|
Intel CPU Powersaving
To all, who are interestet in using the classic way of cpu frequency scaling as it did function for many years now ...
following the KISS principal and NEVER CHANGE A RUNNING SYSTEM if it is not realy better than it was before ...
So since Intel invented it's "intel P-States" witch could not be switched off in kernel anymore some of you
maybe want to change back to the good old days when it did what it should do ...
So i found the "best way for me" to get it back working.
Short history - why i want to change the default behavior - and maybe you too.
I have got three Computers running with different jobs to do ...
One for dayli work witch did run 24/7/365 just to wait fot me to interacting ...
It is protected by a Universal Power Suply -->Dayli
The second one is for keeping my backups , some file systems , some media streams , it is
like a big file server and more and did run 24/7/365 because i always forget to switch it off ...
and for sure it is also protected by the USV(UPS) --> Server
And the third - i rearly just bought to do kernel compiling and testing ...
so - yes it is rearly powerful ...
It did not need to be protected by the USV ... --> Knecht
All these have Intel CPU's because of their flexible power usage - witch i like since i first
have had an intel centrino with intel speedstep technology ...
But since the newer kernels with Intel P-State - these computers did not work as expected anymore !
Here are the technical problems in short described ...
The Dayli running system consume to mutch power all the time ...
The Knecht did not compile fast enough anymore ...
The server did run on higher cpu frequency than it should ...
I did read many documents about the intel p-state - original from intel
I check the command line parameters of the kernel and so on ...
If this is the frequencie spectrum of an intel cpu ...
200 400 800 1000 ...... 2000 ...... 2200 and maybe 2500 (turbo)
it is now (with p-states) just possible to switch from one region to the other with its 'powersave' or 'performance' governor:
|__________________________|_____| performance |
powersave |__________________________________|
And if you want to use the 'old' well known 'ondemand' or 'conservative' governors, you can't reach them anymore ...
| conservative , ondemand |
|_____________________________________________________________|
To get these back - you have to use the following kernel parameter 'intel_pstate=passive'
Yes i do have read the tips all over the net and also here - inside this forum - witch did say other things ...
Following are my Notices witch i do have found in REAL DOCUMENTATIONS :
- Operation Modes
* Active Mode ~ default
indication is cat /sys/devices/system/cpu/cpufreq/policy[0-7]/scaling_driver
=>intel_pstate
There are two P-state selection algorithms provided by intel_pstate
in the active mode: powersave and performance
cat /sys/devices/system/cpu/cpufreq/policy0/scaling_governor
MyDaily: powersave # could just be changed to the oposit
Kernel Config CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE => performance
if not set #CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE => powersave
HWP - Hardware Managed P-states
Knecht cpuinfo: hwp hwp_notify hwp_act_window hwp_epp
* Active Mode With HWP ~ If the processor supports the HWP feature.
* HWP + performance
Energy-Performance Preference (EPP) knob (if supported)
which means that the processors internal P-state selection logic
is expected to focus entirely on performance.
* HWP + powersave
Energy-Performance Bias (EPB) knob (otherwise)
will set the processors Energy Preference to whatever value
it was previously set to via sysfs
* Active Mode Without HWP ~ default if processors not support the HWP feature.
also used by default with the intel_pstate=no_hwp
In this mode intel_pstate registers utilization update callbacks with the
CPU scheduler in order to run a P-state selection algorithm,
either powersave or performance, depending on the scaling_governor policy setting
in sysfs.
The current CPU frequency information to be made available from the scaling_cur_freq
policy attribute in sysfs is periodically updated by those
utilization update callbacks too.
* performance
Without HWP, this P-state selection algorithm is always the same.
selects the maximum P-state it is allowed to use, subject to limits set via sysfs
* powersave
Without HWP, this P-state selection algorithm is similar to implemented
generic schedutil scaling governor
It generally selects P-states proportional to the current CPU utilization.
* Passive Mode
This mode is used if intel_pstate=passive is passed to the kernel command line
(it implies the intel_pstate=no_hwp setting too).
indication is cat /sys/devices/system/cpu/cpufreq/policy[0-7]/scaling_driver
=>intel_cpufreq
Then, the driver behaves like a regular CPUFreq scaling driver.
it is invoked by generic scaling governors when necessary to talk to the hardware
in order to change the P-state of a CPU
it provides the CPUFreq core (and the scaling governors attached to the policy objects)
with accurate information on the maximum and minimum operating frequencies
supported by the hardware (including the so-called "turbo" frequency ranges)
in the passive mode the entire range of available P-states
is exposed by intel_pstate to the CPUFreq core
the scaling_cur_freq information comes from the CPUFreq core and is
the last frequency selected by the current scaling governor for the given policy
--
I hope this did help some people who search for a better powersaving solution with there intel cpu's
Last edited by Wiser Slacker; 08-12-2019 at 05:06 PM.
|
