two PWM lines: CPU and GPU (or tree if SLI).

 

I want to drive water loop over maximal value of one of these PWM signals. Looking for some kind of hardware (read as reliable) based implementation of SpeedFan. 

 

Is there anyone who found silver bullet for this problem?

[Mira Yurizaki]

Where is the PWM signal coming from?

 

Edit: In waiting for an answer, if it's coming from the fan headers, you're also going to have to feed back a value in order for the fan controllers to be happy.

 

If this is the case, you're going to need to roll your own setup using a microcontroller. Even then, a better solution would be to use a thermal probe and adjust the cooling system that way. Trying to sample PWMs requires knowing what the base frequency of said PWM is and sampling it over a reasonable period of time. With a thermal probe you can make adjustments instantaneously.

[memberOfEarthSince1970]

29 minutes ago, M.Yurizaki said:

Where is the PWM signal coming from?

CPU: PWM header on MB

GPU: PWM header on GPU

 

Regarding thermal probe: is it so fast to react as internal reading from chip by BIOS?

[memberOfEarthSince1970]

1 hour ago, M.Yurizaki said:

you're going to need to roll your own setup using a microcontroller

https://shop.aquacomputer.de/product_info.php?products_id=2673

https://shop.aquacomputer.de/product_info.php?products_id=3481

something like this? what is the bast placement for thermal probe on cpu/gpu?

[Mira Yurizaki]

43 minutes ago, memberOfEarthSince1970 said:

https://shop.aquacomputer.de/product_info.php?products_id=2673

https://shop.aquacomputer.de/product_info.php?products_id=3481

something like this? what is the bast placement for thermal probe on cpu/gpu?

That wasn't what I was thinking, but one of those is better than what I had in mind anyway.

 

The best placement for a thermal probe is as close to the die as possible, without compromising the amount of contact between the processor die and the heatsink.

[0ld_Chicken]

Is there a reason that you can't monitor and use water temps?  Maybe I'm not understanding, but controlling from water temps only makes sense, since that is what you're cooling. Aquaero makes this very easy to do

[memberOfEarthSince1970]

7 hours ago, 0ld_Chicken said:

Is there a reason that you can't monitor and use water temps?  Maybe I'm not understanding, but controlling from water temps only makes sense, since that is what you're cooling. Aquaero makes this very easy to do

temp of water changes very slowly, I don't think that this is reliable metric for PWM, what do you think? how do you drive your loop?

[0ld_Chicken]

4 hours ago, memberOfEarthSince1970 said:

temp of water changes very slowly, I don't think that this is reliable metric for PWM, what do you think? how do you drive your loop?

Well since the water is what is actually cooling the components, and the fans are cooling the water it really makes sense to me.  I like that it changes slowly because the fan speeds also change very slowly.

 

The component temps are directly correlated with the water temps, as my water temps rise my component temps rise by the same amount

 

 

 

I have an aquaero (5lt) and 5 temp sensors (2 water, 3 ambient air).  Each sensor type is averaged and then I created a virtual temperature that is the difference of the two averaged values (known as DeltaT).  Basically the aquaero uses this DeltaT instead, so it also adjusts with ambient temps.  This way I don't have to readjust my fan curves in different seasons. 

 

[memberOfEarthSince1970]

6 minutes ago, 0ld_Chicken said:

Well since the water is what is actually cooling the components, and the fans are cooling the water it really makes sense to me.  I like that it changes slowly because the fan speeds also change very slowly.

 

The component temps are directly correlated with the water temps, as my water temps rise my component temps rise by the same amount

 

 

 

I have an aquaero (5lt) and 5 temp sensors (2 water, 3 ambient air).  Each sensor type is averaged and then I created a virtual temperature that is the difference of the two averaged values (known as DeltaT).  Basically the aquaero uses this DeltaT instead, so it also adjusts with ambient temps.  This way I don't have to readjust my fan curves in different seasons. 

 

thank you very much, you did my day, now I have a plan

[memberOfEarthSince1970]

22 hours ago, M.Yurizaki said:

That wasn't what I was thinking, but one of those is better than what I had in mind anyway.

what did you mean in origin, looking for cheaper solution

[Mira Yurizaki]

14 minutes ago, memberOfEarthSince1970 said:

what did you mean in origin, looking for cheaper solution

Buy a microcontroller kit like an Arduino and program it yourself.

[asand1]

Speed fan can use both GPU temp and CPU temp to control the same case (radiator) fans. The hotter of the two components will dictate fan speed. Set the according headers to react to both temp sensors. Me need of extra hardware.

[memberOfEarthSince1970]

main idea is get rid of speed fan, because it is just software