Zitat von »Remayz«

What is that blue "Intake" sensor? fan speed or temperature?
It looks unstable AF. I still feel if your sensors were stable you wouldn't even need to bother with overcomplicated sensors to try to find a useful signal out of that noise '
If it's a fan speed.. same deal, teh readouts on my octo are dead stable.
Temp sensors see variations only at the second decimal, and still, on water it's like -+0.01°C . This is what i am smoothing with that 10sec averaging.

Zitat von »Remayz«

weeeeeeell that jagged curve on your data viewer is super unstable
But as you installed the sensor on a fan intake, it makes sense. There's so much disturbance there, you can't get a stable reading without a lot of averaging (and lose responsiveness).

Why take the highest and lowest temperature of the room? the only temperature that matters is the one your radiators are intaking in real time. This is the only one that affects the radiator efficiency. real time water temperature, and real time air temperature.
If you calculate any other value, you are modulating fan speed based on a temperature that is not the one of the air they push, so it basically, by design, becomes either inefficient on the lower end, or needlessly noisy on the upper end.

This is where i have a problem with all the calculations. The most efficient cooling is always governed by dT. It is self regulating, winter or summer, smoothing is provided by the water thermal mass (you can introduce some more with averaging) but the more you deviate from that, the more inefficient/unstable the cooling becomes.

Maybe with air cooling several discrete components with independent cooling it can be useful, but with watercooling, i still fail to see the point '

Zitat von »Remayz«

uhm.. the sensor doesn't increase fan speed before the water temp on the graph, it either does nothing (staying flat because it never decreased fans speed when it should have) or lags behind every single time as i experienced.

All i see is a water temp that barely changes, and a fan curve that jumps all over the place instead of being as flat as the temperatures. Of course, that's masked by the very slight fan response because in absolute the control temp varies by little, but same deal : on a smaller loop with larger temp variations, and larger fan speed changes, the sensor will be noisy and very unstable.

Zitat von »Remayz«

Yes.. the regulator is directly related to fan speed. that's what a fan curve does. th fan curve will be a direct replicate of the temperature curve.
Yours is very flat because you did set it that way and as a consequence you basically heavily mask the defects of that sensor. with smooth fan curves, i can make the most discombobulated sensor look like the best thing since sliced bread

A steeper fan curve will just make a noisy mess. I mean look at your fan curve, it's not a curve it's a bar graph basically. it jumps values instead of transitionning smoothly as fan curves do. that is pure sound nuisance.

I still fail to see the relevance of the sensor. It doesn't control the fans well, it acts on imaginary temperatures (even if they are called hot/cold/ or "temperature map", they are not the real temp that the radiators are experiencing).

With such small temperature variations and a very gentle fan curve, you can't assess properly how the thing works or fails.

On my loop with 3 slim rads, and a 10900k and 3090, the thing was just bonkers. And from the graphs you showed, it still is. It just doesn't react well to heavy loads and faster temperature changes. And disabling 3 out of your 4 rads won't do much. there's just too much thermal mass in your loop, and not enough heat load to experience it. I mean you could set the fans to a quiet fixed speed and not even control them, it would work fine.

Take your last graph, and imagine the PWM% variations do not span 1% but 20%. you'll see what i mean.

Zitat von »Remayz«

If you can't see the defects in your sensor from your own graphs then what is there to show? you already have them.
I told you, because you have a fat loop, very small thermal load in comparison, your fan curve is very flat, so you don't hear how bad it is since your fans speed vary very little.

just look at the bar graph your fan speed is, now apply that to a more steep curve and you'll get it.
Read my last phrase in the previous response, you'll understand why your sensor is unuseable unless you are deaf, or have a loop like yours with relativelylow load and huge thermal mass.

Your sensor has defects everywhere. it is not smooth, depending on the fan curve it is annoying sound wise, accelerating and decelerating fans steeply, it's noisy again because it doesn't reduce fan speeds quick enough when load decreases it lags behind thermal load, and it is very inefficient since it regulates fan speed based off virtual temperatures instead of real ones.

The way you implemented in your loop hides all those flaws, again because of your very flat fan curve.
I'm not going to lose time making graphs since you already showed them. It only takes a little bit of imagination to see what that sensor response curve does on a steeper fan curve. speed jumping 5-10% every 5 minutes constantly.

Now if you want to defend that sensor as the best thing ever made, you do you, i was just trying to help you see it does worse than a simple sensor with only two inputs and a substraction.

Zitat von »Remayz«

you seem to only consider feedback that goes your way, and refuse feedback that shows your sensor doesn't work well..

by the way you added an input called "Hour" that is in fact a temperature. No idea what that is.

nevermind, it's the hour of the day. still no idea what it does in a cooling sensor.

Zitat von »Remayz«

Well.. setting up the cooling exactly as you do is quite the adventure.

Your fan curve ensures the fans run between 80 and 100% speed all the time. Indeed it cools well but who likes sitting next to a turbine?
Setting the curve to 0 arch makes it quieter but it's still a good 64% speed at idle which is too much. that's what i have at full load usually.
at + 13 i have something decent, but that only makes the fan acceleration noisier since they reach over 2000 RPM in a matter of seconds. And when stopping the load, the sensor holds that racket for minutes... off we go to reset the sensor..




I've made a curve that would make sense if i wasn't using dT, just based off water temp. the coolant never goes below 20°C, and as you i keep 40°C as full speed.
No arch, straight curve, otherwise it's too noisy at idle, and already, the fans run too fast at 50% PWM.
That's a bonus for you, i already push the cooling from the start.
25% at 20°C, 100% at 40°C, no arch.


Starting a game, the fans take off quickly because, despite the water being at 28, the sensor goes to 36 - 37 on the fan curve. Since the radiators aren't even that warm, it only causes noise and has no cooling benefit yet.
Now in less than 3 minutes, my fans are at 94% PWM because the sensor drives them as if the water was 38.3°C.



The water temp keeps climbing obviously since the 3090 is pushing over 350W but i'm being blasted by 2400 rpm fans while just following the water temp (no virdual sensor) they would be gently climbing around 65 - 70%.

What should i do ? make an arched curve? that would be even steeper accelerations and more nuisances. Raise the max temperature? I shouldn't have to if i want to limit the water temp to 40°C, but your sensor just uses "imaginary" temps, so i can't set a curve based off real temps. Right now the water stabilizes around 31, but the sensor keeps 38.
While i was typing that, the speeds dropped 10% because the sensor finally left its plateau.

But guess what? dropping that fast, the RTX does what it does best, the water temp goes back upand the fan speeds too.



Now the water is at 31.5, and the sensor at 33 which is closer to reality.
Usually under full load, my fans run close to 1700 rpm, right now they are just under 2000. i won't lose hours trying to get a similar result, it's close enough.

Waiting for the water to cool down, again, the sensor lags behind and keeps the PC noisy. the water is back to the temp i started (around 24) at but the fans are still at 70% because the sensors holds a 32°C temp.
And again, as i write that, the sensor drops and the fans slow down fast, and i have to stop the test there because the PC is chilling me, blowing air so hard on my side..

The point stands :

Your fan curve, that i completely missed from the first post, is what makes you believe you sensor is useable.
If you are at 90% PWM at idle, of course the sensor will work. those big swings will be so small you won't even notice them, but nobody does a curve like that. it's just too noisy, it clogs dust filters too fast and it defeats the purpose of having a fan controller. you could set a fixed 80 or 90% and be done.

With such a fan curve you can't assess at all how your sensor works. You are basing your own conclusions off a test that is basically rigged to succeed everytime, since no fan control is needed once your base speed is so fast..
You could make the most ridiculous sensor using every single module available in the playground and still have good temps.

You are using Vardars, they have a 0RPM feature, so, let's disregard it and try using your controller, varying your fan speed between 26% at whatever your water idle temp is, and 100% at 40°C. See what it does.

Now when you use delta T (which i remind you is a one operation virtual sensor) you get something like this :




the fan speed increases gradually, you barely notice it, and it follows the radiator efficiency which, again, is directly red as the water over ambient temperature difference.
This alone dictates how efficient your radiators are at dissipating heat. Blasting the fans while that delta is low only makes noise and has NO cooling benefit, none, at all.
This is why using imaginary calculated temperatures makes absolutely no sense. you disregard the one value that governs radiator efficiency and provides the best base for fan control.

Now if you want to manipulate that value to have a different fan curve between 8 AM and 10 PM, have fun but i'll maintain that the fan curve you used is too extreme to warrant the use of any control variable, and the virtual sensor is way too unstable and extremely inefficient at cooling if you use any fan curve that doesn't pierce your ears constantly at idle.

Zitat von »Remayz«

I didn't follow what the hour sensor was doing because it makes no sense to have a timer in a cooling sensor, that's all.

Who cares what temperature my room was yesterday? it's the temperature it sits at RIGHT NOW that affects cooling. So you can remove that memory shenanigan from the sensor. it serves no purpose.
By the way, in a hot day, you don't accelerate fans. Again, it is the delta between ambient and water that governs how much heat the rads can dissipate. if my room is at 35°C and the water at 37, guess what benefit i'll have to set my fans at full speed? none. 2°C, that's an idling computer.
That's why everyone uses dT. it auto adjusts without needing overcomplicated calculations and random high speeds.

And i gave you my fan curve in my post : 25% at 20°C, 100% at 40°C, no arch.

It's not hasty assumptions.. it's confusion about throwing useless stuff in a cooling sensor that only serve at making it unuseable.

My loop is prefectly fine keeping the temperatures low at 1500 RPM. Why does your sensor assume it is NEEDED to run them at 2300 rpm for the same result?
That is with a curve that has a maximum fan speed at 40°, yet the sensor sends them at that speed while the water temp isn't even half that. This is what i call inefficient and noisy from the start.

But if you don't know how to make a fan curve, why even bother make a sensor? and one as complicated and useless at that?
I mean, we're still here trying to help, probably because we're idiots at this point ' but please, do not come and post your sensor as if it was the best non compromise thing ever done. and you tested it for one year without noticing it is just a massive flaw, from end to end. The fan curves, the sensor behaviour, nothing makes sense at all.

It's fine being new to watercooling, there's nothing wrong with that, but don't try to play expert with such a unuseable sensor. It's really embarassing.
just use common sense and look how cooling works on a radiator (yes, they are radiators, not intercoolers). only 3 factors play any role. water temperature, air temperature, and air flow through the thing (fan speed). two inputs, one output. Need more tweaking? direct PID control does that perfectly already.

The hour of the day, the temperature from yesterday, the truncated square root of the average coldest temp times the hottest temp squared isn't helpful. neither is blasting fans at full speed on cold water.

No need to review data really, just look at the response graph of the sensor, we both used the same one. it's all over the place, it doesn't cool better. It's just a convoluted brute force method that may work if it was an aircooled build.. even then it would still be noisy.

I mean, have fun, test again, review data, make graphs, but as long as you don't know how to assess their relevance in the context of cooling a PC, it won't be useful.