Powertop
Collection of different Powertop results
Results from running powertop on various machines/operating systems/configurations.
See http://www.lesswatts.org/projects/powertop/known.php for a list of known offenders.
bigbox Intrepid LiveCD (2008-12-17)
- disabled HAL-CD-polling (after seeing that pata_amd was near top of table)
- increased writeback time to 1500 cs (after seeing some related entries in powertop)
- disabled binary nvidia driver (after seeing that it was on top of table)
- powertop running in gnome-terminal; Firefox running with one minimized window with four tabs
- idle desktop, no mouse or keyboard usage
PowerTOP version 1.10 (C) 2007 Intel Corporation < Detailed C-state information is not P-states (frequencies) Wakeups-from-idle per second : 15.5 interval: 10.0s no ACPI power usage estimate available Top causes for wakeups: 25.6% ( 2.0) <kernel core> : clocksource_check_watchdog (clocksource_watc 21.8% ( 1.7) gnome-terminal : schedule_timeout (process_timeout) 14.1% ( 1.1) firefox : futex_wait (hrtimer_wakeup) 7.7% ( 0.6) <kernel core> : neigh_table_init_no_netlink (neigh_periodic_ 6.4% ( 0.5) pulseaudio : schedule_timeout (process_timeout) 3.8% ( 0.3) <kernel module> : neigh_table_init_no_netlink (neigh_periodic_ 2.6% ( 0.2) <kernel core> : neigh_add_timer (neigh_timer_handler) 2.6% ( 0.2) firefox : schedule_timeout (process_timeout) 2.6% ( 0.2) gnome-panel : schedule_timeout (process_timeout) 2.6% ( 0.2) update-notifier : schedule_timeout (process_timeout) 2.6% ( 0.2) nautilus : schedule_timeout (process_timeout) 2.6% ( 0.2) gvfsd-trash : futex_wait (hrtimer_wakeup) 1.3% ( 0.1) hald : schedule_timeout (process_timeout) 1.3% ( 0.1) ssh-agent : schedule_timeout (process_timeout) 1.3% ( 0.1) gconfd-2 : schedule_timeout (process_timeout) 1.3% ( 0.1) gnome-power-man : schedule_timeout (process_timeout) Suggestion: Enable USB autosuspend by pressing the U key or adding usbcore.autosuspend=1 to the kernel command line in the grub config Q - Quit R - Refresh U - Enable USB suspend
bigbox Debian Testing (2008-12-17)
- KDE 3.5.9, my default desktop running
- using binary nvidia driver
- HAL-CD-polling still enabled (machine has two CD drives, ie. two hald-addon-storage processes running)
- only few apps running
- powertop running in konsole
- idle desktop, no mouse or keyboard usage
- results:
- nvidia driver is bad (need to find bug report)
see http://ubuntuforums.org/showthread.php?t=694776: enabling the OnDemandVBlankInterrupts option is supposed to help (didn't try it yet, though)
- HAL-CD-polling is bad (why does it cost so much? Shouldn't it cause maybe 1 wakeup/second?)
kicker: systray contains has some monitoring applets, with 500ms or 1000ms update frequency -> shouldn't these be synchronized? like with g_timer_add_seconds()?
- EMU10K1, eth0: no idea what's it doing... maybe "stray" packets on the net? Hmm... maybe ssh client was already running (need to check with wireshark)
- this is mostly caused by an ssh screen session which displays network throughput every second (when using ssh without screen, wakeups for eth0 are far lower)
- the few other packets (NTP, CUPS announcements, IRC pings, email fetching) don't do much
- aptitude: weird, I don't think it should wake up (strace needed)
konsole: weird as well (maybe caused by aptitude but I don't think this is right (strace needed)
- mysqld: should be idle process, but MySQL is probably not optimized for this (probably best to just keep this process disabled for most time)
- dirmngr: I don't think this is right... this process should stay idle
PowerTOP version 1.10 (C) 2007 Intel Corporation < Detailed C-state information is not P-states (frequencies) Wakeups-from-idle per second : 124.4 interval: 10.0s no ACPI power usage estimate available Top causes for wakeups: 56.9% ( 61.2) <interrupt> : nvidia 16.7% ( 18.0) <interrupt> : ide1 3.0% ( 3.2) kicker : schedule_timeout (process_timeout) 1.9% ( 2.0) <interrupt> : EMU10K1, eth0 1.9% ( 2.0) aptitude : futex_wait (hrtimer_wakeup) 1.9% ( 2.0) <kernel core> : clocksource_register (clocksource_watchdog) 1.9% ( 2.0) konsole : schedule_timeout (process_timeout) 1.4% ( 1.5) mysqld : schedule_timeout (process_timeout) 1.0% ( 1.1) dirmngr : schedule_timeout (process_timeout) 0.9% ( 1.0) kdesktop : schedule_timeout (process_timeout) 0.9% ( 1.0) kwin : schedule_timeout (process_timeout) 0.9% ( 1.0) hald-addon-stor : schedule_timeout (process_timeout) 0.9% ( 1.0) Xorg : nv_start_rc_timer (nv_kern_rc_timer) 0.9% ( 1.0) kwrapper : do_nanosleep (hrtimer_wakeup) 0.9% ( 1.0) apache2 : schedule_timeout (process_timeout) 0.9% ( 1.0) lircd : schedule_timeout (process_timeout) 0.9% ( 1.0) kcpuload : schedule_timeout (process_timeout) 0.9% ( 1.0) ntpd : do_setitimer (it_real_fn) 0.9% ( 1.0) knetload : schedule_timeout (process_timeout) 0.6% ( 0.6) <interrupt> : ide0 0.5% ( 0.5) <kernel module> : neigh_table_init_no_netlink (neigh_periodic_ 0.5% ( 0.5) <kernel core> : schedule_delayed_work_on (delayed_work_timer 0.5% ( 0.5) <kernel core> : neigh_table_init_no_netlink (neigh_periodic_ 0.5% ( 0.5) uml_switch : do_setitimer (it_real_fn) 0.4% ( 0.4) ksensors : schedule_timeout (process_timeout) 0.2% ( 0.2) portmap : inet_twsk_schedule (inet_twdr_hangman) 0.2% ( 0.2) ssh-agent : schedule_timeout (process_timeout) 0.2% ( 0.2) init : schedule_timeout (process_timeout) 0.2% ( 0.2) <kernel core> : page_writeback_init (wb_timer_fn) 0.2% ( 0.2) kded : schedule_timeout (process_timeout) 0.2% ( 0.2) <kernel module> : ide_do_rw_disk (ledtrig_ide_timerfunc) 0.1% ( 0.1) sh : start_this_handle (commit_timeout) 0.1% ( 0.1) nmbd : schedule_timeout (process_timeout) 0.1% ( 0.1) klauncher : schedule_timeout (process_timeout)
- nvidia driver is bad (need to find bug report)
EeePc 701 (4G) Hardy (2008-12-22)
- Desktop Effects enabled
- HAL polls /dev/sdb (card reader?)
Wifi connection active, NetworkManager running, four Wifi APs in range (using madwifi driver 0.3862-1)
- powertop is running in gnome-terminal
- only gnome-terminal and xterm running
- idle desktop, no mouse or keyboard usage
- results:
Intel driver causes lots of wakeups in 3D mode (due to Desktop Effects); see http://www.lesswatts.org/projects/powertop/known.php#intelgfx
- Wifi causes lots of wakeups; either because of constant scanning (by NM) or due to driver bug
- multiload-applet is the CPU load monitor in panel (500ms refresh rate)
the acpi interrupts are maybe caused by HAL polling for battery state; see http://bugzilla.kernel.org/show_bug.cgi?id=9263
- C3 state is never reached
- could be caused by drivers preventing C3; suspects: madwifi, snd-hda-intel
- rough list of actions to improve power usage:
- disable Desktop Effects
- disable Wifi (in nm-applet menu)
- disable HAL polling the card reader
- remove CPU monitor applet
- increase writeback to 1500cs (this decreases the libata wakeups)
- rmmod uvcvideo module (this removes the ehci_work line from powertop output); or better: enable some powersaving feature for this module
actually it seems to suffice to run "echo auto > /sys/bus/usb/devices/5-8/power/level" to enable powersaving mode for the camera (the uvcvideo driver must be loaded before that)
disable snd-hda-intel module which prevents C3 state; or better: enable powersaving for this module - see http://forum.eeeuser.com/viewtopic.php?id=31898
- better: upgrade BIOS to version 1101 (0910 is not enough), and load snd-hda-intel with "power_save=10 power_save_controller=Y" options
PowerTOP version 1.9 (C) 2007 Intel Corporation Cn Verweildauer P-States (Frequenzen) C0 (Prozessor läuft) ( 1,3%) C1 0,0ms ( 0,0%) C2 16,7ms (98,7%) C3 0,0ms ( 0,0%) Aufwachen pro Sekunde : 61,9 Intervall: 15,0s Keine ACPI Stromverbrauch-Schätzung verfügbar Häufigste Ursachen für das Aufwachen: 39,9% ( 36,1) <interrupt> : uhci_hcd:usb4, HDA Intel, i915@pci:0000:00:0 11,0% ( 10,0) <interrupt> : uhci_hcd:usb3, wifi0 11,0% ( 10,0) <kernel module> : ehci_work (ehci_watchdog) 9,4% ( 8,5) <interrupt> : uhci_hcd:usb1, ehci_hcd:usb5 6,6% ( 6,0) USB Gerät 5-5 : UB6225 (ENE) 4,6% ( 4,2) compiz.real : schedule_timeout (process_timeout) 2,2% ( 2,0) multiload-apple : schedule_timeout (process_timeout) 1,8% ( 1,7) gnome-terminal : schedule_timeout (process_timeout) 1,5% ( 1,4) wpa_supplicant : schedule_timeout (process_timeout) 1,2% ( 1,1) hald-addon-stor : schedule_timeout (process_timeout) 1,1% ( 1,0) <interrupt> : acpi 1,1% ( 1,0) dhcdbd : schedule_timeout (process_timeout) 1,1% ( 1,0) <kernel core> : ath_newstate (ath_calibrate) 1,1% ( 1,0) Xorg : schedule_timeout (process_timeout) Vorschlag: Hindern sie mit: "hal-disable-polling --device /dev/cdrom" hal daran das CD-Laufwerk abzufragen. hal startet nach dem Einlegen einer CD autom. einen Dateimanager. Die Abfragen verhindern den SATA Energiesparmodus. Q - Beenden R - Aktualisieren K - kille hald-addon-storage
Toshiba L40-14N Gutsy (2008-12-23)
- no Desktop Effects
- using "intel" graphics driver
- HAL polls CD-ROM drive
Wifi connection active, NetworkManager (0.6.5) running, two Wifi APs in range (using ndiswrapper driver)
- powertop is running in gnome-terminal
- only gnome-terminal running
- idle desktop, no mouse or keyboard usage
PowerTOP version 1.8 (C) 2007 Intel Corporation Cn Verweildauer P-States (Frequenzen) C0 (Prozessor läuft) ( 0,5%) 1467 MHz 1,2% C1 0,0ms ( 0,0%) 1067 MHz 0,0% C2 13,8ms (18,8%) 800 MHz 98,8% C3 13,1ms (80,7%) Aufwachen pro Sekunde : 75,0 Intervall: 15,0s Keine ACPI Stromverbrauch-Schätzung verfügbar Häufigste Ursachen für das Aufwachen: 21,4% ( 15,9) <kernel core> : ehci_work (ehci_watchdog) 15,5% ( 11,5) S20powernowd : queue_delayed_work_on (delayed_work_timer_fn 14,6% ( 10,9) <interrupt> : uhci_hcd:usb5, ehci_hcd:usb6 13,8% ( 10,3) <interrupt> : libata 5,7% ( 4,3) wrapndis_wq/0 : wrap_set_timer (timer_proc) 5,3% ( 3,9) <kernel module> : usb_hcd_poll_rh_status (rh_timer_func) 2,9% ( 2,1) Xorg : do_setitimer (it_real_fn) 2,7% ( 2,0) <kernel core> : queue_delayed_work_on (delayed_work_timer_fn 2,2% ( 1,7) gnome-terminal : schedule_timeout (process_timeout) 2,0% ( 1,5) wpa_supplicant : schedule_timeout (process_timeout) 1,8% ( 1,3) <interrupt> : acpi 1,3% ( 1,0) dhcdbd : schedule_timeout (process_timeout) 1,3% ( 1,0) nm-applet : schedule_timeout (process_timeout) 1,3% ( 1,0) <kernel module> : wrap_set_timer (timer_proc) 1,3% ( 1,0) Xorg : schedule_timeout (process_timeout) 0,8% ( 0,6) NetworkManager : schedule_timeout (process_timeout) 0,8% ( 0,6) <kernel core> : neigh_table_init_no_netlink (neigh_periodic_ 0,6% ( 0,5) hald-addon-stor : schedule_timeout (process_timeout) 0,6% ( 0,5) update-notifier : schedule_timeout (process_timeout) 0,6% ( 0,5) <interrupt> : extra timer interrupt 0,5% ( 0,4) gnome-panel : schedule_timeout (process_timeout) 0,4% ( 0,3) gnome-power-man : schedule_timeout (process_timeout) Vorschlag: Aktivieren Sie "USB autosuspend" durch drücken der U-Taste oder durch anhängen der Bootoption "usbcore autosuspend=1" an die Kernel- kommandozeile oder in der GRUB Konfiguration Q - Beenden R - Aktualisieren U - USB suspend einschalten
Fujitsu-Siemens Amilo ???, Intrepid LiveCD (2008-12-28)
- disabled HAL-CD-polling
- increased writeback time to 1500 cs
- powertop running in gnome-terminal; Firefox running with one minimized window with one tab; one Nautilus window open
disabled Wifi in NetworkManager (Netgear MA111 adapter attached via USB); wired Ethernet connected
- idle desktop, no mouse or keyboard usage
- results:
- no HPET enabled (probably reduces power saving)
- power measure device shows ~30W power usage (dimming display saves ~2W)
PowerTOP version 1.10 (C) 2007 Intel Corporation Cn Verweildauer P-States (Frequenzen) C0 (Prozessor läuft) ( 0,5%) 1128 MHz 0,0% polling 0,0ms ( 0,0%) 995 MHz 0,0% C1 3,1ms ( 0,2%) 862 MHz 0,0% C2 21,8ms (99,3%) 663 MHz 100,0% Aufwachen pro Sekunde : 46,1 Intervall: 15,0s Keine ACPI Stromverbrauch-Schätzung verfügbar Häufigste Ursachen für das Aufwachen: 34,5% ( 11,1) <interrupt> : extra timer interrupt 25,4% ( 8,1) <kernel module> : usb_hcd_poll_rh_status (rh_timer_func) 9,1% ( 2,9) <interrupt> : uhci_hcd:usb1, eth0 6,7% ( 2,1) <interrupt> : acpi 5,2% ( 1,7) gnome-terminal : schedule_timeout (process_timeout) 3,1% ( 1,0) cpufreq-applet : schedule_timeout (process_timeout) 2,7% ( 0,9) /sys/bus/usb/devices/1-2 2,3% ( 0,7) <kernel core> : uhci_scan_schedule (uhci_fsbr_timeout) 2,1% ( 0,7) <kernel core> : neigh_table_init_no_netlink (neigh_periodic_ 1,7% ( 0,5) firefox : futex_wait (hrtimer_wakeup) 1,5% ( 0,5) <kernel core> : __netdev_watchdog_up (dev_watchdog) 1,0% ( 0,3) gnome-power-man : schedule_timeout (process_timeout) 0,8% ( 0,3) gnome-panel : schedule_timeout (process_timeout) 0,6% ( 0,2) update-notifier : schedule_timeout (process_timeout) 0,6% ( 0,2) nautilus : schedule_timeout (process_timeout) 0,6% ( 0,2) gvfsd-trash : futex_wait (hrtimer_wakeup) 0,6% ( 0,2) <kernel module> : neigh_table_init_no_netlink (neigh_periodic_ 0,4% ( 0,1) ssh-agent : schedule_timeout (process_timeout) 0,4% ( 0,1) NetworkManager : schedule_timeout (process_timeout) Vorschlag: Aktivieren Sie HPET (Multimedia Timer) in Ihrem BIOS oder benutzen Sie den Kernel-Patch um HPET zu erzwingen. HPET Unterstützung erlaubt es dem Linux Kernel längere Ruhepausen einzulegen. Q - Beenden R - Aktualisieren
OGerlich/Powertop (last edited 2008-12-28 15:50:17 by p548C5F4C)