2

I installed Ubuntu 20.04 and have the following cpu:

grep -E '^model name|^cpu MHz' /proc/cpuinfo

model name  : Intel(R) Core(TM) i7-8550U CPU @ 1.80GHz
cpu MHz     : 3000.025
model name  : Intel(R) Core(TM) i7-8550U CPU @ 1.80GHz
cpu MHz     : 2721.680
model name  : Intel(R) Core(TM) i7-8550U CPU @ 1.80GHz
cpu MHz     : 2999.579
model name  : Intel(R) Core(TM) i7-8550U CPU @ 1.80GHz
cpu MHz     : 3000.027
model name  : Intel(R) Core(TM) i7-8550U CPU @ 1.80GHz
cpu MHz     : 2993.468
model name  : Intel(R) Core(TM) i7-8550U CPU @ 1.80GHz
cpu MHz     : 3000.568
model name  : Intel(R) Core(TM) i7-8550U CPU @ 1.80GHz
cpu MHz     : 2989.114
model name  : Intel(R) Core(TM) i7-8550U CPU @ 1.80GHz
cpu MHz     : 3001.278

I use the following command to set governor and frequency range:

sudo cpupower frequency-set --governor performance --min 2600000 --max 3000000

And the following command to verify it:

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:  Cannot determine or is not supported.
  hardware limits: 400 MHz - 4.00 GHz
  available cpufreq governors: performance powersave
  current policy: frequency should be within 2.60 GHz and 3.00 GHz.
                  The governor "performance" may decide which speed to use
                  within this range.
  current CPU frequency: Unable to call hardware
  current CPU frequency: 3.00 GHz (asserted by call to kernel)
  boost state support:
    Supported: yes
    Active: yes

Current frequency is in specified range, this is true under no-load conditions.

However, after I start program (gmic) that loads cpu, current frequency drops below specified range:

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:  Cannot determine or is not supported.
  hardware limits: 400 MHz - 4.00 GHz
  available cpufreq governors: performance powersave
  current policy: frequency should be within 2.60 GHz and 3.00 GHz.
                  The governor "performance" may decide which speed to use
                  within this range.
  current CPU frequency: Unable to call hardware
  current CPU frequency: 1.80 GHz (asserted by call to kernel)
  boost state support:
    Supported: yes
    Active: yes

Under Windows I run the same application and achieve much higher performance. Under no-load conditions frequency is around 1GHz, and under load frequency jumps to 2.6GHz. So I don't think it is hardware or bios related issue.

I rebooted Ubuntu to discard my cpu frequency changes (so they are defaults again). I will monitor cpu frequency and other important info with turbostat utility.

Here is turbostat output:

sudo turbostat --Summary --quiet --show Busy%,Bzy_MHz,IRQ,PkgWatt,PkgTmp,RAMWatt,GFXWatt --interval 5

Busy%  Bzy_MHz    IRQ   PkgTmp PkgWatt GFXWatt RAMWatt
0.85     703     1663     45    1.22    0.00    0.00
0.81     654     1565     46    1.22    0.00    0.00
0.63     656     1338     45    1.19    0.00    0.00
0.74     664     1440     45    1.20    0.00    0.00
2.71     1650    2587     46    1.76    0.00    0.00
4.44     842     4362     46    1.53    0.00    0.00
5.35     1667    4806     46    2.45    0.00    0.00
5.47     831     4510     46    1.53    0.00    0.00
5.50     991     5130     47    1.68    0.00    0.00
2.73     1093    3055     48    1.48    0.00    0.00
2.49     1025    2780     57    1.47    0.00    0.00
2.28     2651    2098     47    2.53    0.00    0.00
1.35     1908    1685     47    1.63    0.00    0.00
1.03     1063    1823     47    1.29    0.00    0.00
1.41     995     2057     47    1.37    0.00    0.00
1.00     1003    1628     46    1.29    0.00    0.00
1.48     1132    1946     46    1.36    0.00    0.00
0.88     935     1488     46    1.25    0.00    0.00
1.13     957     1740     46    1.30    0.00    0.00
1.07     904     1660     46    1.28    0.00    0.00
2.85     895     3249     46    1.52    0.00    0.00
1.19     755     1686     46    1.25    0.00    0.00
2.25     1224    2992     70    1.57    0.00    0.00


the moment program that loads cpu started


82.89    3703    13909    71    23.55   0.00    0.00
99.70    2589    11348    54    14.88   0.00    0.00
99.64    1960    11179    53    6.04    0.00    0.00
99.67    2122    11149    53    6.37    0.00    0.00
98.61    1851    11558    54    6.59    0.00    0.00
99.71    1577    11087    53    6.38    0.00    0.00
99.72    1546    11028    54    6.37    0.00    0.00
99.70    1740    11231    54    6.28    0.00    0.00
99.63    1945    11376    54    6.45    0.00    0.00
99.56    1939    11804    52    6.51    0.00    0.00
99.60    1949    11816    52    6.46    0.00    0.00
99.63    1971    11599    54    6.49    0.00    0.00
99.74    1833    11475    52    6.51    0.00    0.00
99.76    1711    11436    53    6.45    0.00    0.00
99.74    1800    11383    52    6.28    0.00    0.00
99.71    2117    11446    53    6.45    0.00    0.00
99.45    2031    11728    55    6.64    0.00    0.00
99.61    1646    11629    53    6.54    0.00    0.00
99.74    1562    11550    53    6.37    0.00    0.00
99.74    1599    11559    53    6.45    0.00    0.00
99.69    1670    11379    52    6.37    0.00    0.00
99.67    1866    11664    52    6.40    0.00    0.00
99.62    1963    11871    53    6.44    0.00    0.00
99.60    1961    11956    54    6.53    0.00    0.00
99.64    1919    11791    52    6.41    0.00    0.00
99.72    1831    11575    54    6.54    0.00    0.00
99.74    1794    11408    54    6.40    0.00    0.00
99.74    1802    11737    53    6.44    0.00    0.00
99.71    1991    11491    52    6.38    0.00    0.00
99.70    2005    11542    53    6.61    0.00    0.00
97.88    1706    11697    53    6.51    0.00    0.00
99.73    1571    12302    53    6.38    0.00    0.00
99.72    1550    12159    54    6.47    0.00    0.00
94.27    1650    12576    50    6.04    0.00    0.00


the moment program that loads cpu terminated


3.03     827     3706     48    1.49    0.00    0.00
1.73     789     2349     47    1.32    0.00    0.00
2.88     818     3389     47    1.38    0.00    0.00
1.82     777     2359     46    1.27    0.00    0.00
1.74     730     2233     47    1.25    0.00    0.00
2.28     756     2906     46    1.31    0.00    0.00
2.59     813     2916     45    1.34    0.00    0.00
2.59     909     3089     46    1.39    0.00    0.00
2.32     752     3077     46    1.33    0.00    0.00
2.09     899     2738     45    1.31    0.00    0.00
1.79     808     2245     45    1.27    0.00    0.00
1.73     818     2126     44    1.26    0.00    0.00
1.78     808     2231     44    1.26    0.00    0.00
1.88     870     2320     45    1.31    0.00    0.00

Now I will execute the following command again:

sudo cpupower frequency-set --governor performance --min 2600000 --max 3000000

Let's see turbostat output under these conditions:

sudo turbostat --Summary --quiet --show Busy%,Bzy_MHz,IRQ,PkgWatt,PkgTmp,RAMWatt,GFXWatt --interval 5

Busy%   Bzy_MHz  IRQ    PkgTmp PkgWatt GFXWatt RAMWatt
0.67     2999    2148     45    1.44    0.00    0.00
0.74     2999    2251     45    1.44    0.00    0.00
0.69     2999    2130     45    1.42    0.00    0.00
1.81     2998    3719     45    2.01    0.00    0.00
2.76     2993    5245     45    2.49    0.00    0.00
0.96     2997    3300     45    1.62    0.00    0.00
0.99     2996    3248     45    1.63    0.00    0.00
1.56     2988    4440     46    1.93    0.00    0.00
0.61     2999    2106     45    1.41    0.00    0.00
0.60     2998    2200     45    1.40    0.00    0.00
0.60     2974    2105     45    1.42    0.00    0.00
0.69     2958    2280     45    1.45    0.00    0.00
0.73     2976    2829     45    1.46    0.00    0.00
0.69     3000    2127     45    1.43    0.00    0.00
0.67     2998    2071     45    1.43    0.00    0.00
0.89     3000    2709     45    1.61    0.00    0.00
1.27     2997    3685     45    1.85    0.00    0.00


the moment program that loads cpu started


11.67    3000    4679     58    3.48    0.00    0.00
99.74    3000    12823    70    14.83   0.00    0.00
99.69    3000    11618    64    16.38   0.00    0.00
99.69    2623    11406    54    9.63    0.00    0.00
95.83    1975    11912    56    6.35    0.00    0.00
99.02    1650    11763    53    6.36    0.00    0.00
99.73    1525    11397    54    6.34    0.00    0.00
99.73    1593    12664    53    6.32    0.00    0.00
99.67    1812    13476    52    6.40    0.00    0.00
99.63    1913    13013    53    6.39    0.00    0.00
99.51    1934    11901    54    6.48    0.00    0.00
99.62    1973    11892    52    6.44    0.00    0.00
99.70    1899    11586    54    6.59    0.00    0.00
99.76    1775    11497    54    6.45    0.00    0.00
99.76    1702    11638    54    6.40    0.00    0.00
99.71    1975    11709    53    6.28    0.00    0.00
99.69    2134    11636    53    6.52    0.00    0.00
99.37    1841    11991    53    6.66    0.00    0.00
99.66    1577    11541    53    6.43    0.00    0.00
99.69    1580    11430    53    6.38    0.00    0.00
99.73    1576    11357    53    6.43    0.00    0.00
99.69    1759    11356    52    6.34    0.00    0.00
99.62    1934    11787    53    6.46    0.00    0.00
99.49    1955    11894    52    6.50    0.00    0.00
99.63    1939    11815    52    6.48    0.00    0.00
99.70    1938    11668    53    6.48    0.00    0.00
99.75    1751    11530    54    6.47    0.00    0.00
99.75    1734    11578    53    6.45    0.00    0.00
44.89    1778    8422     49    3.88    0.00    0.00


the moment program that loads cpu terminated


0.61     2997    2309     48    1.49    0.00    0.00
0.88     2999    4576     47    1.54    0.00    0.00
0.68     3000    1906     47    1.43    0.00    0.00
0.85     2995    2237     46    1.54    0.00    0.00
0.89     2987    2430     46    1.53    0.00    0.00
0.61     2989    2091     46    1.41    0.00    0.00
0.66     2934    1990     46    1.46    0.00    0.00
1.58     2980    4030     46    2.05    0.00    0.00
1.49     2995    4678     46    2.01    0.00    0.00
1.46     2991    4713     45    1.97    0.00    0.00
0.74     2994    2481     46    1.49    0.00    0.00
0.82     2999    3804     45    1.48    0.00    0.00
0.69     2997    2095     45    1.44    0.00    0.00
0.75     2997    2688     45    1.49    0.00    0.00
0.59     3000    1957     45    1.39    0.00    0.00
0.67     2996    2041     45    1.43    0.00    0.00
1.11     2999    3236     45    1.68    0.00    0.00
1.31     2997    3892     46    1.79    0.00    0.00

So it looks like Ubuntu is throttling cpu power consumption. Since it's not possible to stay under high load and within specified frequency range (when power consumption is cut), frequency drops to satisfy power constraint.

Now I reboot Ubuntu again, so cpu frequency goes to defaults.

I indeed was running thermald:

sudo service thermald status

● thermald.service - Thermal Daemon Service
     Loaded: loaded (/lib/systemd/system/thermald.service; enabled; vendor preset: enabled)
     Active: active (running) since Fri 2021-04-09 09:11:43 EEST; 9h ago
   Main PID: 889 (thermald)
      Tasks: 2 (limit: 19017)
     Memory: 6.7M
     CGroup: /system.slice/thermald.service
             └─889 /usr/sbin/thermald --no-daemon --dbus-enable


кві 09 09:11:43 igor-HP-ProBook-450-G5 thermald[889]: [WARN]sensor id 16 : No temp sysfs for reading raw temp
кві 09 09:11:43 igor-HP-ProBook-450-G5 systemd[1]: Started Thermal Daemon Service.
кві 09 09:11:43 igor-HP-ProBook-450-G5 thermald[889]: [WARN]sensor id 16 : No temp sysfs for reading raw temp
кві 09 09:11:43 igor-HP-ProBook-450-G5 thermald[889]: I/O warning : failed to load external entity "/etc/thermald/thermal-conf.xml"
кві 09 09:11:43 igor-HP-ProBook-450-G5 thermald[889]: [WARN]error: could not parse file /etc/thermald/thermal-conf.xml
кві 09 09:11:43 igor-HP-ProBook-450-G5 thermald[889]: [WARN]sysfs open failed
кві 09 09:11:43 igor-HP-ProBook-450-G5 thermald[889]: I/O warning : failed to load external entity "/etc/thermald/thermal-conf.xml"
кві 09 09:11:43 igor-HP-ProBook-450-G5 thermald[889]: [WARN]error: could not parse file /etc/thermald/thermal-conf.xml
кві 09 09:11:43 igor-HP-ProBook-450-G5 thermald[889]: I/O warning : failed to load external entity "/etc/thermald/thermal-conf.xml"
кві 09 09:11:43 igor-HP-ProBook-450-G5 thermald[889]: [WARN]error: could not parse file /etc/thermald/thermal-conf.xml

So I stopped it (it did not solve cpu throttling issue):

sudo service thermald stop

And verified it's no longer active:

sudo service thermald status

● thermald.service - Thermal Daemon Service
     Loaded: loaded (/lib/systemd/system/thermald.service; enabled; vendor preset: enabled)
     Active: inactive (dead) since Fri 2021-04-09 18:37:15 EEST; 2s ago
    Process: 889 ExecStart=/usr/sbin/thermald --no-daemon --dbus-enable (code=exited, status=0/SUCCESS)
   Main PID: 889 (code=exited, status=0/SUCCESS)


кві 09 09:11:43 igor-HP-ProBook-450-G5 thermald[889]: [WARN]sysfs open failed
кві 09 09:11:43 igor-HP-ProBook-450-G5 thermald[889]: I/O warning : failed to load external entity "/etc/thermald/thermal-conf.xml"
кві 09 09:11:43 igor-HP-ProBook-450-G5 thermald[889]: [WARN]error: could not parse file /etc/thermald/thermal-conf.xml
кві 09 09:11:43 igor-HP-ProBook-450-G5 thermald[889]: I/O warning : failed to load external entity "/etc/thermald/thermal-conf.xml"
кві 09 09:11:43 igor-HP-ProBook-450-G5 thermald[889]: [WARN]error: could not parse file /etc/thermald/thermal-conf.xml
кві 09 18:37:13 igor-HP-ProBook-450-G5 systemd[1]: Stopping Thermal Daemon Service...
кві 09 18:37:13 igor-HP-ProBook-450-G5 thermald[889]: [WARN]Terminating ...
кві 09 18:37:14 igor-HP-ProBook-450-G5 thermald[889]: [WARN]terminating on user request ..
кві 09 18:37:15 igor-HP-ProBook-450-G5 systemd[1]: thermald.service: Succeeded.
кві 09 18:37:15 igor-HP-ProBook-450-G5 systemd[1]: Stopped Thermal Daemon Service.

To collect the following data:

1. turbostat header
2. rdmsr (MSR_CORE_PERF_LIMIT_REASONS)
3. params in /sys/devices/system/cpu/intel_pstate directory
4. params in /sys/class/powercap/intel-rapl/intel-rapl:0 directory
5. scaling limits in /sys/devices/system/cpu/cpu*/cpufreq directory

I wrote the following script (collect_cpu_info.sh):

#!/bin/bash

exec_turbostat()
{
    echo "turbostat --Summary --show Busy%,Bzy_MHz,IRQ,PkgWatt,PkgTmp,RAMWatt,GFXWatt --interval 0.001 --num_iterations 1"
    turbostat --Summary --show Busy%,Bzy_MHz,IRQ,PkgWatt,PkgTmp,RAMWatt,GFXWatt --interval 0.001 --num_iterations 1
}


exec_rdmsr()
{
    echo "rdmsr 0x64f"
    rdmsr 0x64f
}


watch_intel_pstate()
{
    echo "grep . /sys/devices/system/cpu/intel_pstate/"
    grep . /sys/devices/system/cpu/intel_pstate/
}


watch_intel_rapl()
{
    echo "grep . /sys/class/powercap/intel-rapl/intel-rapl:0/"
    grep . /sys/class/powercap/intel-rapl/intel-rapl:0/
}


watch_cpufreq()
{
    echo "grep . /sys/devices/system/cpu/cpu/cpufreq/scaling_m"
    grep . /sys/devices/system/cpu/cpu/cpufreq/scaling_m
}


while sleep 5
do
    echo "------------------------------------------------------------------------------------------------"
    exec_turbostat
    echo ""
    exec_rdmsr
    echo ""
    watch_intel_pstate
    echo ""
    watch_intel_rapl
    echo ""
    watch_cpufreq
done

Let's run it (full log is not provided due to its size, I just will show load - no_load transition points):

<<< see log below in my next post, because post size is limited >>>

My laptop stays plugged in all the time, so on-battery related power limitations should not occur.

Do you know how to turn off cpu power constraint on Ubuntu? So cpu will consume max amount of power it needs to stay on high frequency. Any comments will be appreciated. Thank you.

igntec
  • 121
  • 3
  • there seem to be several similar questions in the last few days. See this one, including the comments. Edit your question with the same requested information, then flag me with a comment and I'll have a look. – Doug Smythies Apr 06 '21 at 14:13
  • @DougSmythies please take a look. – igntec Apr 07 '21 at 15:29
  • Yes, it appears to be power and/or temperature throttling, but then not recovering properly on the fly, which it should. I am just on my way out, and will look in more detail later. In your first example it undershot to 6 watts and did not recover. That is wrong. – Doug Smythies Apr 07 '21 at 15:44
  • @DougSmythies ok, thank you – igntec Apr 07 '21 at 15:50
  • please watch grep . /sys/devices/system/cpu/intel_pstate/* and grep . /sys/devices/system/cpu/cpu*/cpufreq/scaling_m* during these events also. I'll be back later. – Doug Smythies Apr 07 '21 at 15:50
  • Are you running any thermal daemon, such as thermald or tlp or other? In addition to above request, please obtain the turbostat spew of startup stuff while throttling is occurring. We need to see the bits of the PACKAGE_THERM_STATUS msr, which contrary to the register name include power limiting bits. Your issue does NOT appear to be power throttling. Are you willing to try some things, which might turn out to be a waste of time? Are there any temperature related messages in any logs file in /var/log, say kern.log and/or syslog and/or ...? – Doug Smythies Apr 07 '21 at 18:47
  • Please also provide the results for sudo rdmsr 0x64f (MSR_CORE_PERF_LIMIT_REASONS) during throttling, as I don't think turbostat does so. – Doug Smythies Apr 08 '21 at 00:05
  • @DougSmythies sorry for delay, I was busy with work. Sure, let's try to fix this. Anyway it already works on Windows. I will be back tomorrow and provide all requested info. – igntec Apr 08 '21 at 22:30
  • @DougSmythies I uploaded all requested info, please take a look – igntec Apr 09 '21 at 18:40
  • It is power limiting after all. the first power limit stuff is all working correctly, but the second is out by a factor of about 2. I'll fully decode the bits of MSR 0x64F in about an hour. Yes, check you have latest microcode and BIOS. Also try changing constraint_0_power_limit_uw a little, just as a test. – Doug Smythies Apr 09 '21 at 19:13
  • MSR 0x64F decodes exactly as expected. For as yet unknown reasons PKG Limit #1 is servoing to the incorrect package power level. Or the information given by turbostat is incorrect. We need to go back to basics and determine the numbers for ourselves, similar to method 1 here, but even checking the conversion factor ourselves. But check micrcode and BIOS versions first. – Doug Smythies Apr 09 '21 at 20:19
  • @DougSmythies I added cpu microcode version. BIOS should not be a problem, since throttling does not occur on Windows. However if you need it, I will provide it. – igntec Apr 09 '21 at 21:13
  • @DougSmythies On Windows microcode version looks strange, I can't compare it to what Ubuntu shows. There is utility called Intel processor identification utility however I can't download it right now, because Intel download center is temporary unavailable. When I will be able to download from Intel's site, I will let you know the results. – igntec Apr 09 '21 at 21:28
  • @DougSmythies I tried to change constraint_0_power_limit_uw from 18 millions to 28 millions. The file was written successfully, however it appears to have no effect - throttling still occurs. turbostat just shows values from those files. – igntec Apr 09 '21 at 21:31
  • I am researching the basics, but got sidetracked. need msr 0x606. A long shot is to try to disable HWP via grub commandline. intel_pstate=active intel_pstate=no_hwp – Doug Smythies Apr 09 '21 at 21:40
  • @DougSmythies I tried intel_pstate=active intel_pstate=no_hwp No luck. Also it is not a clock modulation, because sudo rdmsr -a 0x19a reads 0 on all cpus all the time. Maybe it is Idle Injection Framework (introduced in kernel 4.19) that gives this behavior? – igntec Apr 10 '21 at 23:52
  • Let us continue this discussion in chat. – Doug Smythies Apr 11 '21 at 00:11
  • I don't know if it helps but I noticed a change a couple years ago where Intel Skylake showed high CPU frequency and low temperatures with no load. With moderate loads CPU frequency decreases and temperature increases a little. A dual fan laptop means I don't really hear the fans under load though. At first this behavior concerned me but after many hours of research I just didn't care anymore. – WinEunuuchs2Unix Apr 11 '21 at 02:07

4 Answers4

2

Readers: please be patient, eventually the question and answer will hopefully be reduced to some concise content.

Of all the information provided, this is the most relevant:

cpu0: MSR_RAPL_POWER_UNIT: 0x000a0e03 (0.125000 Watts, 0.000061 Joules, 0.000977 sec.)
cpu0: MSR_PKG_POWER_INFO: 0x00000078 (15 W TDP, RAPL 0 - 0 W, 0.000000 sec.)
cpu0: MSR_PKG_POWER_LIMIT: 0x42016000dc8090 (UNlocked)
cpu0: PKG Limit #1: ENabled (18.000000 Watts, 28.000000 sec, clamp DISabled)
cpu0: PKG Limit #2: DISabled (44.000000 Watts, 0.002441* sec, clamp DISabled)
...
Busy%  Bzy_MHz   IRQ  PkgTmp  PkgWatt  GFXWatt   RAMWatt
99.77   1700     17    55      7.24     0.00      0.00

rdmsr 0x64f
3d030400

Indicating that the processor is currently throttling due to Processor package power limit #1, while at the same time the processor is nowhere near package power limit #1, at least if we believe what turbostat is telling us.

Manually decoding MSR 0x64F gives, among others, bit 10 is set. From the 64-ia-32-architectures-software-developer-manual-325462:

Package/Platform-Level Power Limiting PL1 Status (R0) When set, frequency is reduced below the operating system request due to package/platform-level power limiting PL1.

The other bits are all latched log bits and not relevant just now.

The RAPL_POWER_UNIT MSR (0x606) gave 0x000a0e03, which does decode correctly to 0.125 watts per bit.

The PKG_POWER_LIMIT MSR (0x610) gave 0x42016000dc8090. Let's decode the power limit 1 part manually. Bits 14:0 define the limit, or 90h or 144 decimal. Times .125 = 18 watts, as expected. Bit 15 is 1, so it is enabled.

So, something appears to be wrong and the processor appears to be not behaving the way it should.

Could it be somehow related to the unusually high processor package limit #2? Normally it would be set to about 1.25 times power limit #1. such a high value will likely give issues.

EDIT 1: returning to MSR 0x64F, the other latched bits indicate the temperature, PROCHOT (always bad), and "OTHER" conditions had previously occurred. Myself, I would turn down power limit 2 from 44 watts to about 22 or 23 watts. Also suggest taking a long slow look at this, coming from the low side of CPU frequency. For example, set a very low CPU maximum frequency:

echo 10 | sudo tee /sys/devices/system/cpu/intel_pstate/max_perf_pct

and then apply the heavy load, all the while watching with turbostat and watching MSR 0x64f, simultaneously in two terminal windows. Gradually increase the CPU frequency:

echo 15 | sudo tee /sys/devices/system/cpu/intel_pstate/max_perf_pct
echo 20 | sudo tee /sys/devices/system/cpu/intel_pstate/max_perf_pct
echo 25 | sudo tee /sys/devices/system/cpu/intel_pstate/max_perf_pct
...

until it throttles.

EDIT 2: Clock Modulation is probably involved, if so likely from the BIOS and likely due to the extraordinarily high setting for power limit #2. The way to know is to read MSR 0x19A during the odd throttling event. Do sudo rdmsr -a 0x19a. Example:

doug@s19:~$ sudo /home/doug/temp-k-git/linux/tools/power/x86/turbostat/turbostat --Summary --quiet --show Busy%,Bzy_MHz,IRQ,PkgWatt,PkgTmp,RAMWatt,GFXWatt --interval 6
Busy%   Bzy_MHz IRQ     PkgTmp  PkgWatt GFXWatt RAMWatt
99.76   1246    72344   60      60.43   0.00    0.89
99.76   1246    72144   61      60.40   0.00    0.89  <<< Why is the CPU frequency so low at 100% load?
99.76   1246    72147   61      60.43   0.00    0.89
99.76   1246    72145   60      60.05   0.00    0.89
99.04   2931    72183   80      98.51   0.00    0.89
98.41   4507    72132   80      133.71  0.00    0.89  <<< O.K. that is more like it. Clock Modulation removed.
98.43   4509    72135   80      133.79  0.00    0.89
99.25   4483    72147   77      127.63  0.00    0.89
99.57   4466    72127   75      124.90  0.00    0.89 <<< And power limit #1 kicks in.
99.56   4461    72158   75      124.91  0.00    0.89
99.56   4463    72211   76      124.91  0.00    0.89

Meanwhile:

root@s19:/home/doug# wrmsr -a 0x19a 0x14  <<< enable CM at 25%
... apply load and start turbstat ...
root@s19:/home/doug# wrmsr -a 0x19a 0     <<< disable CM

Anticipated question: Then why does it not correct itself until after the load is removed.
Answer: Due to the brutal conditions which caused Clock Modulation to be required in the first place, it doesn't get removed until things are really calm and cool. On some computers, it never gets reset.

Doug Smythies
  • 15,448
  • 5
  • 44
  • 61
0

Short log (full log is not provided due to its size, I just will show load - no_load transition points):

sudo ./collect_cpu_info.sh



turbostat --Summary --show Busy%,Bzy_MHz,IRQ,PkgWatt,PkgTmp,RAMWatt,GFXWatt --interval 0.001 --num_iterations 1
turbostat version 20.03.20 - Len Brown <lenb@kernel.org>
CPUID(0): GenuineIntel 0x16 CPUID levels; 0x80000008 xlevels; family:model:stepping 0x6:8e:a (6:142:10)
CPUID(1): SSE3 MONITOR - EIST TM2 TSC MSR ACPI-TM HT TM
CPUID(6): APERF, TURBO, DTS, PTM, HWP, HWPnotify, HWPwindow, HWPepp, No-HWPpkg, EPB
cpu7: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MWAIT PREFETCH TURBO)
CPUID(7): SGX
cpu7: MSR_IA32_FEATURE_CONTROL: 0x00000005 (Locked )
CPUID(0x15): eax_crystal: 2 ebx_tsc: 166 ecx_crystal_hz: 0
TSC: 1992 MHz (24000000 Hz * 166 / 2 / 1000000)
CPUID(0x16): base_mhz: 2000 max_mhz: 4000 bus_mhz: 100
cpu7: MSR_MISC_PWR_MGMT: 0x00401cc0 (ENable-EIST_Coordination DISable-EPB DISable-OOB)
RAPL: 17476 sec. Joule Counter Range, at 15 Watts
cpu7: MSR_PLATFORM_INFO: 0x4043df1011400
4 * 100.0 = 400.0 MHz max efficiency frequency
20 * 100.0 = 2000.0 MHz base frequency
cpu7: MSR_IA32_POWER_CTL: 0x0024005d (C1E auto-promotion: DISabled)
cpu7: MSR_TURBO_RATIO_LIMIT: 0x25252828
37 * 100.0 = 3700.0 MHz max turbo 4 active cores
37 * 100.0 = 3700.0 MHz max turbo 3 active cores
40 * 100.0 = 4000.0 MHz max turbo 2 active cores
40 * 100.0 = 4000.0 MHz max turbo 1 active cores
cpu7: MSR_CONFIG_TDP_NOMINAL: 0x00000012 (base_ratio=18)
cpu7: MSR_CONFIG_TDP_LEVEL_1: 0x00080050 (PKG_MIN_PWR_LVL1=0 PKG_MAX_PWR_LVL1=0 LVL1_RATIO=8 PKG_TDP_LVL1=80)
cpu7: MSR_CONFIG_TDP_LEVEL_2: 0x001400c8 (PKG_MIN_PWR_LVL2=0 PKG_MAX_PWR_LVL2=0 LVL2_RATIO=20 PKG_TDP_LVL2=200)
cpu7: MSR_CONFIG_TDP_CONTROL: 0x00000000 ( lock=0)
cpu7: MSR_TURBO_ACTIVATION_RATIO: 0x00000011 (MAX_NON_TURBO_RATIO=17 lock=0)
cpu7: MSR_PKG_CST_CONFIG_CONTROL: 0x1e008008 (UNdemote-C3, UNdemote-C1, demote-C3, demote-C1, locked, pkg-cstate-limit=8 (unlimited))
cpu7: cpufreq driver: intel_pstate
cpu7: cpufreq governor: powersave
cpufreq intel_pstate no_turbo: 0
cpu7: MSR_MISC_FEATURE_CONTROL: 0x00000000 (L2-Prefetch L2-Prefetch-pair L1-Prefetch L1-IP-Prefetch)
cpu0: MSR_PM_ENABLE: 0x00000001 (HWP)
cpu0: MSR_HWP_CAPABILITIES: 0x01081228 (high 40 guar 18 eff 8 low 1)
cpu0: MSR_HWP_REQUEST: 0x80002804 (min 4 max 40 des 0 epp 0x80 window 0x0 pkg 0x0)
cpu0: MSR_HWP_INTERRUPT: 0x00000000 (Dis_Guaranteed_Perf_Change, Dis_Excursion_Min)
cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min)
cpu0: MSR_IA32_ENERGY_PERF_BIAS: 0x00000006 (balanced)
cpu0: MSR_RAPL_POWER_UNIT: 0x000a0e03 (0.125000 Watts, 0.000061 Joules, 0.000977 sec.)
cpu0: MSR_PKG_POWER_INFO: 0x00000078 (15 W TDP, RAPL 0 - 0 W, 0.000000 sec.)
cpu0: MSR_PKG_POWER_LIMIT: 0x42016000dc8090 (UNlocked)
cpu0: PKG Limit #1: ENabled (18.000000 Watts, 28.000000 sec, clamp DISabled)
cpu0: PKG Limit #2: DISabled (44.000000 Watts, 0.002441* sec, clamp DISabled)
cpu0: MSR_DRAM_POWER_LIMIT: 0x5400de00000000 (UNlocked)
cpu0: DRAM Limit: DISabled (0.000000 Watts, 0.000977 sec, clamp DISabled)
cpu0: MSR_PP0_POLICY: 0
cpu0: MSR_PP0_POWER_LIMIT: 0x00000000 (UNlocked)
cpu0: Cores Limit: DISabled (0.000000 Watts, 0.000977 sec, clamp DISabled)
cpu0: MSR_PP1_POLICY: 0
cpu0: MSR_PP1_POWER_LIMIT: 0x00000000 (UNlocked)
cpu0: GFX Limit: DISabled (0.000000 Watts, 0.000977 sec, clamp DISabled)
cpu0: MSR_IA32_TEMPERATURE_TARGET: 0x02640000 (100 C)
cpu0: MSR_IA32_PACKAGE_THERM_STATUS: 0x88370808 (45 C)
cpu0: MSR_IA32_PACKAGE_THERM_INTERRUPT: 0x00000003 (100 C, 100 C)
cpu7: MSR_PKGC3_IRTL: 0x0000884e (valid, 79872 ns)
cpu7: MSR_PKGC6_IRTL: 0x00008876 (valid, 120832 ns)
cpu7: MSR_PKGC7_IRTL: 0x00008894 (valid, 151552 ns)
cpu7: MSR_PKGC8_IRTL: 0x000088fa (valid, 256000 ns)
cpu7: MSR_PKGC9_IRTL: 0x0000894c (valid, 339968 ns)
cpu7: MSR_PKGC10_IRTL: 0x00008bf2 (valid, 1034240 ns)
Busy%   Bzy_MHz IRQ   PkgTmp  PkgWatt  GFXWatt  RAMWatt
7.41     2176    17    46      16.54    0.00       0.00


rdmsr 0x64f
3d010000


grep . /sys/devices/system/cpu/intel_pstate/*
/sys/devices/system/cpu/intel_pstate/hwp_dynamic_boost:0
/sys/devices/system/cpu/intel_pstate/max_perf_pct:100
/sys/devices/system/cpu/intel_pstate/min_perf_pct:10
/sys/devices/system/cpu/intel_pstate/no_turbo:0
/sys/devices/system/cpu/intel_pstate/num_pstates:37
/sys/devices/system/cpu/intel_pstate/status:active
/sys/devices/system/cpu/intel_pstate/turbo_pct:60


grep . /sys/class/powercap/intel-rapl/intel-rapl:0/*
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_0_max_power_uw:15000000
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_0_name:long_term
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_0_power_limit_uw:18000000
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_0_time_window_us:27983872
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_1_max_power_uw:0
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_1_name:short_term
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_1_power_limit_uw:44000000
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_1_time_window_us:2440
grep: /sys/class/powercap/intel-rapl/intel-rapl:0/device: Is a directory
/sys/class/powercap/intel-rapl/intel-rapl:0/enabled:1
/sys/class/powercap/intel-rapl/intel-rapl:0/energy_uj:31842193630
grep: /sys/class/powercap/intel-rapl/intel-rapl:0/intel-rapl:0:0: Is a directory
grep: /sys/class/powercap/intel-rapl/intel-rapl:0/intel-rapl:0:1: Is a directory
/sys/class/powercap/intel-rapl/intel-rapl:0/max_energy_range_uj:262143328850
/sys/class/powercap/intel-rapl/intel-rapl:0/name:package-0
grep: /sys/class/powercap/intel-rapl/intel-rapl:0/power: Is a directory
grep: /sys/class/powercap/intel-rapl/intel-rapl:0/subsystem: Is a directory


grep . /sys/devices/system/cpu/cpu/cpufreq/scaling_m
/sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu0/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu1/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu1/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu2/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu2/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu3/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu3/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu4/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu4/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu5/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu5/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu6/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu6/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu7/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu7/cpufreq/scaling_min_freq:400000




turbostat --Summary --show Busy%,Bzy_MHz,IRQ,PkgWatt,PkgTmp,RAMWatt,GFXWatt --interval 0.001 --num_iterations 1
turbostat version 20.03.20 - Len Brown <lenb@kernel.org>
CPUID(0): GenuineIntel 0x16 CPUID levels; 0x80000008 xlevels; family:model:stepping 0x6:8e:a (6:142:10)
CPUID(1): SSE3 MONITOR - EIST TM2 TSC MSR ACPI-TM HT TM
CPUID(6): APERF, TURBO, DTS, PTM, HWP, HWPnotify, HWPwindow, HWPepp, No-HWPpkg, EPB
cpu6: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MWAIT PREFETCH TURBO)
CPUID(7): SGX
cpu6: MSR_IA32_FEATURE_CONTROL: 0x00000005 (Locked )
CPUID(0x15): eax_crystal: 2 ebx_tsc: 166 ecx_crystal_hz: 0
TSC: 1992 MHz (24000000 Hz * 166 / 2 / 1000000)
CPUID(0x16): base_mhz: 2000 max_mhz: 4000 bus_mhz: 100
cpu6: MSR_MISC_PWR_MGMT: 0x00401cc0 (ENable-EIST_Coordination DISable-EPB DISable-OOB)
RAPL: 17476 sec. Joule Counter Range, at 15 Watts
cpu6: MSR_PLATFORM_INFO: 0x4043df1011400
4 * 100.0 = 400.0 MHz max efficiency frequency
20 * 100.0 = 2000.0 MHz base frequency
cpu6: MSR_IA32_POWER_CTL: 0x0024005d (C1E auto-promotion: DISabled)
cpu6: MSR_TURBO_RATIO_LIMIT: 0x25252828
37 * 100.0 = 3700.0 MHz max turbo 4 active cores
37 * 100.0 = 3700.0 MHz max turbo 3 active cores
40 * 100.0 = 4000.0 MHz max turbo 2 active cores
40 * 100.0 = 4000.0 MHz max turbo 1 active cores
cpu6: MSR_CONFIG_TDP_NOMINAL: 0x00000012 (base_ratio=18)
cpu6: MSR_CONFIG_TDP_LEVEL_1: 0x00080050 (PKG_MIN_PWR_LVL1=0 PKG_MAX_PWR_LVL1=0 LVL1_RATIO=8 PKG_TDP_LVL1=80)
cpu6: MSR_CONFIG_TDP_LEVEL_2: 0x001400c8 (PKG_MIN_PWR_LVL2=0 PKG_MAX_PWR_LVL2=0 LVL2_RATIO=20 PKG_TDP_LVL2=200)
cpu6: MSR_CONFIG_TDP_CONTROL: 0x00000000 ( lock=0)
cpu6: MSR_TURBO_ACTIVATION_RATIO: 0x00000011 (MAX_NON_TURBO_RATIO=17 lock=0)
cpu6: MSR_PKG_CST_CONFIG_CONTROL: 0x1e008008 (UNdemote-C3, UNdemote-C1, demote-C3, demote-C1, locked, pkg-cstate-limit=8 (unlimited))
cpu6: cpufreq driver: intel_pstate
cpu6: cpufreq governor: powersave
cpufreq intel_pstate no_turbo: 0
cpu6: MSR_MISC_FEATURE_CONTROL: 0x00000000 (L2-Prefetch L2-Prefetch-pair L1-Prefetch L1-IP-Prefetch)
cpu0: MSR_PM_ENABLE: 0x00000001 (HWP)
cpu0: MSR_HWP_CAPABILITIES: 0x01091228 (high 40 guar 18 eff 9 low 1)
cpu0: MSR_HWP_REQUEST: 0x80002804 (min 4 max 40 des 0 epp 0x80 window 0x0 pkg 0x0)
cpu0: MSR_HWP_INTERRUPT: 0x00000000 (Dis_Guaranteed_Perf_Change, Dis_Excursion_Min)
cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min)
cpu0: MSR_IA32_ENERGY_PERF_BIAS: 0x00000006 (balanced)
cpu0: MSR_RAPL_POWER_UNIT: 0x000a0e03 (0.125000 Watts, 0.000061 Joules, 0.000977 sec.)
cpu0: MSR_PKG_POWER_INFO: 0x00000078 (15 W TDP, RAPL 0 - 0 W, 0.000000 sec.)
cpu0: MSR_PKG_POWER_LIMIT: 0x42016000dc8090 (UNlocked)
cpu0: PKG Limit #1: ENabled (18.000000 Watts, 28.000000 sec, clamp DISabled)
cpu0: PKG Limit #2: DISabled (44.000000 Watts, 0.002441* sec, clamp DISabled)
cpu0: MSR_DRAM_POWER_LIMIT: 0x5400de00000000 (UNlocked)
cpu0: DRAM Limit: DISabled (0.000000 Watts, 0.000977 sec, clamp DISabled)
cpu0: MSR_PP0_POLICY: 0
cpu0: MSR_PP0_POWER_LIMIT: 0x00000000 (UNlocked)
cpu0: Cores Limit: DISabled (0.000000 Watts, 0.000977 sec, clamp DISabled)
cpu0: MSR_PP1_POLICY: 0
cpu0: MSR_PP1_POWER_LIMIT: 0x00000000 (UNlocked)
cpu0: GFX Limit: DISabled (0.000000 Watts, 0.000977 sec, clamp DISabled)
cpu0: MSR_IA32_TEMPERATURE_TARGET: 0x02640000 (100 C)
cpu0: MSR_IA32_PACKAGE_THERM_STATUS: 0x88210808 (67 C)
cpu0: MSR_IA32_PACKAGE_THERM_INTERRUPT: 0x00000003 (100 C, 100 C)
cpu6: MSR_PKGC3_IRTL: 0x0000884e (valid, 79872 ns)
cpu6: MSR_PKGC6_IRTL: 0x00008876 (valid, 120832 ns)
cpu6: MSR_PKGC7_IRTL: 0x00008894 (valid, 151552 ns)
cpu6: MSR_PKGC8_IRTL: 0x000088fa (valid, 256000 ns)
cpu6: MSR_PKGC9_IRTL: 0x0000894c (valid, 339968 ns)
cpu6: MSR_PKGC10_IRTL: 0x00008bf2 (valid, 1034240 ns)
Busy%  Bzy_MHz  IRQ   PkgTmp  PkgWatt  GFXWatt   RAMWatt
97.55   3700     9     68      43.25    0.00      0.00


rdmsr 0x64f
3d011000


grep . /sys/devices/system/cpu/intel_pstate/*
/sys/devices/system/cpu/intel_pstate/hwp_dynamic_boost:0
/sys/devices/system/cpu/intel_pstate/max_perf_pct:100
/sys/devices/system/cpu/intel_pstate/min_perf_pct:10
/sys/devices/system/cpu/intel_pstate/no_turbo:0
/sys/devices/system/cpu/intel_pstate/num_pstates:37
/sys/devices/system/cpu/intel_pstate/status:active
/sys/devices/system/cpu/intel_pstate/turbo_pct:60


grep . /sys/class/powercap/intel-rapl/intel-rapl:0/*
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_0_max_power_uw:15000000
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_0_name:long_term
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_0_power_limit_uw:18000000
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_0_time_window_us:27983872
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_1_max_power_uw:0
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_1_name:short_term
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_1_power_limit_uw:44000000
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_1_time_window_us:2440
grep: /sys/class/powercap/intel-rapl/intel-rapl:0/device: Is a directory
/sys/class/powercap/intel-rapl/intel-rapl:0/enabled:1
/sys/class/powercap/intel-rapl/intel-rapl:0/energy_uj:31914117396
grep: /sys/class/powercap/intel-rapl/intel-rapl:0/intel-rapl:0:0: Is a directory
grep: /sys/class/powercap/intel-rapl/intel-rapl:0/intel-rapl:0:1: Is a directory
/sys/class/powercap/intel-rapl/intel-rapl:0/max_energy_range_uj:262143328850
/sys/class/powercap/intel-rapl/intel-rapl:0/name:package-0
grep: /sys/class/powercap/intel-rapl/intel-rapl:0/power: Is a directory
grep: /sys/class/powercap/intel-rapl/intel-rapl:0/subsystem: Is a directory


grep . /sys/devices/system/cpu/cpu/cpufreq/scaling_m
/sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu0/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu1/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu1/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu2/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu2/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu3/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu3/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu4/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu4/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu5/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu5/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu6/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu6/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu7/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu7/cpufreq/scaling_min_freq:400000




turbostat --Summary --show Busy%,Bzy_MHz,IRQ,PkgWatt,PkgTmp,RAMWatt,GFXWatt --interval 0.001 --num_iterations 1
turbostat version 20.03.20 - Len Brown <lenb@kernel.org>
CPUID(0): GenuineIntel 0x16 CPUID levels; 0x80000008 xlevels; family:model:stepping 0x6:8e:a (6:142:10)
CPUID(1): SSE3 MONITOR - EIST TM2 TSC MSR ACPI-TM HT TM
CPUID(6): APERF, TURBO, DTS, PTM, HWP, HWPnotify, HWPwindow, HWPepp, No-HWPpkg, EPB
cpu6: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MWAIT PREFETCH TURBO)
CPUID(7): SGX
cpu6: MSR_IA32_FEATURE_CONTROL: 0x00000005 (Locked )
CPUID(0x15): eax_crystal: 2 ebx_tsc: 166 ecx_crystal_hz: 0
TSC: 1992 MHz (24000000 Hz * 166 / 2 / 1000000)
CPUID(0x16): base_mhz: 2000 max_mhz: 4000 bus_mhz: 100
cpu6: MSR_MISC_PWR_MGMT: 0x00401cc0 (ENable-EIST_Coordination DISable-EPB DISable-OOB)
RAPL: 17476 sec. Joule Counter Range, at 15 Watts
cpu6: MSR_PLATFORM_INFO: 0x4043df1011400
4 * 100.0 = 400.0 MHz max efficiency frequency
20 * 100.0 = 2000.0 MHz base frequency
cpu6: MSR_IA32_POWER_CTL: 0x0024005d (C1E auto-promotion: DISabled)
cpu6: MSR_TURBO_RATIO_LIMIT: 0x25252828
37 * 100.0 = 3700.0 MHz max turbo 4 active cores
37 * 100.0 = 3700.0 MHz max turbo 3 active cores
40 * 100.0 = 4000.0 MHz max turbo 2 active cores
40 * 100.0 = 4000.0 MHz max turbo 1 active cores
cpu6: MSR_CONFIG_TDP_NOMINAL: 0x00000012 (base_ratio=18)
cpu6: MSR_CONFIG_TDP_LEVEL_1: 0x00080050 (PKG_MIN_PWR_LVL1=0 PKG_MAX_PWR_LVL1=0 LVL1_RATIO=8 PKG_TDP_LVL1=80)
cpu6: MSR_CONFIG_TDP_LEVEL_2: 0x001400c8 (PKG_MIN_PWR_LVL2=0 PKG_MAX_PWR_LVL2=0 LVL2_RATIO=20 PKG_TDP_LVL2=200)
cpu6: MSR_CONFIG_TDP_CONTROL: 0x00000000 ( lock=0)
cpu6: MSR_TURBO_ACTIVATION_RATIO: 0x00000011 (MAX_NON_TURBO_RATIO=17 lock=0)
cpu6: MSR_PKG_CST_CONFIG_CONTROL: 0x1e008008 (UNdemote-C3, UNdemote-C1, demote-C3, demote-C1, locked, pkg-cstate-limit=8 (unlimited))
cpu6: cpufreq driver: intel_pstate
cpu6: cpufreq governor: powersave
cpufreq intel_pstate no_turbo: 0
cpu6: MSR_MISC_FEATURE_CONTROL: 0x00000000 (L2-Prefetch L2-Prefetch-pair L1-Prefetch L1-IP-Prefetch)
cpu0: MSR_PM_ENABLE: 0x00000001 (HWP)
cpu0: MSR_HWP_CAPABILITIES: 0x01091228 (high 40 guar 18 eff 9 low 1)
cpu0: MSR_HWP_REQUEST: 0x80002804 (min 4 max 40 des 0 epp 0x80 window 0x0 pkg 0x0)
cpu0: MSR_HWP_INTERRUPT: 0x00000000 (Dis_Guaranteed_Perf_Change, Dis_Excursion_Min)
cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min)
cpu0: MSR_IA32_ENERGY_PERF_BIAS: 0x00000006 (balanced)
cpu0: MSR_RAPL_POWER_UNIT: 0x000a0e03 (0.125000 Watts, 0.000061 Joules, 0.000977 sec.)
cpu0: MSR_PKG_POWER_INFO: 0x00000078 (15 W TDP, RAPL 0 - 0 W, 0.000000 sec.)
cpu0: MSR_PKG_POWER_LIMIT: 0x42016000dc8090 (UNlocked)
cpu0: PKG Limit #1: ENabled (18.000000 Watts, 28.000000 sec, clamp DISabled)
cpu0: PKG Limit #2: DISabled (44.000000 Watts, 0.002441* sec, clamp DISabled)
cpu0: MSR_DRAM_POWER_LIMIT: 0x5400de00000000 (UNlocked)
cpu0: DRAM Limit: DISabled (0.000000 Watts, 0.000977 sec, clamp DISabled)
cpu0: MSR_PP0_POLICY: 0
cpu0: MSR_PP0_POWER_LIMIT: 0x00000000 (UNlocked)
cpu0: Cores Limit: DISabled (0.000000 Watts, 0.000977 sec, clamp DISabled)
cpu0: MSR_PP1_POLICY: 0
cpu0: MSR_PP1_POWER_LIMIT: 0x00000000 (UNlocked)
cpu0: GFX Limit: DISabled (0.000000 Watts, 0.000977 sec, clamp DISabled)
cpu0: MSR_IA32_TEMPERATURE_TARGET: 0x02640000 (100 C)
cpu0: MSR_IA32_PACKAGE_THERM_STATUS: 0x882d0e88 (55 C)
cpu0: MSR_IA32_PACKAGE_THERM_INTERRUPT: 0x00000003 (100 C, 100 C)
cpu6: MSR_PKGC3_IRTL: 0x0000884e (valid, 79872 ns)
cpu6: MSR_PKGC6_IRTL: 0x00008876 (valid, 120832 ns)
cpu6: MSR_PKGC7_IRTL: 0x00008894 (valid, 151552 ns)
cpu6: MSR_PKGC8_IRTL: 0x000088fa (valid, 256000 ns)
cpu6: MSR_PKGC9_IRTL: 0x0000894c (valid, 339968 ns)
cpu6: MSR_PKGC10_IRTL: 0x00008bf2 (valid, 1034240 ns)
Busy%  Bzy_MHz   IRQ  PkgTmp  PkgWatt  GFXWatt   RAMWatt
99.77   1700     17    55      7.24     0.00      0.00


rdmsr 0x64f
3d030400


grep . /sys/devices/system/cpu/intel_pstate/*
/sys/devices/system/cpu/intel_pstate/hwp_dynamic_boost:0
/sys/devices/system/cpu/intel_pstate/max_perf_pct:100
/sys/devices/system/cpu/intel_pstate/min_perf_pct:10
/sys/devices/system/cpu/intel_pstate/no_turbo:0
/sys/devices/system/cpu/intel_pstate/num_pstates:37
/sys/devices/system/cpu/intel_pstate/status:active
/sys/devices/system/cpu/intel_pstate/turbo_pct:60


grep . /sys/class/powercap/intel-rapl/intel-rapl:0/*
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_0_max_power_uw:15000000
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_0_name:long_term
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_0_power_limit_uw:18000000
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_0_time_window_us:27983872
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_1_max_power_uw:0
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_1_name:short_term
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_1_power_limit_uw:44000000
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_1_time_window_us:2440
grep: /sys/class/powercap/intel-rapl/intel-rapl:0/device: Is a directory
/sys/class/powercap/intel-rapl/intel-rapl:0/enabled:1
/sys/class/powercap/intel-rapl/intel-rapl:0/energy_uj:32027108134
grep: /sys/class/powercap/intel-rapl/intel-rapl:0/intel-rapl:0:0: Is a directory
grep: /sys/class/powercap/intel-rapl/intel-rapl:0/intel-rapl:0:1: Is a directory
/sys/class/powercap/intel-rapl/intel-rapl:0/max_energy_range_uj:262143328850
/sys/class/powercap/intel-rapl/intel-rapl:0/name:package-0
grep: /sys/class/powercap/intel-rapl/intel-rapl:0/power: Is a directory
grep: /sys/class/powercap/intel-rapl/intel-rapl:0/subsystem: Is a directory


grep . /sys/devices/system/cpu/cpu/cpufreq/scaling_m
/sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu0/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu1/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu1/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu2/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu2/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu3/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu3/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu4/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu4/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu5/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu5/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu6/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu6/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu7/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu7/cpufreq/scaling_min_freq:400000




turbostat --Summary --show Busy%,Bzy_MHz,IRQ,PkgWatt,PkgTmp,RAMWatt,GFXWatt --interval 0.001 --num_iterations 1
turbostat version 20.03.20 - Len Brown <lenb@kernel.org>
CPUID(0): GenuineIntel 0x16 CPUID levels; 0x80000008 xlevels; family:model:stepping 0x6:8e:a (6:142:10)
CPUID(1): SSE3 MONITOR - EIST TM2 TSC MSR ACPI-TM HT TM
CPUID(6): APERF, TURBO, DTS, PTM, HWP, HWPnotify, HWPwindow, HWPepp, No-HWPpkg, EPB
cpu7: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MWAIT PREFETCH TURBO)
CPUID(7): SGX
cpu7: MSR_IA32_FEATURE_CONTROL: 0x00000005 (Locked )
CPUID(0x15): eax_crystal: 2 ebx_tsc: 166 ecx_crystal_hz: 0
TSC: 1992 MHz (24000000 Hz * 166 / 2 / 1000000)
CPUID(0x16): base_mhz: 2000 max_mhz: 4000 bus_mhz: 100
cpu7: MSR_MISC_PWR_MGMT: 0x00401cc0 (ENable-EIST_Coordination DISable-EPB DISable-OOB)
RAPL: 17476 sec. Joule Counter Range, at 15 Watts
cpu7: MSR_PLATFORM_INFO: 0x4043df1011400
4 * 100.0 = 400.0 MHz max efficiency frequency
20 * 100.0 = 2000.0 MHz base frequency
cpu7: MSR_IA32_POWER_CTL: 0x0024005d (C1E auto-promotion: DISabled)
cpu7: MSR_TURBO_RATIO_LIMIT: 0x25252828
37 * 100.0 = 3700.0 MHz max turbo 4 active cores
37 * 100.0 = 3700.0 MHz max turbo 3 active cores
40 * 100.0 = 4000.0 MHz max turbo 2 active cores
40 * 100.0 = 4000.0 MHz max turbo 1 active cores
cpu7: MSR_CONFIG_TDP_NOMINAL: 0x00000012 (base_ratio=18)
cpu7: MSR_CONFIG_TDP_LEVEL_1: 0x00080050 (PKG_MIN_PWR_LVL1=0 PKG_MAX_PWR_LVL1=0 LVL1_RATIO=8 PKG_TDP_LVL1=80)
cpu7: MSR_CONFIG_TDP_LEVEL_2: 0x001400c8 (PKG_MIN_PWR_LVL2=0 PKG_MAX_PWR_LVL2=0 LVL2_RATIO=20 PKG_TDP_LVL2=200)
cpu7: MSR_CONFIG_TDP_CONTROL: 0x00000000 ( lock=0)
cpu7: MSR_TURBO_ACTIVATION_RATIO: 0x00000011 (MAX_NON_TURBO_RATIO=17 lock=0)
cpu7: MSR_PKG_CST_CONFIG_CONTROL: 0x1e008008 (UNdemote-C3, UNdemote-C1, demote-C3, demote-C1, locked, pkg-cstate-limit=8 (unlimited))
cpu7: cpufreq driver: intel_pstate
cpu7: cpufreq governor: powersave
cpufreq intel_pstate no_turbo: 0
cpu7: MSR_MISC_FEATURE_CONTROL: 0x00000000 (L2-Prefetch L2-Prefetch-pair L1-Prefetch L1-IP-Prefetch)
cpu0: MSR_PM_ENABLE: 0x00000001 (HWP)
cpu0: MSR_HWP_CAPABILITIES: 0x01081228 (high 40 guar 18 eff 8 low 1)
cpu0: MSR_HWP_REQUEST: 0x80002804 (min 4 max 40 des 0 epp 0x80 window 0x0 pkg 0x0)
cpu0: MSR_HWP_INTERRUPT: 0x00000000 (Dis_Guaranteed_Perf_Change, Dis_Excursion_Min)
cpu0: MSR_HWP_STATUS: 0x00000000 (No-Guaranteed_Perf_Change, No-Excursion_Min)
cpu0: MSR_IA32_ENERGY_PERF_BIAS: 0x00000006 (balanced)
cpu0: MSR_RAPL_POWER_UNIT: 0x000a0e03 (0.125000 Watts, 0.000061 Joules, 0.000977 sec.)
cpu0: MSR_PKG_POWER_INFO: 0x00000078 (15 W TDP, RAPL 0 - 0 W, 0.000000 sec.)
cpu0: MSR_PKG_POWER_LIMIT: 0x42016000dc8090 (UNlocked)
cpu0: PKG Limit #1: ENabled (18.000000 Watts, 28.000000 sec, clamp DISabled)
cpu0: PKG Limit #2: DISabled (44.000000 Watts, 0.002441* sec, clamp DISabled)
cpu0: MSR_DRAM_POWER_LIMIT: 0x5400de00000000 (UNlocked)
cpu0: DRAM Limit: DISabled (0.000000 Watts, 0.000977 sec, clamp DISabled)
cpu0: MSR_PP0_POLICY: 0
cpu0: MSR_PP0_POWER_LIMIT: 0x00000000 (UNlocked)
cpu0: Cores Limit: DISabled (0.000000 Watts, 0.000977 sec, clamp DISabled)
cpu0: MSR_PP1_POLICY: 0
cpu0: MSR_PP1_POWER_LIMIT: 0x00000000 (UNlocked)
cpu0: GFX Limit: DISabled (0.000000 Watts, 0.000977 sec, clamp DISabled)
cpu0: MSR_IA32_TEMPERATURE_TARGET: 0x02640000 (100 C)
cpu0: MSR_IA32_PACKAGE_THERM_STATUS: 0x88340808 (48 C)
cpu0: MSR_IA32_PACKAGE_THERM_INTERRUPT: 0x00000003 (100 C, 100 C)
cpu7: MSR_PKGC3_IRTL: 0x0000884e (valid, 79872 ns)
cpu7: MSR_PKGC6_IRTL: 0x00008876 (valid, 120832 ns)
cpu7: MSR_PKGC7_IRTL: 0x00008894 (valid, 151552 ns)
cpu7: MSR_PKGC8_IRTL: 0x000088fa (valid, 256000 ns)
cpu7: MSR_PKGC9_IRTL: 0x0000894c (valid, 339968 ns)
cpu7: MSR_PKGC10_IRTL: 0x00008bf2 (valid, 1034240 ns)
Busy%   Bzy_MHz  IRQ   PkgTmp  PkgWatt  GFXWatt   RAMWatt
17.53    2200     1     48      1.36     0.00      0.00


rdmsr 0x64f
3d030000


grep . /sys/devices/system/cpu/intel_pstate/*
/sys/devices/system/cpu/intel_pstate/hwp_dynamic_boost:0
/sys/devices/system/cpu/intel_pstate/max_perf_pct:100
/sys/devices/system/cpu/intel_pstate/min_perf_pct:10
/sys/devices/system/cpu/intel_pstate/no_turbo:0
/sys/devices/system/cpu/intel_pstate/num_pstates:37
/sys/devices/system/cpu/intel_pstate/status:active
/sys/devices/system/cpu/intel_pstate/turbo_pct:60


grep . /sys/class/powercap/intel-rapl/intel-rapl:0/*
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_0_max_power_uw:15000000
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_0_name:long_term
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_0_power_limit_uw:18000000
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_0_time_window_us:27983872
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_1_max_power_uw:0
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_1_name:short_term
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_1_power_limit_uw:44000000
/sys/class/powercap/intel-rapl/intel-rapl:0/constraint_1_time_window_us:2440
grep: /sys/class/powercap/intel-rapl/intel-rapl:0/device: Is a directory
/sys/class/powercap/intel-rapl/intel-rapl:0/enabled:1
/sys/class/powercap/intel-rapl/intel-rapl:0/energy_uj:32210401183
grep: /sys/class/powercap/intel-rapl/intel-rapl:0/intel-rapl:0:0: Is a directory
grep: /sys/class/powercap/intel-rapl/intel-rapl:0/intel-rapl:0:1: Is a directory
/sys/class/powercap/intel-rapl/intel-rapl:0/max_energy_range_uj:262143328850
/sys/class/powercap/intel-rapl/intel-rapl:0/name:package-0
grep: /sys/class/powercap/intel-rapl/intel-rapl:0/power: Is a directory
grep: /sys/class/powercap/intel-rapl/intel-rapl:0/subsystem: Is a directory


grep . /sys/devices/system/cpu/cpu/cpufreq/scaling_m
/sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu0/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu1/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu1/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu2/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu2/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu3/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu3/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu4/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu4/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu5/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu5/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu6/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu6/cpufreq/scaling_min_freq:400000
/sys/devices/system/cpu/cpu7/cpufreq/scaling_max_freq:4000000
/sys/devices/system/cpu/cpu7/cpufreq/scaling_min_freq:400000

First lines of dmesg ouput:

[    0.000000] microcode: microcode updated early to revision 0xe0, date = 2020-06-17
[    0.000000] Linux version 5.8.0-48-generic (buildd@lgw01-amd64-008) (gcc (Ubuntu 9.3.0-17ubuntu1~20.04) 9.3.0, GNU ld (GNU Binutils for Ubuntu) 2.34) #54~20.04.1-Ubuntu SMP Sat Mar 20 13:40:25 UTC 2021 (Ubuntu 5.8.0-48.54~20.04.1-generic 5.8.18)
[    0.000000] Command line: BOOT_IMAGE=/boot/vmlinuz-5.8.0-48-generic root=UUID=19c1475f-df39-48e7-947c-ace73270909e ro quiet splash vt.handoff=7

Also I found a lot of ACPI errors in dmesg output, here is the first one:

[    3.055366] ACPI BIOS Error (bug): AE_AML_PACKAGE_LIMIT, Index (0x000000005) is beyond end of object (length 0x5) (20200528/exoparg2-393)
igntec
  • 121
  • 3
0

Also I found bluetooth erros in dmesg output:

[    7.747559] Bluetooth: hci0: command 0xfc05 tx timeout
[    7.747568] Bluetooth: hci0: Reading Intel version information failed (-110)
[    7.747569] Bluetooth: hci0: Intel Read version failed (-110)

I see that cpu thermal monitoring is enabled:

[    0.165359] mce: CPU0: Thermal monitoring enabled (TM1)

and this strange line:

[    6.139480] thermal thermal_zone10: failed to read out thermal zone (-61)

CPU microcode version on Ubuntu:

grep 'model name\|microcode' /proc/cpuinfo

model name  : Intel(R) Core(TM) i7-8550U CPU @ 1.80GHz
microcode   : 0xe0
model name  : Intel(R) Core(TM) i7-8550U CPU @ 1.80GHz
microcode   : 0xe0
model name  : Intel(R) Core(TM) i7-8550U CPU @ 1.80GHz
microcode   : 0xe0
model name  : Intel(R) Core(TM) i7-8550U CPU @ 1.80GHz
microcode   : 0xe0
model name  : Intel(R) Core(TM) i7-8550U CPU @ 1.80GHz
microcode   : 0xe0
model name  : Intel(R) Core(TM) i7-8550U CPU @ 1.80GHz
microcode   : 0xe0
model name  : Intel(R) Core(TM) i7-8550U CPU @ 1.80GHz
microcode   : 0xe0
model name  : Intel(R) Core(TM) i7-8550U CPU @ 1.80GHz
microcode   : 0xe0

CPU microcode version on Windows:

reg query HKLM\HARDWARE\DESCRIPTION\System\CentralProcessor\0 | find "Update Revision"

Update Revision             REG_BINARY    00000000B4000000
Previous Update Revision    REG_BINARY    0000000084000000

Update Revision is current microcode version. Previous Update Revision is microcode version hardcoded in BIOS.

Intel Processor Identification Utility on Windows shows that CPU Revision is indeed B4. So actually Windows was running older version of microcode.

However, power throttling still occurs, after downgrading microcode on Ubuntu from E0 to B4. So buggy microcode is not a problem here.

I removed intel_powerclamp module (just in case):

sudo rmmod intel_powerclamp

In top output I see idle_inject processes for each cpu core:

idle_inject/0
idle_inject/1
idle_inject/2
idle_inject/3

It is Idle Injection Framework (CONFIG_IDLE_INJECT) that was introduced in linux kernel 4.19

It is meant to power clamp the system, so probably it gives such power drop?

igntec
  • 121
  • 3
0

After Ubuntu 20.04 update everything seems to work fine.

sudo stress --cpu 8

System Monitor shows sudden jump to 100% cpu load, and it holds at 100% until I terminate stress app.

uname -r
5.11.0-25-generic

lsb_release -a
No LSB modules are available.
Distributor ID: Ubuntu
Description:    Ubuntu 20.04.2 LTS
Release:    20.04
Codename:   focal
igntec
  • 121
  • 3