View previous topic :: View next topic |
Author |
Message |
Bloot Tux's lil' helper
Joined: 10 Mar 2006 Posts: 99 Location: Barcelona
|
Posted: Fri Mar 27, 2015 12:00 am Post subject: Intel P state always at max freq when idle |
|
|
Hello,
I recently upgraded to a i5-4590 Haswell refresh installed on a Gigabyte Z97P-D3 motherboard, so I activated Intel P state in the kernel. But it always stays at maximum frequency when the system is idle, it only scales down when it's busy which is disconcerting, it behaves the opposite way
My kernel setup
Code: | $ cat /usr/src/linux/.config | grep CPU_FREQ
CONFIG_CPU_FREQ=y
# CONFIG_CPU_FREQ_STAT is not set
CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE=y
# CONFIG_CPU_FREQ_DEFAULT_GOV_USERSPACE is not set
# CONFIG_CPU_FREQ_DEFAULT_GOV_ONDEMAND is not set
# CONFIG_CPU_FREQ_DEFAULT_GOV_CONSERVATIVE is not set
CONFIG_CPU_FREQ_GOV_PERFORMANCE=y
CONFIG_CPU_FREQ_GOV_POWERSAVE=y
# CONFIG_CPU_FREQ_GOV_USERSPACE is not set
# CONFIG_CPU_FREQ_GOV_ONDEMAND is not set
# CONFIG_CPU_FREQ_GOV_CONSERVATIVE is not set |
Code: | $ cat /usr/src/linux/.config | grep PSTATE
CONFIG_X86_INTEL_PSTATE=y |
Code: | $ cat /usr/src/linux/.config | grep NO_HZ
CONFIG_NO_HZ_COMMON=y
CONFIG_NO_HZ_IDLE=y
# CONFIG_NO_HZ_FULL is not set
# CONFIG_NO_HZ is not set
# CONFIG_RCU_FAST_NO_HZ is not set |
Is there something I am missing, or is this driver really buggy? I am turning mad, I've compiled the kernel lots of times with no luck, it always stays at maximum frequency (3700MHz) on idle and only scales down when compiling, surfing the net etc...
Some cpupower info:
Code: | # cpupower frequency-info
analyzing CPU 0:
driver: intel_pstate
CPUs which run at the same hardware frequency: 0
CPUs which need to have their frequency coordinated by software: 0
maximum transition latency: 0.97 ms.
hardware limits: 800 MHz - 3.70 GHz
available cpufreq governors: performance, powersave
current policy: frequency should be within 800 MHz and 3.70 GHz.
The governor "powersave" may decide which speed to use
within this range.
current CPU frequency is 3.69 GHz (asserted by call to hardware).
boost state support:
Supported: yes
Active: yes |
Code: | # cpupower idle-info
CPUidle driver: intel_idle
CPUidle governor: menu
Analyzing CPU 0:
Number of idle states: 6
Available idle states: POLL C1-HSW C1E-HSW C3-HSW C6-HSW C7s-HSW
POLL:
Flags/Description: CPUIDLE CORE POLL IDLE
Latency: 0
Usage: 155
Duration: 148047
C1-HSW:
Flags/Description: MWAIT 0x00
Latency: 2
Usage: 29439
Duration: 3517980
C1E-HSW:
Flags/Description: MWAIT 0x01
Latency: 10
Usage: 13720
Duration: 8908518
C3-HSW:
Flags/Description: MWAIT 0x10
Latency: 33
Usage: 23891
Duration: 23960449
C6-HSW:
Flags/Description: MWAIT 0x20
Latency: 133
Usage: 3355
Duration: 4377164
C7s-HSW:
Flags/Description: MWAIT 0x32
Latency: 166
Usage: 33282
Duration: 88202668 |
Code: | # cpupower monitor
|Nehalem || SandyBridge || Mperf || Idle_Stats
CPU | C3 | C6 | PC3 | PC6 || C7 | PC2 | PC7 || C0 | Cx | Freq || POLL | C1-H | C1E- | C3-H | C6-H | C7s-
0|******|******|******|******||******|******|******||******|******|******|| 0,00| 0,32| 0,12| 0,36| 0,33| 98,49
1|******|******|******|******||******|******|******||******|******|******|| 0,00| 0,00| 0,00| 0,20| 0,00| 103,6
2|******|******|******|******||******|******|******||******|******|******|| 0,00| 0,35| 0,09| 0,21| 0,25| 92,30
3|******|******|******|******||******|******|******||******|******|******|| 0,00| 0,47| 0,01| 0,35| 0,20| 101,9
|
Temperatures are ok though
Code: | $ sensors
coretemp-isa-0000
Adapter: ISA adapter
Physical id 0: +36.0°C (high = +80.0°C, crit = +100.0°C)
Core 0: +32.0°C (high = +80.0°C, crit = +100.0°C)
Core 1: +30.0°C (high = +80.0°C, crit = +100.0°C)
Core 2: +31.0°C (high = +80.0°C, crit = +100.0°C)
Core 3: +33.0°C (high = +80.0°C, crit = +100.0°C) |
I'd really appreciate if someone can shed some ligh, thanks.
Last edited by Bloot on Fri Mar 27, 2015 6:00 pm; edited 2 times in total |
|
Back to top |
|
|
abduct Apprentice
Joined: 19 Mar 2015 Posts: 215
|
Posted: Fri Mar 27, 2015 1:20 am Post subject: |
|
|
In the kernel try setting the "default CPUFreq governor" to "ondemand" so that the system adjusts the processor for you. The option can be found under "Power management and ACPI options -> CPU Frequency Scaling". Also try using the ACPI P-States option as I've read that the Intel P option has a similar bug in which it maxes out the CPU.
Source: I had a similar problem with a laptop and the processor not enabling turbo-boost or scaling automatically. |
|
Back to top |
|
|
Bloot Tux's lil' helper
Joined: 10 Mar 2006 Posts: 99 Location: Barcelona
|
Posted: Fri Mar 27, 2015 9:23 am Post subject: |
|
|
abduct wrote: | In the kernel try setting the "default CPUFreq governor" to "ondemand" so that the system adjusts the processor for you. The option can be found under "Power management and ACPI options -> CPU Frequency Scaling". Also try using the ACPI P-States option as I've read that the Intel P option has a similar bug in which it maxes out the CPU.
Source: I had a similar problem with a laptop and the processor not enabling turbo-boost or scaling automatically. |
Thanks for you reply. I already tried ondemand, in fact I tried so many combinations I could spend the hole thread numering them
Intel themselves don't recommend using any other driver than their own's for the newer chips, so that's why I selected Intel P-State over ACPI P-States.
I think I've made some progress and maybe solved the erratic Intel P-State behaviour, I've disabled acpi processor:
Code: | $ cat /usr/src/linux/.config | grep PROCESSOR
# CONFIG_ACPI_PROCESSOR is not set |
And now intel p state is working as it should, though it never stays permanently at 800MHz (as it did with ondemand and acpi p-states), it's always oscilating and frankly, I don't know if this is the correct behaviour. At least it's not always at full power frequency like before.
Code: | # cpupower frequency-info
analyzing CPU 0:
driver: intel_pstate
CPUs which run at the same hardware frequency: 0
CPUs which need to have their frequency coordinated by software: 0
maximum transition latency: 0.97 ms.
hardware limits: 800 MHz - 3.70 GHz
available cpufreq governors: performance, powersave
current policy: frequency should be within 800 MHz and 3.70 GHz.
The governor "powersave" may decide which speed to use
within this range.
current CPU frequency is 1.19 GHz (asserted by call to hardware).
boost state support:
Supported: yes
Active: yes |
Code: | $ cat /usr/src/linux/.config | grep IDLE
CONFIG_NO_HZ_IDLE=y
CONFIG_GENERIC_SMP_IDLE_THREAD=y
CONFIG_CPU_IDLE=y
CONFIG_CPU_IDLE_GOV_LADDER=y
CONFIG_CPU_IDLE_GOV_MENU=y
# CONFIG_ARCH_NEEDS_CPU_IDLE_COUPLED is not set
# CONFIG_INTEL_IDLE is not set
# CONFIG_I7300_IDLE is not set
# CONFIG_NETFILTER_XT_TARGET_IDLETIMER is not set |
The drawback is that INTEL_IDLE cannot be used, or it will force max cpu frequency again, and thus cpupower is unable to show idle information without intel idle driver. But as long as it works, I don't mind not having access to idle information.
Code: | # cpupower idle-info
CPUidle driver: none
CPUidle governor: menu
Analyzing CPU 0:
CPU 0: No idle states |
Code: | # cpupower monitor
|Nehalem || SandyBridge || Mperf
CPU | C3 | C6 | PC3 | PC6 || C7 | PC2 | PC7 || C0 | Cx | Freq
0|******|******|******|******||******|******|******||******|******|******
1|******|******|******|******||******|******|******||******|******|******
2|******|******|******|******||******|******|******||******|******|******
3|******|******|******|******||******|******|******||******|******|****** |
So I'll mark it as solved, thank you very much and hope it helps someone. |
|
Back to top |
|
|
albright Advocate
Joined: 16 Nov 2003 Posts: 2588 Location: Near Toronto
|
Posted: Fri Mar 27, 2015 11:40 am Post subject: |
|
|
Quote: | And now intel p state is working as it should, though it never stays permanently at 800MHz (as it did with ondemand and acpi p-states), it's always oscilating and frankly, I don't know if this is the correct behaviour |
I think that's right; at least my two machines with p-state are like that _________________ .... there is nothing - absolutely nothing - half so much worth
doing as simply messing about with Linux ...
(apologies to Kenneth Graeme) |
|
Back to top |
|
|
Bloot Tux's lil' helper
Joined: 10 Mar 2006 Posts: 99 Location: Barcelona
|
Posted: Fri Mar 27, 2015 12:11 pm Post subject: |
|
|
albright wrote: | Quote: | And now intel p state is working as it should, though it never stays permanently at 800MHz (as it did with ondemand and acpi p-states), it's always oscilating and frankly, I don't know if this is the correct behaviour |
I think that's right; at least my two machines with p-state are like that |
Thank you very much
Would you mind sharing your kernel config with me? Now powersave works without acpi processor, but can't get past 3500Mhz and this chip should get up to 3700Mhz, at least it did when Intel P-State forced the max frequency on idle. |
|
Back to top |
|
|
Yamakuzure Advocate
Joined: 21 Jun 2006 Posts: 2284 Location: Adendorf, Germany
|
Posted: Fri Mar 27, 2015 1:48 pm Post subject: |
|
|
- You can only use performance and powersave governors, because the p-state driver does not support any other.
If you want other governors, you have to use ACPI and replace it using phc-intel (which is a p-state driver, too).
*but* : phc-intel can not fully power up the cpu. My 2.8 GHz i7 (With turbo up to 3.4 GHz) can not go over 2.6 GHz with phc-intel.
- Therefore I use the intel pstate driver from the kernel.
With performance governor:
Code: | ~ # LC_ALL=C cpupower frequency-info
analyzing CPU 0:
driver: intel_pstate
CPUs which run at the same hardware frequency: 0
CPUs which need to have their frequency coordinated by software: 0
maximum transition latency: 0.97 ms.
hardware limits: 800 MHz - 3.40 GHz
available cpufreq governors: performance, powersave
current policy: frequency should be within 800 MHz and 3.40 GHz.
The governor "performance" may decide which speed to use
within this range.
current CPU frequency is 3.29 GHz (asserted by call to hardware).
boost state support:
Supported: yes
Active: yes | and with powersave governor, AC unplugged: Code: | ~ # LC_ALL=C cpupower frequency-info
analyzing CPU 0:
driver: intel_pstate
CPUs which run at the same hardware frequency: 0
CPUs which need to have their frequency coordinated by software: 0
maximum transition latency: 0.97 ms.
hardware limits: 800 MHz - 3.40 GHz
available cpufreq governors: performance, powersave
current policy: frequency should be within 800 MHz and 3.40 GHz.
The governor "powersave" may decide which speed to use
within this range.
current CPU frequency is 605 MHz (asserted by call to hardware).
boost state support:
Supported: yes
Active: yes |
(Interestingly the tool says that minimum frequency is 800MHz, but it was scaled down to 600MHz. ... conky reports the same, by the way...)
There seems to be something else wrong with your config if you can not use INTEL_IDLE. Or maybe it's the cpu?
Here is my config for reference: Code: | /usr/src/linux # grep -P "^[^#].*(FREQ|IDLE|PSTATE)" .config
CONFIG_NO_HZ_IDLE=y
CONFIG_GENERIC_SMP_IDLE_THREAD=y
CONFIG_CPU_FREQ=y
CONFIG_CPU_FREQ_GOV_COMMON=y
CONFIG_CPU_FREQ_STAT=m
CONFIG_CPU_FREQ_STAT_DETAILS=y
CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE=y
CONFIG_CPU_FREQ_GOV_PERFORMANCE=y
CONFIG_CPU_FREQ_GOV_POWERSAVE=y
CONFIG_CPU_FREQ_GOV_USERSPACE=y
CONFIG_CPU_FREQ_GOV_ONDEMAND=y
CONFIG_CPU_FREQ_GOV_CONSERVATIVE=y
CONFIG_X86_INTEL_PSTATE=y
CONFIG_X86_ACPI_CPUFREQ=m
CONFIG_CPU_IDLE=y
CONFIG_CPU_IDLE_GOV_LADDER=y
CONFIG_CPU_IDLE_GOV_MENU=y
CONFIG_INTEL_IDLE=y | and this is what idle-info comes up with: Code: | ~ # LC_ALL=C cpupower idle-info
CPUidle driver: intel_idle
CPUidle governor: menu
Analyzing CPU 0:
Number of idle states: 6
Available idle states: POLL C1-HSW C1E-HSW C3-HSW C6-HSW C7s-HSW
POLL:
Flags/Description: CPUIDLE CORE POLL IDLE
Latency: 0
Usage: 20346
Duration: 20113894
C1-HSW:
Flags/Description: MWAIT 0x00
Latency: 2
Usage: 7175620
Duration: 2884307784
C1E-HSW:
Flags/Description: MWAIT 0x01
Latency: 10
Usage: 2076175
Duration: 1951980913
C3-HSW:
Flags/Description: MWAIT 0x10
Latency: 33
Usage: 3480621
Duration: 2181767439
C6-HSW:
Flags/Description: MWAIT 0x20
Latency: 133
Usage: 1282988
Duration: 980550922
C7s-HSW:
Flags/Description: MWAIT 0x32
Latency: 166
Usage: 23154988
Duration: 73947612700 | I hope it helps. _________________ Important German:- "Aha" - German reaction to pretend that you are really interested while giving no f*ck.
- "Tja" - German reaction to the apocalypse, nuclear war, an alien invasion or no bread in the house.
|
|
Back to top |
|
|
Bloot Tux's lil' helper
Joined: 10 Mar 2006 Posts: 99 Location: Barcelona
|
Posted: Fri Mar 27, 2015 5:58 pm Post subject: |
|
|
Thanks Yamakuzure, unfortunately when using you config my cpu never goes down 3.7GHz when idle hence the powersave doesn't work at all even if I adjust it manually.
Code: | # cpupower frequency-set --governor powersave && cpupower frequency-info && cat /proc/cpuinfo | grep MHz
Setting cpu: 0
Setting cpu: 1
Setting cpu: 2
Setting cpu: 3
analyzing CPU 0:
driver: intel_pstate
CPUs which run at the same hardware frequency: 0
CPUs which need to have their frequency coordinated by software: 0
maximum transition latency: 0.97 ms.
hardware limits: 800 MHz - 3.70 GHz
available cpufreq governors: performance, powersave
current policy: frequency should be within 800 MHz and 3.70 GHz.
The governor "powersave" may decide which speed to use
within this range.
current CPU frequency is 3.63 GHz (asserted by call to hardware).
boost state support:
Supported: yes
Active: yes
cpu MHz : 3628.066
cpu MHz : 3680.789
cpu MHz : 3588.105
cpu MHz : 3665.835 |
So that's how I found out that disabling acpi_cpu and intel_idle did the trick, trial and error for hours! But then the chip can't get to the maximum frequency of 3.7GHz, just up to 3,5GHz. Maybe the problem is present only in the newer haswell refresh and it's a known bug.
Thanks for your time, I will remove the "solved" mark because it clearly it's not solved yet. |
|
Back to top |
|
|
albright Advocate
Joined: 16 Nov 2003 Posts: 2588 Location: Near Toronto
|
Posted: Fri Mar 27, 2015 6:31 pm Post subject: |
|
|
Here is my config
Code: | /usr/src/linux $ grep -P "^[^#].*(FREQ|IDLE|PSTATE)" .config
CONFIG_NO_HZ_IDLE=y
CONFIG_GENERIC_SMP_IDLE_THREAD=y
CONFIG_CPU_FREQ=y
CONFIG_CPU_FREQ_GOV_COMMON=y
CONFIG_CPU_FREQ_STAT=y
CONFIG_CPU_FREQ_DEFAULT_GOV_ONDEMAND=y
CONFIG_CPU_FREQ_GOV_PERFORMANCE=y
CONFIG_CPU_FREQ_GOV_ONDEMAND=y
CONFIG_X86_INTEL_PSTATE=y
CONFIG_CPU_IDLE=y
CONFIG_CPU_IDLE_GOV_MENU=y
CONFIG_INTEL_IDLE=y
|
_________________ .... there is nothing - absolutely nothing - half so much worth
doing as simply messing about with Linux ...
(apologies to Kenneth Graeme) |
|
Back to top |
|
|
Ant P. Watchman
Joined: 18 Apr 2009 Posts: 6920
|
Posted: Sat Mar 28, 2015 11:32 pm Post subject: |
|
|
Intel cheats, basically. The newer chips use C-states alone for power saving - C2 and higher is effectively 0MHz, though you can't tell that from cpufreq numbers alone. |
|
Back to top |
|
|
Bloot Tux's lil' helper
Joined: 10 Mar 2006 Posts: 99 Location: Barcelona
|
|
Back to top |
|
|
Bloot Tux's lil' helper
Joined: 10 Mar 2006 Posts: 99 Location: Barcelona
|
Posted: Thu Jun 18, 2015 11:43 pm Post subject: |
|
|
4.0.5 kernel finally solved it.
Edit
False alarm... once ati-drivers got compiled against the new kernel, the problem is there again, so I guess it's ati-drivers who prevents the cpu being idle. |
|
Back to top |
|
|
Yamakuzure Advocate
Joined: 21 Jun 2006 Posts: 2284 Location: Adendorf, Germany
|
Posted: Fri Jun 19, 2015 7:27 am Post subject: |
|
|
Bloot wrote: | 4.0.5 kernel finally solved it.
Edit
False alarm... once ati-drivers got compiled against the new kernel, the problem is there again, so I guess it's ati-drivers who prevents the cpu being idle. | There could be another issue.
Even in powersave mode, the CPUs can go up to maximum MHz including Turbo Boost if something wakes them up too often.
Could you try out powertop? Maybe it is something ompletely different?
(P.S : However: Powersave mode does *not* necessarily extend battery life, but may actually shorten it.
The reason is, that a CPU in a lower frequency needs longer to switch to C1, do its work, and switch back. So reducing the frequency actually lengthens the periods the CPUs are in a non-idle state, consuming more power. Theoretically.
Hence, unless undervolting the CPU, doing regular work while using the powersave governor, is a bad idea. ) _________________ Important German:- "Aha" - German reaction to pretend that you are really interested while giving no f*ck.
- "Tja" - German reaction to the apocalypse, nuclear war, an alien invasion or no bread in the house.
|
|
Back to top |
|
|
s4e8 Guru
Joined: 29 Jul 2006 Posts: 311
|
Posted: Fri Jun 19, 2015 9:10 am Post subject: |
|
|
There's NO freq at idle (C-state), all C3/C6/C7 states stop core-clock, C1 will running at low-speed if C1E enable, or high speed if C1E disable. The reported idle freq is useless. |
|
Back to top |
|
|
Bloot Tux's lil' helper
Joined: 10 Mar 2006 Posts: 99 Location: Barcelona
|
Posted: Fri Jun 19, 2015 11:40 pm Post subject: |
|
|
Yamakuzure wrote: | Bloot wrote: | 4.0.5 kernel finally solved it.
Edit
False alarm... once ati-drivers got compiled against the new kernel, the problem is there again, so I guess it's ati-drivers who prevents the cpu being idle. | There could be another issue.
Even in powersave mode, the CPUs can go up to maximum MHz including Turbo Boost if something wakes them up too often.
Could you try out powertop? Maybe it is something ompletely different?
(P.S : However: Powersave mode does *not* necessarily extend battery life, but may actually shorten it.
The reason is, that a CPU in a lower frequency needs longer to switch to C1, do its work, and switch back. So reducing the frequency actually lengthens the periods the CPUs are in a non-idle state, consuming more power. Theoretically.
Hence, unless undervolting the CPU, doing regular work while using the powersave governor, is a bad idea. ) |
Hi, mine's a desktop computer, so no problem with batteries. It's funny, because before I compiled the proprietary ATi drivers, the frequency stayed at 800-900MHz with no cpu load (as it should), and after compiled the video drivers the cpu stays again at 3.7GHz with no cpu load, I can only guess it's a video drivers fault, or maybe the KDE desktop effects has also something to do with it (I tried disabling them, but no luck).
Code: | top - 01:36:20 up 10 min, 1 user, load average: 0,92, 1,72, 1,08
Tasks: 213 total, 1 running, 212 sleeping, 0 stopped, 0 zombie
%Cpu(s): 1,6 us, 0,4 sy, 0,0 ni, 97,4 id, 0,6 wa, 0,0 hi, 0,0 si, 0,0 st
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
5917 debaser 20 0 884192 129712 73484 S 3,0 1,6 0:08.83 chrome
3017 debaser 20 0 3047140 133848 100236 S 1,7 1,7 0:14.88 kwin
3812 debaser 20 0 1070612 268824 109188 S 1,3 3,3 0:13.78 chrome
2604 root 20 0 436124 208408 163136 S 0,7 2,6 0:11.67 X
3734 debaser 20 0 331220 48672 33620 S 0,7 0,6 0:03.68 urxvt
3865 debaser 20 0 437792 65600 51712 S 0,7 0,8 0:03.52 chrome
7225 debaser 20 0 328648 45756 33136 S 0,7 0,6 0:02.87 urxvt
5665 debaser 20 0 920888 154780 54300 S 0,3 1,9 0:01.56 chrome
6068 debaser 20 0 954012 161688 59372 S 0,3 2,0 0:02.05 chrome
7450 debaser 20 0 826720 86344 51108 S 0,3 1,1 0:00.24 chrome
7761 debaser 20 0 328648 46256 33636 S 0,3 0,6 0:02.00 urxvt
1 root 20 0 4168 1456 1352 S 0,0 0,0 0:00.41 init
2 root 20 0 0 0 0 S 0,0 0,0 0:00.00 kthreadd
3 root 20 0 0 0 0 S 0,0 0,0 0:00.06 ksoftirqd/0
4 root 20 0 0 0 0 S 0,0 0,0 0:00.04 kworker/0:0
5 root 0 -20 0 0 0 S 0,0 0,0 0:00.00 kworker/0:0H
6 root 20 0 0 0 0 S 0,0 0,0 0:02.20 kworker/u8:0
7 root 20 0 0 0 0 S 0,0 0,0 0:00.48 rcu_sched
8 root 20 0 0 0 0 S 0,0 0,0 0:00.00 rcu_bh
9 root rt 0 0 0 0 S 0,0 0,0 0:00.00 migration/0
10 root rt 0 0 0 0 S 0,0 0,0 0:00.00 migration/1
11 root 20 0 0 0 0 S 0,0 0,0 0:00.02 ksoftirqd/1
13 root 0 -20 0 0 0 S 0,0 0,0 0:00.00 kworker/1:0H
14 root rt 0 0 0 0 S 0,0 0,0 0:00.00 migration/2
15 root 20 0 0 0 0 S 0,0 0,0 0:00.01 ksoftirqd/2
17 root 0 -20 0 0 0 S 0,0 0,0 0:00.00 kworker/2:0H
18 root rt 0 0 0 0 S 0,0 0,0 0:00.00 migration/3
19 root 20 0 0 0 0 S 0,0 0,0 0:00.02 ksoftirqd/3
21 root 0 -20 0 0 0 S 0,0 0,0 0:00.00 kworker/3:0H
22 root 0 -20 0 0 0 S 0,0 0,0 0:00.00 khelper
23 root 20 0 0 0 0 S 0,0 0,0 0:00.00 kdevtmpfs
24 root 0 -20 0 0 0 S 0,0 0,0 0:00.00 netns
27 root 0 -20 0 0 0 S 0,0 0,0 0:00.00 perf
286 root 20 0 0 0 0 S 0,0 0,0 0:00.00 khungtaskd
287 root 0 -20 0 0 0 S 0,0 0,0 0:00.00 writeback
289 root 0 -20 0 0 0 S 0,0 0,0 0:00.00 crypto
290 root 20 0 0 0 0 S 0,0 0,0 0:00.01 kworker/1:1
291 root 0 -20 0 0 0 S 0,0 0,0 0:00.00 bioset
293 root 0 -20 0 0 0 S 0,0 0,0 0:00.00 kblockd |
I think I'll give up, funniest thing is the chip frequencies works fine on Windows, it's just Linux related. |
|
Back to top |
|
|
s4e8 Guru
Joined: 29 Jul 2006 Posts: 311
|
Posted: Sat Jun 20, 2015 12:54 am Post subject: |
|
|
new intel_idle, export both C1 and C1E states, let the cpuidle gorvenor choose which one to use. ACPI idle driver always combine C1 and C1E, and auto-promotion to C1E if BIOS enable it. Unless cpu wont enter deeper C3 or C6 states, there's NO problem if idle governor prefer to C1 instead C1E. |
|
Back to top |
|
|
Bloot Tux's lil' helper
Joined: 10 Mar 2006 Posts: 99 Location: Barcelona
|
Posted: Sat Jun 20, 2015 11:05 pm Post subject: |
|
|
s4e8 wrote: | new intel_idle, export both C1 and C1E states, let the cpuidle gorvenor choose which one to use. ACPI idle driver always combine C1 and C1E, and auto-promotion to C1E if BIOS enable it. Unless cpu wont enter deeper C3 or C6 states, there's NO problem if idle governor prefer to C1 instead C1E. |
Yes I'm giving up, I guess it's the normal Haswell refresh chips behaviour under Linux, most important temps are OK, thank you all. |
|
Back to top |
|
|
Yamakuzure Advocate
Joined: 21 Jun 2006 Posts: 2284 Location: Adendorf, Germany
|
Posted: Mon Jun 22, 2015 6:43 am Post subject: |
|
|
@Bloot: I meant "sys-power/powertop", not "top".
But as you are on a desktop computer without battery, it wouldn't help that much anyway. I guess you are right, and your desktop hardware simply isn't able to do it like you want. I always assumed you were on a laptop.
However, my laptop came shipped with an ATI card, but both the open source and proprietary drivers gave me too much trouble. I had it replaced by an nvidia card and am happy now. So if you are right, and it is a driver issue with the proprietary ATI-Drivers, I can at least be sure that my decision was a good one.
For completenes' sake:
Output with AC connected on my laptop: Code: | ~ # LC_ALL=C powertop --time=5
PowerTOP 2.6 Overview Idle stats Frequency stats Device stats Tunables
Summary: 4506.7 wakeups/second, 109.7 GPU ops/seconds, 0.0 VFS ops/sec and 23.5% CPU use
Usage Events/s Category Description
137.0 ms/s 2169.6 Process /usr/sbin/mysqld --defaults-file=/etc/mysql/my.cnf
7.7 ms/s 1109.8 Process akonadiserver
43.3 ms/s 429.8 Process /usr/bin/lancelot
1.3 ms/s 255.1 Process /usr/games/bin/wesnoth
2.6 ms/s 204.2 Timer tick_sched_timer
6.8 ms/s 147.8 Process kdeinit4: plasma-desktop [kdeinit]
2.1 ms/s 125.9 Timer hrtimer_wakeup
9.0 ms/s 34.3 Process /usr/bin/firefox
6.9 ms/s 37.6 Process /usr/bin/X :0 -auth /var/run/sddm/:0 -nolisten tcp -background none -noreset vt7
2.0 ms/s 27.7 Interrupt [25] i915
1.0 ms/s 11.2 Process kwin -session 10143143147fd000143149784900000084790000_1434736059_314150
1.5 ms/s 8.4 Process /usr/bin/yakuake -session 10143143147fd000140059895900000131970026_1434736059_305163
4.1 ms/s 7.0 Process /usr/bin/kontact
8.5 us/s 6.6 kWork blk_delay_work
104.6 us/s 6.0 Process /usr/lib/jvm/oracle-jdk-bin-1.8//bin/java -classpath /usr/share/davmail-bin/lib/activation-1.1.1.jar:/usr/share/davmail-bin/lib
16.7 us/s 5.8 Process [rcu_sched]
58.6 us/s 3.4 Process [z_null_iss/0]
529.7 us/s 2.2 Process /usr/bin/conky
154.0 us/s 2.0 Process [z_wr_iss/1]
23.0 us/s 2.0 Process [txg_sync]
651.9 us/s 1.4 Interrupt [30] snd_hda_intel
155.8 us/s 1.4 Process [z_wr_iss/0]
1.9 ms/s 0.6 Process [uksmd]
65.0 us/s 1.0 Process /usr/sbin/syslog-ng --persist-file /var/lib/syslog-ng/syslog-ng.persist --cfgfile /etc/syslog-ng/syslog-ng.conf --pidfile /run/ | Output while on batteries on my laptop with powersave governor (600-800MHz on all 8 CPUs): Code: | ~ # LC_ALL=C powertop --time=5
PowerTOP 2.6 Overview Idle stats Frequency stats Device stats Tunables
The battery reports a discharge rate of 20.0 W
The estimated remaining time is 3 hours, 21 minutes
Summary: 2671.8 wakeups/second, 148.2 GPU ops/seconds, 0.0 VFS ops/sec and 19.9% CPU use
Usage Events/s Category Description
18.9 ms/s 1041.9 Process /usr/games/bin/wesnoth
60.7 ms/s 433.6 Process /usr/bin/firefox
2.5 ms/s 324.9 Timer tick_sched_timer
11.8 ms/s 182.6 Process kdeinit4: plasma-desktop [kdeinit]
6.9 ms/s 162.8 Process /usr/bin/conky
9.7 ms/s 148.8 Process kwin -session 10143143147fd000143149784900000084790000_1434736059_314150
4.8 ms/s 83.9 Process /usr/bin/yakuake -session 10143143147fd000140059895900000131970026_1434736059_305163
1.9 ms/s 73.7 Interrupt [25] i915
1.0 ms/s 62.2 Process [txg_sync]
334.0 us/s 56.0 Process [rcu_sched]
1.2 ms/s 43.7 Process /usr/lib/jvm/oracle-jdk-bin-1.8//bin/java -classpath /usr/share/davmail-bin/lib/activation-1.1.1.jar:/usr/share/davmail-bin/lib
27.4 ms/s 21.7 Process /usr/bin/X :0 -auth /var/run/sddm/:0 -nolisten tcp -background none -noreset vt7
311.6 us/s 25.6 Process /usr/sbin/mysqld --defaults-file=/etc/mysql/my.cnf
1.3 ms/s 22.7 Process /usr/sbin/syslog-ng --persist-file /var/lib/syslog-ng/syslog-ng.persist --cfgfile /etc/syslog-ng/syslog-ng.conf --pidfile /run/
280.3 us/s 21.3 Process [z_null_iss/0]
5.5 ms/s 18.9 Process default.sh
6.8 ms/s 7.4 Process acpid
0.7 ms/s 7.6 Process /usr/sbin/acpid --netlink
7.0 ms/s 3.8 Process sh
130.1 us/s 4.6 kWork acpi_os_execute_deferred
5.1 ms/s 2.4 Process [uksmd] | Idle stats: Code: | Package | Core | CPU 0 CPU 4
| | C0 active 2.6% 0.9%
| | POLL 0.0% 0.0 ms 0.0% 0.0 ms
| | C1E-HSW 0.1% 0.1 ms 19.6% 49.5 ms
C2 (pc2) 22.0% | |
C3 (pc3) 0.0% | C3 (cc3) 0.4% | C3-HSW 1.6% 0.4 ms 0.0% 0.0 ms
C6 (pc6) 0.0% | C6 (cc6) 0.2% | C6-HSW 1.5% 1.0 ms 0.0% 0.0 ms
C7 (pc7) 0.0% | C7 (cc7) 33.4% | C7s-HSW 93.6% 2.5 ms 35.8% 15.0 ms
| Core | CPU 1 CPU 5
| | C0 active 2.3% 0.6%
| | POLL 0.0% 0.0 ms 0.0% 0.0 ms
| | C1E-HSW 5.4% 0.9 ms 1.6% 1.8 ms
| |
| C3 (cc3) 1.6% | C3-HSW 2.1% 0.5 ms 0.0% 0.0 ms
| C6 (cc6) 0.9% | C6-HSW 0.8% 0.6 ms 0.2% 0.7 ms
| C7 (cc7) 84.6% | C7s-HSW 88.3% 3.2 ms 96.3% 60.7 ms
| Core | CPU 2 CPU 6
| | C0 active 4.7% 1.5%
| | POLL 0.0% 0.0 ms 0.0% 0.0 ms
| | C1E-HSW 19.4% 5.7 ms 0.0% 0.0 ms
| |
| C3 (cc3) 1.2% | C3-HSW 0.4% 0.5 ms 1.7% 2.1 ms
| C6 (cc6) 0.5% | C6-HSW 0.5% 0.6 ms 0.3% 1.6 ms
| C7 (cc7) 70.7% | C7s-HSW 73.6% 3.2 ms 96.6% 11.3 ms
| Core | CPU 3 CPU 7
| | C0 active 2.0% 0.2%
| | POLL 0.0% 0.0 ms 0.0% 0.0 ms
| | C1E-HSW 0.5% 0.2 ms 0.0% 0.2 ms
| |
| C3 (cc3) 1.6% | C3-HSW 1.7% 0.4 ms 0.0% 0.1 ms
| C6 (cc6) 0.8% | C6-HSW 0.8% 0.6 ms 0.1% 0.3 ms
| C7 (cc7) 91.0% | C7s-HSW 93.1% 1.6 ms 99.7% 83.8 ms
| GPU |
| |
| Powered On 1.5% |
| RC6 98.5% |
| RC6p 0.0% |
| RC6pp 0.0% |
| |
| | | So you can at least see what wakes up your CPUs and what time they spend in which state. _________________ Important German:- "Aha" - German reaction to pretend that you are really interested while giving no f*ck.
- "Tja" - German reaction to the apocalypse, nuclear war, an alien invasion or no bread in the house.
|
|
Back to top |
|
|
yaclo n00b
Joined: 11 Sep 2015 Posts: 20
|
Posted: Mon Feb 01, 2016 10:08 pm Post subject: |
|
|
same problem here and it's a desktop it's first generation haswell cpu i7 4770 / r9 290
with radeonsi cpu frequency scaling works fine
but with fglrx cpu freq goes to max no matter what
fglrx prevents the cpu being idle |
|
Back to top |
|
|
|
|
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
|
|