[Request] Too much fan speed changing

[oni]

Hi, Martin

I've been using your program for fan control since I got an Aliemware laptop, it's amazing and thanks for your work.

I used to use custom auto but sometimes it's little annonying because it changes fan speed frequently way too much.
For example, I set 70 degrees to trigger 7000rpm, so when temperature goes over 70 degrees, spins up at 7000rpm, and soon temperature drops under 70 degrees thus fan slows down which makes the temperature raise again...it's kinda annoying fans speed up then speed down continueously again and again over and over.
Do you think it might help if you can make it maintain the fan speed 3-5 more seconds when fans intend to slow down (in order to prevent too much fan speed changing)?

Thanks for your time.

 

[VictorVG]

Please, read https://www.hwinfo.com/forum/Thread-IMPORTANT-Fan-Control

 

[oni]

Hi, VictorVG, Thanks for your attention

The program works flawless on my machine. My request is about another concern.
In our daily use or even in gaming, our cpu usage is not always the same, thus sometimes fan speed changes a lot because of some cpu usage spikes. It goes straight up to 7000rpm and then suddenly drops to 4000rpm and then up to 7000rpm again, I mean 7000rpm then 4000rpm then 7000rpm (continueously changing depends on our fan table setting), in case like this, I prefer it stays at 7000rpm all the time instead of fan speed changing (which makes too much noise and vibration).

So I was thinking maybe hopefully Martin could make a little change in its algorithm to make fan speed stay high for a period of time (maintain fan speed 3-5 seconds before it goes down), maybe in this way fan speed won't change that frequent anymore.
I hope I made my mind clear.

Thanks.

 

[Betaminos]

I'd love to see this feature as well
I.e. have the new (higher) fan speed stick for X seconds, before setting a lower speed again.

My fan keeps pulsing due to one measurement of a higher temperature - I guess that DELL implemented a faulty sensor filter, because I will get pretty consistent readings and then one outlier.
(E.g.: 42-43-42-42-65-43-42)
It would be great to get rid of that outlier and not have the fans erratically switch between different speeds within 1 second.

A different approach could be to use the average temperature (having it set to 4 samples, 250ms polling interval) for fan control.

Could you please have a look at this and maybe spend the time?

Thank you very much!

 

[carbonox infernox]

I've been using HWiNFO for a few months now but made an account just for this topic. I have an alienware laptop and my CPU temp spikes around within a range of 10-15 degrees (constantly while under load and intermittently otherwise). I came to the same conclusion as Betaminos that the fan speed should be tied to an average temperature, rather than the instantaneous value.

I looked for a way to do it within the program but couldn't find one. I increased the polling interval to 3000 ms to prevent the fan from changing as often, but it still changes randomly depending on whether the temp is caught at a peak or trough. Please, an averaging would be really great!

 

[Martin]

Sorry, but I don't think that using average values would be a good idea. The CPU temperature can spike very fast when there's a sudden load and using average values could cause overheating.

 

[oni]

Sorry, but I don't think that using average values would be a good idea. The CPU temperature can spike very fast when there's a sudden load and using average values could cause overheating.

I don't think average values could be a good idea, either.
But really hope the program could make the fan speed sticks for X seconds before it setting another lower speed.

 

[carbonox infernox]

That would seem intuitively correct, but it isn't necessarily true. Consider a scenario in which the core temperature increases by 12 degrees per second from 50 to 80. I've prepared a table and plot comparing the actual temperature to the temperature "seen" by a 250 ms averaging of the past 4 values, as well as a simple 1000 ms poll. See my attached picture. 

Half the time the averaged data is higher, and the other half of the time the simple poll is higher. I'd have to declare this a tie. However, in the case where the temperature has no upward or downward trend but is jumping around causing the fans to constantly change, the averaging method would clearly be superior.

I think you should at least put the option in there somewhere. It could be disabled by default and wouldn't inconvenience people who don't want it.