Original distribution site : https://github.com/shibatch/cputemp
Some modern CPUs can reach temperatures near 100 degrees Celsius even when used at stock clock frequency without overclocking. However, it is indeed scary to use a CPU regularly near Tjunction. If possible, I would like to avoid the CPU fan to hum and the high temperature exhaust to spout out from the back of my PC.
There are various theories, but it is said that the safe temperature for regular use of CPUs is about 80 degrees Celsius. If we want to use our CPUs below this temperature, the clock frequency must be lowered. But how much the frequency should be set to keep the temperature below 80 degrees depends on the ambient temperature. If a CPU has many cores, the CPU temperature depends not only on the clock frequency but also on the usage of each core. More to the point, even when all cores are running at 100% load, the CPU temperature can vary considerably depending on what kind of code is being executed. Since the CPU cooler is a piece of metal and has a good amount of heat capacity, the CPU temperature will not exceed 80 degrees if the time of running at a relatively high clock frequency is short.
This tool monitors CPU temperature and dynamically changes the clock frequency to keep the CPU below the specified temperature on a Linux computer. This is basically a simple alternative to thermald, but this tool controls the CPU temperature more aggressively than thermald.
-
Check out the source code from our GitHub repository :
git clone https://github.com/shibatch/cputemp -
Run make :
cd cputemp && make
cputemp [<options>]
--sensor <sensor name>
Specifies the name of sensor that gives the CPU temperature.
--period <seconds>
Specifies the interval at which temperature is checked and CPU frequency is controlled.
--temp <target temperature>
Specify target CPU temperature.
--daemon <pid file name>
Start this tool as a daemon. If this tool is already running as a daemon, it will be restarted with the new setting.
--kill-daemon <pid file name>
Kill already running daemon.
--verbose
Turn on verbose mode.
This tool monitors temperature obtained by the sensor specified by sensor ID every specified period, and control the CPU clock frequency so that the temperature is kept below the specified target temperature.
The CPU clock frequency is controlled by changing the value of
/sys/devices/system/cpu/cpu*/cpufreq/scaling_max_freq. This means
that this tool requires the cpufreq driver to honor the value of
scaling_max_freq in order to work correctly. Some CPUs need kernel
parameters to be set appropriately for this. For example, you may
need to add amd_pstate=passive to the kernel parameters to use
this tool with AMD CPUs.
To use this tool, you first need to check which sensor gives the CPU temperature. You can see the list of available sensor IDs by executing the command without an argument, like the following.
$ ./cputemp
Usage : ./cputemp [<options>]
This utility controls the CPU frequency to make its temperature close to the target
Options :
--sensor <sensor name> Specify sensor name
--period <seconds> Specify period
--temp <target temperature> Specify target CPU temperature
--daemon <pid file name> Daemonize
--kill-daemon <pid file name> Kill already running daemon
--verbose Turn on verbose mode
Available sensors : nvme (41.85 C), k10temp (58.75 C), mt7921_phy0 (50 C), amdgpu (52 C),
$
In the above example, there are 4 sensors, and the sensor ID that gives the CPU temperature is k10temp.
Then, you can specify this sensor ID and start the command as
root. You can see the verbose output of how the temperature is
controlled by --verbose option.
$ sudo ./cputemp --sensor k10temp --temp 80 --verbose
Sensor file name : /sys/class/hwmon/hwmon1/temp1_input
Max freq : 5881 MHz
Min freq : 400 MHz
Cur freq : 5039.77 MHz
CPU freq = 476.98 MHz, scaling_max_freq = 5881 MHz, CPU temp = 64.625 C, target temp = 80 C
CPU freq = 674.05 MHz, scaling_max_freq = 5881 MHz, CPU temp = 64.5 C, target temp = 80 C
CPU freq = 564.43 MHz, scaling_max_freq = 5881 MHz, CPU temp = 64.375 C, target temp = 80 C
CPU freq = 674.05 MHz, scaling_max_freq = 5881 MHz, CPU temp = 64.25 C, target temp = 80 C
CPU freq = 3524.17 MHz, scaling_max_freq = 5881 MHz, CPU temp = 64.25 C, target temp = 80 C
^C
$
You should be able to observe that the CPU frequency is lowered and
the CPU temperature is kept below the specified temperature by
executing some task in another terminal. If you are satisfied with the
results, you can copy the executable to /usr/local/bin and start it in
/etc/rc.local. In this case, --daemon option can be specified to
start this tool as a daemon.
$ cat /etc/rc.local
#!/bin/sh
/usr/local/bin/cputemp --sensor k10temp --temp 80 --daemon /var/run/cputemp.pid
$
Instead of starting cputemp from /etc/rc.local, you can also start it using systemd.
To do this, edit cputemp.service and copy it to the
/etc/systemd/system directory :
sudo cp cputemp.service /etc/systemd/system/
Next, reload the service file :
sudo systemctl daemon-reload
Then, start cputemp service :
sudo systemctl start cputemp.service
To enable cputemp service on every reboot :
sudo systemctl enable cputemp.service
Typically, there is no need to make any special changes to the settings for thermald or cpufreq in order to use cputemp.
The software is distributed under the Zero-Clause BSD, which means that this software is in public domain.
The fact that this software is released under an open source license only means that you can use the current version of the software for free. If you want this software to be maintained, you need to financially support the project. Please see CODE_OF_CONDUCT.md.
Copyright Naoki Shibata 2024.