CPU Power states

Note

System tuning tools are either about better understanding the system’s performance, or applying such knowledge to improve it. See our common system tuning thoughts for some generally applicable considerations about that.

A modern CPU usually can be in one of many P-states (mostly controlling voltage and frequency) when working as well as C-states (mostly defining how much of the chip elements are turned off) when idle. Controlling these has direct influence to performance, but also power consumption and is therefore considered to be part of Power Management

Depending on manufacturer and chip generation different systems might be active - here we outline some common ones, but actually most others still share a lot of the same aspects.

Below we will outline the direct access via sysfs, but there are tools that can inspect and manipulate such settings like EPP or scaling governors on Ubuntu. Please have a look at The cpupower tool as well.

CPU governors

The kernel provides several CPU governors which can be configured, per core, to optimize for different needs.

Governor	Design philosophy
ondemand	This sets the CPU frequency depending on the current system load. This behavior is usually a good balance between the more extreme options.
conservative	Similar to ondemand, but adapting CPU speed more gracefully rather than jumping to max speed the moment there is any load on the CPU. This behavior is more suitable in a battery-powered environment.
performance	This sets the CPU statically to the highest frequency. This behavior is best to optimize for speed and latency, but might waste power if being under-used.
powersave	Sets the CPU statically to the lowest frequency, essentially locking it to P2. This behavior is suitable to save power without compromises.
userspace	Allows a user-space program to control the CPU frequency.

See the Linux CPUFreq Governors Documentation for a more extensive discussion and explanation of the available Linux CPU governors.

While these governors can be checked and changed directly in sysfs at /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor, the command cpupower which comes with the package linux-tools-common makes this easier by providing a command line interface and providing access to several related values.

Setting CPU Governors

cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor

This file will normally contain one of the following named profiles:

• performance - Prioritize responsiveness and raw performance.

• powersave - Prioritize a reduction in power consumption.

The full list of options is provided in:

cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_governors

The value can be set via sysfs as well:

echo performance | sudo tee /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor

Energy Performance Preference

Energy Performance Preference (EPP) is a CPU power-management feature available on many modern processors (particularly Intel and AMD). It allows the operating system or user to influence the CPU’s balance between performance and power consumption. Instead of forcing a strict maximum performance or minimum power mode, EPP provides a hint to the processor’s hardware power control logic about the desired balance. Consider it to be quite similar to scaling_governors, but partially handled in a hardware/firmware assisted way.

The actual effect depends on the CPU model and firmware, but in general, lower EPP values tell a processor to favor performance while higher values tell it to favor lower power consumption.

Caveats

• EPP is only a preference or hint to the hardware. It does not guarantee that the CPU will behave in a precise or deterministic way.

• EPP value interpretations will often vary between different CPUs.

• Hardware support is required. On some processors, EPP may have no effect even if it appears to be configurable in the OS.

• Some systems may require an update to BIOS settings to allow OS control of power regulation.

EPP values on Ubuntu

On Ubuntu, the EPP for each core on a CPU is stored in /sys/devices/system/cpu/cpu#/cpufreq/energy_performance_preference. To check the current status on cpu0, for example, you can run:

cat /sys/devices/system/cpu/cpu0/cpufreq/energy_performance_preference

This file will normally contain one of the following named profiles:

• performance - Prioritize responsiveness and raw performance.

• balance_performance - Lean toward performance, with lower power consumption.

• balance_power - Lean toward reduced power, with some additional performance.

• power - Prioritize a reduction in power consumption.

• default - Use the default mix of power and performance as recommended by Ubuntu.

The full list of options is provided in /sys/devices/system/cpu/cpu#/cpufreq/energy_performance_available_preferences. To see them, run:

cat /sys/devices/system/cpu/cpu0/cpufreq/energy_performance_available_preferences

which will display something like:

default performance balance_performance balance_power power

For more specificity, a raw numeric value from 0-255 can also be provided in the energy_performance_preference file. In general, a value closer to 0 will prioritize performance and a value closer to 255 will prioritize efficiency.

Setting EPP values

EPP values are managed through each core’s energy_performance_preference file, so the EPP values can be updated manually by overriding the contents of that file. For example, to set cpu0 to the performance profile, run the following:

echo performance | sudo tee /sys/devices/system/cpu/cpu0/cpufreq/energy_performance_preference

On Ubuntu Desktop, GNOME provides a power management menu in settings with options that may modify EPP values.

These settings will normally map to the following EPP profiles:

Power Mode setting	EPP profile
Performance	performance
Balanced	balance_performance
Power Saver	power