- Intuitive YAML based configuration
- Massive range of supported devices
- lm-sensors (hwmon) based sensors and fans
- File based fan/sensor for control/measurement of custom devices
- Command based fan/sensor
- Per fan user-defined speed curves
- Fully customizable and composable curve definitions
- Works after resume from suspend
- Stable device paths after reboot
- Automatic analysis of fan properties, like:
- RPM curve
- minimum and maximum PWM
- Error notifications
- Prometheus exporter
- (optional) REST Api
fan2go is first and foremost a daemon that runs in the background. However, it also provides a small set of CLI commands (see CLI Commands) as well as an optional local HTTP API to allow external programs to interact with it. This API can be used to create UI clients or other tools.
fan2go-tui - (Official) Terminal UI for fan2go.
fan2go relies on lm-sensors to get both temperature and RPM sensor readings, as well as PWM controls, so you will have to set it up first.
yay -S fan2go-git
Community Maintained Packages
- Nix with Flakes:
nix profile install nixpkgs#fan2go
- Nix stable:
nix-env -f '<nixpkgs>' -iA fan2go
Download the latest release from GitHub:
# Install dependencies
sudo pacman -S libnotify
curl -L -o fan2go https://github.com/markusressel/fan2go/releases/latest/download/fan2go-linux-amd64
chmod +x fan2go
sudo cp ./fan2go /usr/bin/fan2go
fan2go -h
Or compile yourself:
git clone https://github.com/markusressel/fan2go.git
cd fan2go
make build
sudo cp ./bin/fan2go /usr/bin/fan2go
sudo chmod ug+x /usr/bin/fan2go
Then configure fan2go by creating a YAML configuration file in one of the following locations:
/etc/fan2go/fan2go.yaml
(recommended)/root/.fan2go/fan2go.yaml
./fan2go.yaml
sudo mkdir /etc/fan2go
sudo nano /etc/fan2go/fan2go.yaml
The most important configuration options you need to define are the fans:
, sensors:
and curves:
sections.
Under fans:
you need to define a list of fan devices that you want to control using fan2go. To detect fans on your
system run fan2go detect
, which will print a list of devices exposed by the hwmon filesystem backend:
> fan2go detect
nct6798
Fans Index Channel Label RPM PWM Auto
1 1 hwmon4/fan1 0 153 false
2 2 hwmon4/fan2 1223 104 false
3 3 hwmon4/fan3 677 107 false
Sensors Index Label Value
1 SYSTIN 41000
2 CPUTIN 64000
amdgpu-pci-0031
Fans Index Channel Label RPM PWM Auto
1 1 hwmon8/fan1 561 43 false
Sensors Index Label Value
1 edge 58000
2 junction 61000
3 mem 56000
The fan index is based on device enumeration and is not stable for a given fan if hardware configuration changes. The Linux kernel hwmon channel is a better identifier for configuration as it is largely based on the fan headers in use.
Fan RPM, PWM, and temperature sensors are independent and Linux does not associate them automatically. A given PWM may control more than one fan, and a fan may not be under the control of a PWM. By default, fan2go guesses and sets the pwm channel number for a given fan to the fan's RPM sensor channel. You can override this in the config.
To use detected devices in your configuration, use the hwmon
fan type:
# A list of fans to control
fans:
# A user defined ID.
# Used for logging only
- id: cpu
# The type of fan configuration, one of: hwmon | file
hwmon:
# A regex matching a controller platform displayed by `fan2go detect`, f.ex.:
# "nouveau", "coretemp", "it8620", "corsaircpro-*" etc.
platform: nct6798
# The channel of this fan's RPM sensor as displayed by `fan2go detect`
rpmChannel: 1
# The pwm channel that controls this fan; fan2go defaults to same channel number as fan RPM
pwmChannel: 1
# Indicates whether this fan should never stop rotating, regardless of
# how low the curve value is
neverStop: true
# The curve ID that should be used to determine the
# speed of this fan
curve: cpu_curve
fans:
- id: file_fan
file:
# Path to a file to get/set the PWM target for this fan
path: /tmp/file_fan
# Path to a file to read the current RPM value of this fan
rpmPath: /tmp/file_fan_rpm
> cat /tmp/file_fan
255
> cat /tmp/file_fan_rpm
3421
Please also make sure to read the section about considerations for using the cmd sensor/fan.
fans:
- id: cmd_fan
cmd:
# Command to apply a new PWM value (0..255)
# Use "%pwm%" to specify where the target pwm value should be used withing the arguments
setPwm:
exec: /usr/bin/some-program
args: [ "--set", "%pwm%" ]
# Command to retrieve the current PWM value (0..255)
getPwm:
exec: /usr/bin/nvidia-settings
args: [ "-a", "someargument" ]
# (optional) Command to retrieve the current RPM value
getRpm:
exec: /usr/bin/nvidia-settings
args: [ "-a", "someargument" ]
If the automatic fan curve analysis doesn't provide a good enough estimation for how the fan behaves, you can use the following configuration options (per fan definition) to correct it:
fans:
- id: ...
...
# (Optional) Override for the lowest PWM value at which the
# fan is able to maintain rotation if it was spinning previously.
minPwm: 30
# (Optional) Override for the lowest PWM value at which the
# fan will still be able to start rotating.
# Note: Settings this to a value that is too small
# may damage your fans. Use at your own risk!
startPwm: 30
# (Optional) Override for the highest PWM value which still yields
# an increased rotational speed compared to lower values.
# Note: you can also use this to limit the max speed of a fan.
maxPwm: 255
# (Optional) Override for the PWM map used by fan2go for
# mapping the expected [0..255] value range to values actually supported by this fan.
# This can be used to compensate for fans with a very limited set of supported values
# (f.ex. off, low, high). If not set manually, fan2go will try to compute this mapping
# automatically during fan initialization. This process is not perfect though and may
# result in suboptimal fan control.
# Note: The mapping must be strictly monotonically increasing and its Key-Set should cover the full
# range of values from 0 to 255. If keys are missing, fan2go will select a key that most closely
# matches the required target value during operation.
pwmMap:
0: 0
64: 128
192: 255
Under sensors:
you need to define a list of temperature sensor devices that you want to monitor and use to adjust
fanspeeds. Like with fans, you can find usable devices using fan2go detect
.
# A list of sensors to monitor
sensors:
# A user defined ID, which is used to reference
# a sensor in a curve configuration (see below)
- id: cpu_package
# The type of sensor configuration, one of: hwmon | file | cmd
hwmon:
# A regex matching a controller platform displayed by `fan2go detect`, f.ex.:
# "coretemp", "it8620", "corsaircpro-*" etc.
platform: coretemp
# The index of this sensor as displayed by `fan2go detect`
index: 1
sensors:
- id: file_sensor
file:
# Path to the file containing sensor values
path: /tmp/file_sensor
The file contains a value in milli-units, like f.ex. milli-degrees.
> cat /tmp/file_sensor
10000
Please also make sure to read the section about considerations for using the cmd sensor/fan.
Just like the file
sensor, the command must output the sensor value in milli-units,
like f.ex. milli-degrees.
sensors:
- id: cmd_fan
cmd:
# Path to the executable to run to retrieve the current sensor value
exec: /usr/bin/bash
# (optional) arguments to pass to the executable
args: [ '/home/markus/myscript.sh' ]
Under curves:
you need to define a list of fan speed curves, which represent the speed of a fan based on one or more
temperature sensors.
To create a simple, linear speed curve, use a curve of type linear
.
This curve type can be used with a min/max sensor value, where the min temp will result in a curve value of 0
and the
max temp will result in a curve value of 255
:
curves:
- id: cpu_curve
# The type of the curve, one of: linear | function
linear:
# The sensor ID to use as a temperature input
sensor: cpu_package
# Sensor input value at which the curve is at minimum speed
min: 40
# Sensor input value at which the curve is at maximum speed
max: 80
You can also define the curve in multiple, linear sections using the steps
parameter:
curves:
- id: cpu_curve
# The type of the curve
linear:
# The sensor ID to use as a temperature input
sensor: cpu_package
# Steps to define a section-wise defined speed curve function.
steps:
# Sensor value -> Speed (in pwm)
- 40: 0
- 50: 50
- 80: 255
If you want to get your hands dirty and use a PID based curve, you can use pid
:
curves:
- id: pid_curve
pid:
sensor: cpu_package
setPoint: 60
p: -0.05
i: -0.005
d: -0.005
Unlike the other curve types, this one does not use the average of the sensor data to calculate its value, which allows you to create a completely custom behaviour. Keep in mind though that the fan controller may also be PID based which could also affect how the curve is applied to the fan.
To create more complex curves you can combine exising curves using a curve of type function
:
curves:
- id: case_avg_curve
function:
# Type of aggregation function to use, one of: minimum | maximum | average | delta | sum | difference
type: average
# A list of curve IDs to use
curves:
- cpu_curve
- mainboard_curve
- ssd_curve
An example configuration file including more detailed documentation can be found in fan2go.yaml.
To check whether your configuration is correct before actually running fan2go you can use:
> sudo fan2go config validate
INFO Using configuration file at: /etc/fan2go/fan2go.yaml
SUCCESS Config looks good! :)
or to validate a specific config file:
> fan2go -c "./my_config.yaml" config validate
INFO Using configuration file at: ./my_config.yaml
WARNING Unused curve configuration: m2_first_ssd_curve
ERROR Validation failed: Curve m2_ssd_curve: no curve definition with id 'm2_first_ssd_curve123' found
fan2go supports using external executables for use as both sensor input, as well as fan output (and rpm input). There are some considerations you should take into account before using this feature though:
Since fan2go requires root permissions to interact with lm-sensors, executables run by fan2go are also executed as root. To prevent some malicious actor from taking advantage of this fan2go will only allow the execution of files that only allow the root user (UID 0) to modify the file.
Running external commands repeatedly through fan2go can have unintended side effects. F.ex., on a laptop using hybrid
graphics, running nvidia-settings
can cause the dedicated GPU to wake up, resulting in substantial increase in power
consumption while on battery. Also, fan2go expects to be able to update sensor values with a minimal delay, so using a
long running script or some network call with a long timeout could also cause problems. With great power comes great
responsibility, always remember that :)
After successfully verifying your configuration you can launch fan2go from the CLI and make sure the initial setup is
working as expected. Assuming you put your configuration file in /etc/fan2go/fan2go.yaml
run:
> fan2go help
fan2go is a simple daemon that controls the fans
on your computer based on temperature sensors.
Usage:
fan2go [flags]
fan2go [command]
Available Commands:
completion Generate the autocompletion script for the specified shell
config Configuration related commands
curve Curve related commands
detect Detect fans and sensors
fan Fan related commands
help Help about any command
sensor Sensor related commands
version Print the version number of fan2go
Flags:
-c, --config string config file (default is $HOME/.fan2go.yaml)
-h, --help help for fan2go
--no-color Disable all terminal output coloration
--no-style Disable all terminal output styling
-v, --verbose More verbose output
Use "fan2go [command] --help" for more information about a command.
> sudo fan2go
Alternatively you can specify the path to your configuration file like this:
> sudo fan2go -c /home/markus/my_fan2go_config.yaml
When installing fan2go using a package, it comes with a systemd unit file. To enable it simply run:
sudo systemctl daemon-reload
sudo systemctl enable --now fan2go
# follow logs
journalctl -u fan2go -f
NOTE: If you want to use a config path that differs from the default one, make sure to edit the unit file and point the
-c
flag to the correct path.
Although fan2go is a fan controller daemon at heart, it also provides some handy cli commands to interact with the devices that you have specified within your config.
> fan2go fan --id cpu speed 100
> fan2go fan --id cpu speed
255
> fan2go fan --id cpu rpm
546
> fan2go fan --id cpu mode
No control, 100% all the time (0)
> fan2go fan --id cpu mode auto
Automatic control by integrated hardware (2)
> fan2go sensor --id cpu_package
46000
For each newly configured fan fan2go measures its fan curve and stores it in a db for future reference. You can take a look at this measurement using the following command:
> sudo fan2go fan --id cpu curve
cpu
Min PWM 30
Start PWM 30
Max PWM 255
No fan curve data yet...
> sudo fan2go fan --id in_front curve
nct6798 -> pwm2
Min PWM 0
Start PWM 0
Max PWM 194
1994 ┤ ╭────────────────────────
1900 ┤ ╭──╯
1805 ┤ ╭────╯
1711 ┤ ╭────╯
1616 ┤ ╭────╯
1522 ┤ ╭───╯
1427 ┤ ╭────╯
1333 ┤ ╭────╯
1238 ┤ ╭─────╯
1144 ┤ ╭────╯
1049 ┤ ╭─────╯
955 ┤ ╭─────╯
860 ┤ ╭─────╯
766 ┤ ╭─────╯
671 ┤ ╭─────╯
577 ┼─╯
RPM / PWM
fan2go has a prometheus exporter built in, which you can use to extract data over time. Simply enable it in your configuration and you are good to go:
statistics:
# Whether to enable the prometheus exporter or not
enabled: true
# The port to expose the exporter on
port: 9000
You can then see the metics on http://localhost:9000/metrics while the fan2go daemon is running.
fan2go comes with a built-in REST Api. This API can be used by third party tools to display (and in the future possibly modify) the state of fans, sensors and curves within fan2go.
api:
# Whether to enable the API or not
enabled: false
# The host to listen for connections
host: localhost
# The port to listen for connections
port: 9001
Currently, this API is read-only and only provides REST endpoints. If there is demand for it, this might be expanded to also support realtime communication via websockets.
Endpoint | Type | Description |
---|---|---|
/fan |
GET | Returns a list of all currently configured fans |
/fan/<id> |
GET | Returns the fan with the given id , if it exists |
Endpoint | Type | Description |
---|---|---|
/sensor |
GET | Returns a list of all currently configured sensors |
/sensor/<id> |
GET | Returns the sensor with the given id , if it exists |
Endpoint | Type | Description |
---|---|---|
/curve |
GET | Returns a list of all currently configured curves |
/curve/<id> |
GET | Returns the curve with the given id , if it exists |
fan2go uses gosensors to directly interact with lm-sensors.
To properly control a fan which fan2go has not seen before, its speed curve is analyzed. This means
- spinning down the fans to 0
- slowly ramping up the speed and monitoring RPM changes along the way
Note that this takes approx. 8 1/2 minutes, since we have to wait for the fan speed to settle before taking measurements. Measurements taken during this process will then be used to determine the lowest PWM value at which the fan is still running, as well as the highest PWM value that still yields a change in RPM.
All of this is saved to a local database (path given by the dbPath
config option), so it is only needed once per fan
configuration.
To reduce the risk of runnin the whole system on low fan speeds for such a long period of time, you can force fan2go to
initialize only one fan at a time, using the runFanInitializationInParallel: false
config option.
Some PWM controllers or fans may require more time to react to PWM changes. If fan2go is failing to characterize a fan,
you can try increasing the fan response delay by passing --fan-response-delay <seconds>
to the fan init
command or
by setting fanResponseDelay
in the config. The default value is 2 seconds.
Temperature and RPM sensors are polled continuously at the rate specified by the tempSensorPollingRate
config option.
tempRollingWindowSize
/rpmRollingWindowSize
amount of measurements are always averaged and stored as the average
sensor value.
The speed of a Fan is controlled using a combination of its curve, a control algorithm and the properties of the fan controller itself.
The curve is used as the target value for the control algorithm to reach. The control algorithm then calculates the
next PWM value to apply to the fan to reach this target value. The fan controller then applies this PWM value to the
fan, while respecting constraints like the minimum and maximum PWM values, as well as the neverStop
flag.
A control algorithm is a function that returns the next PWM value to apply based on the target value calculated by the curve. The simplest control algorithm is the direct control algorithm, which simply forwards the target value to the fan.
The simplest control algorithm is the direct control algorithm. It simply forwards the curve value to the fan controller.
fans:
- id: some_fan
...
controlAlgorithm: direct
This control algorithm can also be used to approach the curve value more slowly:
fans:
- id: some_fan
...
controlAlgorithm:
direct:
maxPwmChangePerCycle: 10
The PID control algorithm uses a PID loop to approach the target value. The default configuration is pretty non-aggressive using the following values:
P | I | D |
---|---|---|
0.3 |
0.02 |
0.005 |
If you don't like the default behaviour you can configure your own in the config:
fans:
- id: some_fan
...
controlAlgorithm:
pid:
p: 0.3
i: 0.02
d: 0.005
The loop is advanced at a constant rate, specified by the controllerAdjustmentTickRate
config option, which
defaults to 200ms
.
TL;DR: modprobe drivetemp
While lm-sensors doesn't provide temperature sensors of SATA drives by default, you can use the kernel module
drivetemp
to enable this. See here
If you see this log message while running fan2go, fan2go detected a change of the PWM value for the given fan that was not caused by fan2go itself. This usually means that fan2go is not the only program controlling the fan and something else (like f.ex. the mainboard or another fan control software) is also running and changing the speed of the fan, competing with fan2go. Since fan2go cannot figure out what other software is, you have to investigate this yourself.
Another common reason this message can occur is when the driver of the fan in question does not actually support setting the PWM directly and uses some kind of virtual PWM instead. This has been a problem mostly on AMD graphics cards but is probably not limitied to them. See #64 for more detail.
TL;DR: Skip the initialization and configure your fans manually.
The initialization phase measures the RPM curve of each fan and tries to estimate the minimum and maximum
boundaries. This can take quite a while though and can lead to overheating of components if they are
under load. To avoid this use the minPwm
and maxPwm
fan config options to set the boundaries yourself.
That way the initialization phase will be skipped and the control algorithm will start right away.
See go.mod
fan2go
Copyright (C) 2021 Markus Ressel
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU Affero General Public License as published
by the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU Affero General Public License for more details.
You should have received a copy of the GNU Affero General Public License
along with this program. If not, see <https://www.gnu.org/licenses/>.