I've got the following idea in mind. I have a water-cooled PC, I wish to minimize fan/pump noise by lowering their speeds, while maximizing the cooling that can be done.

There is the simple solution of setting a target water to air delta, (water temp taken from a reservoir, operating under the assumption that this will be the average water temperature) and allowing for (PID?) control to ramp up speeds until that delta is met.

But is there a better way? For example I can sample die temperatures of the CPU/GPU, the inlet and outlet temperature of each heat exchanger (to within +/- .2 degrees C), and obtain flow speed (+/- 2%) . Can I do something interesting with these data points in real time?

What equations should I be looking at to solve this? Am I over-engineering by attempting to calculate all of this? For example I know from test data that at most the delta across most radiators for this application will be approximately 1-3 degrees C

  • $\begingroup$ What exactly do you mean by delta? Delta T from ambient to coolant is 3 °C, max.. What will work is: Get really big radiators. I can't remember the name right now, but there are radiators designed for 9x120mm fans, these will rarely spin up during load and only for a limited time with low RPM. Edit: Something like this : HX-1080 Radiator $\endgroup$ – idkfa Aug 7 '17 at 6:58

In practice this is likely to be empirical problem as it;s hard to predict exactly what nose is related to a given output or indeed how annoying a particular type of noise will be.

There is also the fact that with water cooling you have a lot of flexibility in where you actually place you heat exchange ie you can make the pipes a bit longer with relatively little additional pump load.

This sounds like an optimization problem so hard to generalise . My intuition is that pump speed will be better for dealing with bulk load while fan speed will be able to respond faster to load spikes.

Ultimately the best thing is to write some software which will allow you to test and optimise the system alternatively you could manually tweak it with pots and have fail-safe override which turns everything up to full if temps get iffy, which might actually be more fun.

You also wouldn't really want to miss the opportunity to have an alarm which said 'core temperature approaching critical'. at which point you can mutter 'setting coolant flow to 90%'....'temperature approaching normal'.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.