You are not logged in.
Pages: 1
Hello,
I'm trying to optimize my power consumption on a freshed arch install on hp 830 G5 laptop.
Unfortunately, I cannot trust powertop... it reports these results :
The battery reports a discharge rate of 132 W
The energy consumed was 0.00 J
The estimated remaining time is 0 hours, 13 minutes
Summary: 1271,6 wakeups/second, 0,0 GPU ops/seconds, 0,0 VFS ops/sec and 39,3% CPU use
Power est. Usage Events/s Category Description
109 W 0,0 pkts/s Device Network interface: enp0s31f6 (e1000e)
15.2 W 25,2% Device Display backlight
8.49 W 11,9 ms/s 633,9 Process [PID 356] [irq/138-iwlwifi]
6.00 W 100,0% Device Radio device: btusb
4.63 W 5,0 ms/s 345,8 Timer tick_sched_timer
710 mW 0,9 ms/s 53,0 kWork iwl_pcie_rx_allocator_work
400 mW 207,7 ms/s 29,9 Process [PID 3061] /usr/lib/thunderbird/thunderbird
245 mW 226,2 µs/s 18,3 Process [PID 17] [rcu_preempt]
219 mW 475,2 µs/s 16,4 Interrupt [134] i915
219 mW 0,0 µs/s 16,4 kWork handle_update
207 mW 161,3 µs/s 15,4 kWork engine_retire
181 mW 40,3 ms/s 13,5 Interrupt [3] net_rx(softirq)
155 mW 402,1 µs/s 11,6 Interrupt [0] HI_SOFTIRQ
...
so for it, ethernet card consumes 109W... whereas card is disabled :
remi@hp-zoo ~ :( $ ip link show
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
2: wlp1s0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP mode DORMANT group default qlen 1000
link/ether 3c:6a:a7:8d:f4:05 brd ff:ff:ff:ff:ff:ff
3: enp0s31f6: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc fq_codel state DOWN mode DEFAULT group default qlen 1000
link/ether f4:39:09:82:19:aa brd ff:ff:ff:ff:ff:ff
I did powertop calibration. Here are the prior logs when starting powertop :
powertop
[sudo] Mot de passe de remi :
modprobe cpufreq_stats failed
Loaded 750 prior measurements
RAPL device for cpu 0
RAPL Using PowerCap Sysfs : Domain Mask d
RAPL device for cpu 0
RAPL Using PowerCap Sysfs : Domain Mask d
Devfreq not enabled
glob returned GLOB_ABORTED
Leaving PowerTOP
And here is a report about battery
native-path: BAT0
vendor: Hewlett-Packard
model: Primary
serial: 26309 2018/06/12
power supply: yes
updated: ven. 22 sept. 2023 22:22:42 (13 seconds ago)
has history: yes
has statistics: yes
battery
present: yes
rechargeable: yes
state: discharging
warning-level: none
energy: 28,4939 Wh
energy-empty: 0 Wh
energy-full: 40,0092 Wh
energy-full-design: 40,0092 Wh
energy-rate: 6,91986 W
voltage: 11,52 V
charge-cycles: N/A
time to empty: 4,1 hours
percentage: 71%
capacity: 100%
technology: lithium-ion
icon-name: 'battery-full-symbolic'
History (rate):
1695414162 6,920 discharging
1695414132 7,182 discharging
1695414109 10,210 discharging
1695414106 11,081 discharging
1695414076 5,539 discharging
I've read some user had overestimated power consumption... But for my case it's so huge. Calibration does not solve anything.
So I'm stuck with powertop.
How can I solve this issue?
Offline
The card isn't "disabled", there's just no carrier.
Google has this for e1000e in the gazillios, "modprobe -r e1000e" and monitor the impact of that.
Online
Now it's display backlight that consumes more than 100W. What rubbish!
he battery reports a discharge rate of 129 W
The energy consumed was 2.58 kJ
The estimated remaining time is 0 hours, 10 minutes
Summary: 729,6 wakeups/second, 0,0 GPU ops/seconds, 0,0 VFS ops/sec and 14,7% CPU use
Power est. Usage Events/s Category Description
115 W 25,2% Device Display backlight
12.6 W 100,0% Device Radio device: btusb
3.47 W 1,0 ms/s 205,8 Interrupt [17] idma64.1
3.40 W 3,7 ms/s 201,8 Timer tick_sched_timer
1.02 W 4,5 ms/s 60,2 Process [PID 333] [irq/137-ALP0018]
829 mW 11,2 ms/s 49,1 Process [PID 3630] java -jar /opt/muon/muon.jar
529 mW 16,7 ms/s 31,3 Interrupt [14] INT344B:00
353 mW 269,2 µs/s 20,9 kWork engine_retire
329 mW 0,8 ms/s 19,5 Interrupt [0] HI_SOFTIRQ
245 mW 1,0 ms/s 14,5 Process [PID 3600] java -jar /opt/muon/muon.jar
200 mW 678,1 µs/s 11,9 Interrupt [134] i915
173 mW 63,0 µs/s 10,3 Timer intel_uncore_fw_release_timer
171 mW 0,9 ms/s 10,1 Process [PID 3611] java -jar /opt/muon/muon.jar
139 mW 113,2 µs/s 8,2 kWork intel_atomic_cleanup_work
136 mW 1,8 ms/s 8,1 kWork intel_atomic_commit_work
136 mW 6,3 ms/s 8,0 Process [PID 644] xfwm4 --display :0.0 --sm-client-id 2dbdb0cad-8e21-43ce-aa78-be24f9e9c497
113 mW 279,4 µs/s 6,7 kWork __intel_wakeref_put_work
Offline
Did you re-calibrate?
Online
Yes, with following result :
Score: 13,8 (18216586,8)
Guess: 128,9
Actual: 142,0
----------------------------------
Leaving PowerTOP
Offline
I was just having a similar issue where backlight was reporting 10W even at 5% backlight level. ran rm -r /var/cache/powertop in elevated permissions, rebooted the computer, and re-ran powertop --calibrate and it seems to, for now, not be giving me strange result. I'd recommend not touching the computer at ALL until powertop shows its usage tui after running calibrate, just leave it alone completely.
Offline
I spent half of my career I spent in modeling. So often I cringed inside, when software had to be given to laymen. That rarely ends well.
PowerTOP doesn’t measure energy consumption. Power is estimated from a model. As with all maths put in an auto mode and given out to the general public, some trouble arise.
While rarely an issue with solutions honestly developed by professional researchers in a project dedicated to the given subject, with models added to other software as an extra feature it’s often the case they suck. PowetTOP tries to squeeze out much values from little data, suggesting the model may be weak and fragile. No idea, how much research and validation was put into this either. Take the results with a grain of salt.⁽¹⁾
This is not a magical black box; this is maths. Using a calculator must not replace understanding of the underlying model. PowerTOP misses a plain documentation for methods it uses and doesn’t discuss their quirks and shortcomings. The user has an excuse for not knowing it. But still some basic precautions have to be observed. Take these into account.
GIGO applies. If a model is built from inadequate or poor data, the result will not be better. Ensure possibly noiseless environment during calibration. The computer should be running for some time already and there should be no programs, which could wake up and cause even slight change in activity.
____
⁽¹⁾ Do not take this against the authors. They do a great job with the primary part of the program and I believe they have best intentions regarding this feature. It’s just a general problem with the approach, which the user must take into account.
Sometimes I seem a bit harsh — don’t get offended too easily!
Offline
Pages: 1