fan control for x62 on linux

[bitonic]

Hi,

I've reverse engineered the windows utility posted on 51nb so that I could control the fan speed on linux: https://github.com/bitonic/x62-fancontrol . I don't know what the fan speed numbers represent exactly, but it seems to work. It also reads the temperature in the same way that the windows utility does. It seems to be vaguely consistent with the package temperature from coretemp.

I didn't test if the behavior was consistent with windows since I never even ran the windows utility, but it's just setting IO ports so I'd be inclined to think that if it works at all it's doing the right thing.

Using that code it should be trivial to write a daemon that automatically regulates the fan speed based on the temperature in a way that is less fidgety than the default one. The windows utility polls the temp and adjust the fan accordingly every second, so I'll probably write it in the same style once I figure out what the fan speed numbers are.

Francesco

[bitonic]

ok, i've added a "manager" mode, you can run it with:

Code: Select all

sudo ./x62-fancontrol manager

This is already _much better_ than the default, the laptop is silent most of the time and the fans are not as jumpy. I can tweak the daemon more to wait a bit between switching levels but this is probably enough for now.

[NonesensE]

Great work, thank you!

An idea: On my 3rd batch X62s, the fan is pretty much perfect, but when the temperature is low, sometimes the fan turns on at a high speed for a second or so. After monitoring the temperature for longer periods of time, I think this is due to occasional erroneous temperature readings (up to 255°C, but can be anything from 0 to 255°C). Neither the EC fan control nor the 51nb fan control program seem to be designed to handle wrong readings. To cancel out those non-real temperature spikes, I'd recommend a moving average, e.g. over the last 5-10 temperature readings. Of course, this makes refresh rates faster than 1 per second necessary, thus it would be better to implement it in the EC, but I don't think we'll ever see this.

[bitonic]

Great work, thank you!

An idea: On my 3rd batch X62s, the fan is pretty much perfect, but when the temperature is low, sometimes the fan turns on at a high speed for a second or so. After monitoring the temperature for longer periods of time, I think this is due to occasional erroneous temperature readings (up to 255°C, but can be anything from 0 to 255°C). Neither the EC fan control nor the 51nb fan control program seem to be designed to handle wrong readings. To cancel out those non-real temperature spikes, I'd recommend a moving average, e.g. over the last 5-10 temperature readings. Of course, this makes refresh rates faster than 1 per second necessary, thus it would be better to implement it in the EC, but I don't think we'll ever see this.

that'd be very easy to do, and a good idea. i'm not sure when i'll get around to it, but obviously PRs are welcome. i think we're far from being a performance drag, even once every 250ms would be perfectly fine.

[bitonic]

ok, i've written a small script that survives wakeups from suspend -- see readme. the script works great so i doubt i'll update anytime soon, but contributions are welcome!

[NonesensE]

In theory, there is Notebook Fan Control (NBFC) for this purpose (and is available for Win&Linux), no need to re-invent the wheel. But I don't know yet how to condense your code into a NBFC config file
Tried to set a few registers and received surprise hibernation

[mifritscher]

I've hacked something together for my board. For me, the fan starts at 50, 100 is the max.

Code: Select all

<?php

function getTemp() {
	$temp = exec('./x62-fancontrol temp');
	$temp=explode("\n",$temp);
	$temp=$temp[count($temp)-1];
	$temp=explode(" ",$temp);
	$temp=$temp[count($temp)-1];
	return $temp;
}

function setLuefter($staerke) {
	$temp = exec('./x62-fancontrol set-fan-speed ' . $staerke);
}

$wartezeit=5;

//vorspannen, damit beim start der Luefter eher zu viel dreht
$count_extrem=$wartezeit;
$count_heiss=$wartezeit;
$count_warm=$wartezeit;

while (true) {
	$luefter=0;
	
	$temp1=getTemp();
	sleep(1);
	$temp2=getTemp();

	$temp=min($temp1,$temp2);

	echo "Temp: $temp\n";

	if ($temp>85) {
		$count_extrem++;
		$count_heiss+=2;
		$count_warm+=3;
	} else if ($temp>80) {
		$count_extrem--;
		$count_heiss++;
		$count_warm+=2;
	} else if ($temp>75) {
		$count_extrem--;
		$count_heiss--;
		$count_warm++;
	} else {
		$count_extrem--;
		$count_heiss--;
		$count_warm--;
	}

	$count_extrem =min($wartezeit*2, $count_extrem);
	$count_extrem =max(0, $count_extrem);

	$count_heiss =min($wartezeit*2, $count_heiss);
	$count_heiss =max(0, $count_heiss);

	$count_warm =min($wartezeit*2, $count_warm);
	$count_warm =max(0, $count_warm);

	if ($count_extrem > $wartezeit) {
		echo "extrem\n";
		$luefter=85;
	} else if ($count_heiss > $wartezeit) {
		echo "heiss\n";
		$luefter=63;
	} else if ($count_warm > $wartezeit) {
		echo "warm\n";
		$luefter=51;
	} else {
		echo ("kalt\n");
		$luefter=0;
	}
	setLuefter($luefter);
}

?>

[OhneHerren]

Is it possible to read the fan speed?

Using your tool on my third batch X62, fan speed 0 is maximum, fan speed 1 is off and fan speeds 2 to 100 are increasing levels with 2 being the minimum and 100 being the maximum. Fan speeds 101-255 are equivalent to fan off.

[bitonic]

Is it possible to read the fan speed?

nope, i just reverse-engineered the windows app -- it was quite clear from the binary what code sets the speed and reads the temperature. sadly the windows application doesn't include anything regarding reading the fan speed.

somebody with more familiarity with how these controllers usually work could probably spot some patterns and make a guess on how to read the fan speed, but i have no clue.

[wujstefan]

Does this 51nb app work for you well?

I can't make it work under Windows. Did nnot try out the Linux version yet.

[bitonic]

Does this 51nb app work for you well?

I can't make it work under Windows. Did nnot try out the Linux version yet.

yes, it works great. the only problem is that you have to restart it after resume, but the x62-fancontrol-manager.py script i added takes care of that. i have it start up on boot using a simple systemd service.

[wujstefan]

Then I must do something wrong way. I set all the data in ThermalTable, run the .exe file, and it works OK... for a short while (5-10mins). I've set everything to keep my system silent until it reaches 60C, and then pump out the fan with speed that ommits the low-pitch noise.

EDIT: it seems that the application stops working properly when going down with temps. This is my termal table:
00,0,0,0,
49,0,0,0,
60,50,50,0,
73,70,20,0,
82,75,2,0,
255,0,100,0
I've got the app in the "autostart", generally a bit annoying I need to minimize or close the window directly after start, but it is as it is. The fan settings work OK just as long as they are going up - when they are turning back to the state where the fan shoud not spin at all, I've got the noisy, low-speed fan movement that should not occur with such settings at all.

I should move this one to another thread - sorry for hijack

[bitonic]

Then I must do something wrong way. I set all the data in ThermalTable, run the .exe file, and it works OK... for a short while (5-10mins). I've set everything to keep my system silent until it reaches 60C, and then pump out the fan with speed that ommits the low-pitch noise.

EDIT: it seems that the application stops working properly when going down with temps. This is my termal table:
00,0,0,0,
49,0,0,0,
60,50,50,0,
73,70,20,0,
82,75,2,0,
255,0,100,0
I've got the app in the "autostart", generally a bit annoying I need to minimize or close the window directly after start, but it is as it is. The fan settings work OK just as long as they are going up - when they are turning back to the state where the fan shoud not spin at all, I've got the noisy, low-speed fan movement that should not occur with such settings at all.

I should move this one to another thread - sorry for hijack

mhm, if i have the time i might take a look at the binary again to see if i spot a bug -- then maybe i can patch the windows binary to fix it .

however i'm not sure it's worth putting the time into the proprietary utility since i reproduced its functionality... it should be fairly easy to port the code to windows.

[jaspen-meyer]

Code: Select all

void unknown_communication(void) {
  printf("Sending unknown commands\n");
  set_0x6C(0x33);
  outb(0x06, 0x68);
}

What do we know about this function?

[bitonic]

Code: Select all

void unknown_communication(void) {
  printf("Sending unknown commands\n");
  set_0x6C(0x33);
  outb(0x06, 0x68);
}

What do we know about this function?

nothing, i just replicated what the windows app does. for some parts of the code i can take i guess, but i have no clue what this does, hence the name.

[vavet]

did anyone managed to come up with a table that eliminates the noise at all?
I have tried different combinations but the noise does come back every time.