2010-10-23 11:53:03 +08:00
.TH TURBOSTAT 8
.SH NAME
turbostat \- Report processor frequency and idle statistics
.SH SYNOPSIS
.ft B
.B turbostat
2012-09-22 13:25:08 +08:00
.RB [ Options ]
2010-10-23 11:53:03 +08:00
.RB command
.br
.B turbostat
2012-09-22 13:25:08 +08:00
.RB [ Options ]
2015-02-10 14:56:38 +08:00
.RB [ "\--interval seconds" ]
2010-10-23 11:53:03 +08:00
.SH DESCRIPTION
tools/power turbostat: v3.0: monitor Watts and Temperature
Show power in Watts and temperature in Celsius
when hardware support is present.
Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL
(Run-Time-Average-Power-Limiting). Per the Intel SDM
(Intel® 64 and IA-32 Architectures Software Developer Manual)
RAPL provides hardware energy counters and power control MSRs
(Model Specific Registers). RAPL MSRs are designed primarily
as a method to implement power capping. However, they are useful
for monitoring system power whether or not power capping is used.
In addition, Turbostat now shows temperature from DTS
(Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware,
if present.
As before, turbostat reads MSRs, and never writes MSRs.
New columns are present in turbostat output:
The Pkg_W column shows Watts for each package (socket) in the system.
On multi-socket systems, the system summary on the 1st row shows the sum
for all sockets together.
The Cor_W column shows Watts due to processors cores.
Note that Core_W is included in Pkg_W.
The optional GFX_W column shows Watts due to the graphics "un-core".
Note that GFX_W is included in Pkg_W.
The optional RAM_W column on server processors shows Watts due to DRAM DIMMS.
As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W.
The optional PKG_% and RAM_% columns on server processors shows the % of time
in the measurement interval that RAPL power limiting is in effect on the
package and on DRAM.
Note that the RAPL energy counters have some limitations.
First, hardware updates the counters about once every milli-second.
This is fine for typical turbostat measurement intervals > 1 sec.
However, when turbostat is used to measure events that approach
1ms, the counters are less useful.
Second, the 32-bit energy counters are subject to wrapping.
For example, a counter incrementing 15 micro-Joule units
on a 130 Watt TDP server processor could (in theory)
roll over in about 9 minutes. Turbostat detects and handles
up to 1 counter overflow per measurement interval.
But when the measurement interval exceeds the guaranteed
counter range, we can't detect if more than 1 overflow occured.
So in this case turbostat indicates that the results are
in question by replacing the fractional part of the Watts
in the output with "**":
Pkg_W Cor_W GFX_W
3** 0** 0**
Third, the RAPL counters are energy (Joule) counters -- they sum up
weighted events in the package to estimate energy consumed. They are
not analong power (Watt) meters. In practice, they tend to under-count
because they don't cover every possible use of energy in the package.
The accuracy of the RAPL counters will vary between product generations,
and between SKU's in the same product generation, and with temperature.
turbostat's -v (verbose) option now displays more power and thermal configuration
information -- as shown on the turbostat.8 manual page.
For example, it now displays the Package and DRAM Thermal Design Power (TDP):
cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.)
cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.)
cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.)
cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.)
Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
\fB turbostat \fP reports processor topology, frequency,
2015-01-23 14:33:58 +08:00
idle power-state statistics, temperature and power on X86 processors.
There are two ways to invoke turbostat.
The first method is to supply a
\fB command\fP , which is forked and statistics are printed
upon its completion.
The second method is to omit the command,
2015-02-10 14:56:38 +08:00
and turbostat displays statistics every 5 seconds.
The 5-second interval can be changed using the --interval option.
2015-02-23 13:34:57 +08:00
.PP
2015-02-10 14:56:38 +08:00
Some information is not available on older processors.
2010-10-23 11:53:03 +08:00
.SS Options
2015-02-23 13:34:57 +08:00
Options can be specified with a single or double '-', and only as much of the option
name as necessary to disambiguate it from others is necessary. Note that options are case-sensitive.
2015-02-10 14:56:38 +08:00
\fB --Counter MSR#\fP shows the delta of the specified 64-bit MSR counter.
tools/power: turbostat v2 - re-write for efficiency
Measuring large profoundly-idle configurations
requires turbostat to be more lightweight.
Otherwise, the operation of turbostat itself
can interfere with the measurements.
This re-write makes turbostat topology aware.
Hardware is accessed in "topology order".
Redundant hardware accesses are deleted.
Redundant output is deleted.
Also, output is buffered and
local RDTSC use replaces remote MSR access for TSC.
From a feature point of view, the output
looks different since redundant figures are absent.
Also, there are now -c and -p options -- to restrict
output to the 1st thread in each core, and the 1st
thread in each package, respectively. This is helpful
to reduce output on big systems, where more detail
than the "-s" system summary is desired.
Finally, periodic mode output is now on stdout, not stderr.
Turbostat v2 is also slightly more robust in
handling run-time CPU online/offline events,
as it now checks the actual map of on-line cpus rather
than just the total number of on-line cpus.
Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
.PP
2015-02-10 14:56:38 +08:00
\fB --counter MSR#\fP shows the delta of the specified 32-bit MSR counter.
tools/power: turbostat v2 - re-write for efficiency
Measuring large profoundly-idle configurations
requires turbostat to be more lightweight.
Otherwise, the operation of turbostat itself
can interfere with the measurements.
This re-write makes turbostat topology aware.
Hardware is accessed in "topology order".
Redundant hardware accesses are deleted.
Redundant output is deleted.
Also, output is buffered and
local RDTSC use replaces remote MSR access for TSC.
From a feature point of view, the output
looks different since redundant figures are absent.
Also, there are now -c and -p options -- to restrict
output to the 1st thread in each core, and the 1st
thread in each package, respectively. This is helpful
to reduce output on big systems, where more detail
than the "-s" system summary is desired.
Finally, periodic mode output is now on stdout, not stderr.
Turbostat v2 is also slightly more robust in
handling run-time CPU online/offline events,
as it now checks the actual map of on-line cpus rather
than just the total number of on-line cpus.
Signed-off-by: Len Brown <len.brown@intel.com>
2012-06-04 12:56:40 +08:00
.PP
2015-02-10 14:56:38 +08:00
\fB --Dump\fP displays the raw counter values.
2012-02-07 07:37:16 +08:00
.PP
2015-02-10 14:56:38 +08:00
\fB --debug\fP displays additional system configuration information. Invoking this parameter
more than once may also enable internal turbostat debug information.
2010-10-23 11:53:03 +08:00
.PP
2015-02-10 14:56:38 +08:00
\fB --interval seconds\fP overrides the default 5-second measurement interval.
2012-10-07 03:26:31 +08:00
.PP
2015-02-10 14:56:38 +08:00
\fB --help\fP displays usage for the most common parameters.
2012-09-22 13:25:08 +08:00
.PP
2015-02-10 14:56:38 +08:00
\fB --Joules\fP displays energy in Joules, rather than dividing Joules by time to print power in Watts.
2012-09-22 13:25:08 +08:00
.PP
2015-02-10 14:56:38 +08:00
\fB --MSR MSR#\fP shows the specified 64-bit MSR value.
2010-10-23 11:53:03 +08:00
.PP
2015-02-10 14:56:38 +08:00
\fB --msr MSR#\fP shows the specified 32-bit MSR value.
2010-10-23 11:53:03 +08:00
.PP
2015-02-10 14:56:38 +08:00
\fB --Package\fP limits output to the system summary plus the 1st thread in each Package.
.PP
\fB --processor\fP limits output to the system summary plus the 1st thread in each processor of each package. Ie. it skips hyper-threaded siblings.
.PP
\fB --Summary\fP limits output to a 1-line System Summary for each interval.
.PP
\fB --TCC temperature\fP sets the Thermal Control Circuit temperature for systems which do not export that value. This is used for making sense of the Digital Thermal Sensor outputs, as they return degrees Celsius below the TCC activation temperature.
.PP
\fB --version\fP displays the version.
.PP
The \fB command\fP parameter forks \fB command\fP , and upon its exit,
2010-10-23 11:53:03 +08:00
displays the statistics gathered since it was forked.
.PP
2015-02-23 13:34:57 +08:00
.SH DEFAULT FIELD DESCRIPTIONS
2010-10-23 11:53:03 +08:00
.nf
2015-02-23 13:34:57 +08:00
\fB CPU\fP Linux CPU (logical processor) number. Yes, it is okay that on many systems the CPUs are not listed in numerical order -- for efficiency reasons, turbostat runs in topology order, so HT siblings appear together.
2014-02-06 13:55:19 +08:00
\fB AVG_MHz\fP number of cycles executed divided by time elapsed.
2015-02-10 14:56:38 +08:00
\fB %Busy\fP percent of the interval that the CPU retired instructions, aka. % of time in "C0" state.
2014-02-06 13:55:19 +08:00
\fB Bzy_MHz\fP average clock rate while the CPU was busy (in "c0" state).
\fB TSC_MHz\fP average MHz that the TSC ran during the entire interval.
2015-02-23 13:34:57 +08:00
.fi
.PP
.SH DEBUG FIELD DESCRIPTIONS
.nf
\fB Package\fP processor package number.
\fB Core\fP processor core number.
Note that multiple CPUs per core indicate support for Intel(R) Hyper-Threading Technology (HT).
2014-02-06 13:55:19 +08:00
\fB CPU%c1, CPU%c3, CPU%c6, CPU%c7\fP show the percentage residency in hardware core idle states.
\fB CoreTmp\fP Degrees Celsius reported by the per-core Digital Thermal Sensor.
\fB PkgTtmp\fP Degrees Celsius reported by the per-package Package Thermal Monitor.
\fB Pkg%pc2, Pkg%pc3, Pkg%pc6, Pkg%pc7\fP percentage residency in hardware package idle states.
\fB PkgWatt\fP Watts consumed by the whole package.
\fB CorWatt\fP Watts consumed by the core part of the package.
\fB GFXWatt\fP Watts consumed by the Graphics part of the package -- available only on client processors.
\fB RAMWatt\fP Watts consumed by the DRAM DIMMS -- available only on server processors.
tools/power turbostat: v3.0: monitor Watts and Temperature
Show power in Watts and temperature in Celsius
when hardware support is present.
Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL
(Run-Time-Average-Power-Limiting). Per the Intel SDM
(Intel® 64 and IA-32 Architectures Software Developer Manual)
RAPL provides hardware energy counters and power control MSRs
(Model Specific Registers). RAPL MSRs are designed primarily
as a method to implement power capping. However, they are useful
for monitoring system power whether or not power capping is used.
In addition, Turbostat now shows temperature from DTS
(Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware,
if present.
As before, turbostat reads MSRs, and never writes MSRs.
New columns are present in turbostat output:
The Pkg_W column shows Watts for each package (socket) in the system.
On multi-socket systems, the system summary on the 1st row shows the sum
for all sockets together.
The Cor_W column shows Watts due to processors cores.
Note that Core_W is included in Pkg_W.
The optional GFX_W column shows Watts due to the graphics "un-core".
Note that GFX_W is included in Pkg_W.
The optional RAM_W column on server processors shows Watts due to DRAM DIMMS.
As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W.
The optional PKG_% and RAM_% columns on server processors shows the % of time
in the measurement interval that RAPL power limiting is in effect on the
package and on DRAM.
Note that the RAPL energy counters have some limitations.
First, hardware updates the counters about once every milli-second.
This is fine for typical turbostat measurement intervals > 1 sec.
However, when turbostat is used to measure events that approach
1ms, the counters are less useful.
Second, the 32-bit energy counters are subject to wrapping.
For example, a counter incrementing 15 micro-Joule units
on a 130 Watt TDP server processor could (in theory)
roll over in about 9 minutes. Turbostat detects and handles
up to 1 counter overflow per measurement interval.
But when the measurement interval exceeds the guaranteed
counter range, we can't detect if more than 1 overflow occured.
So in this case turbostat indicates that the results are
in question by replacing the fractional part of the Watts
in the output with "**":
Pkg_W Cor_W GFX_W
3** 0** 0**
Third, the RAPL counters are energy (Joule) counters -- they sum up
weighted events in the package to estimate energy consumed. They are
not analong power (Watt) meters. In practice, they tend to under-count
because they don't cover every possible use of energy in the package.
The accuracy of the RAPL counters will vary between product generations,
and between SKU's in the same product generation, and with temperature.
turbostat's -v (verbose) option now displays more power and thermal configuration
information -- as shown on the turbostat.8 manual page.
For example, it now displays the Package and DRAM Thermal Design Power (TDP):
cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.)
cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.)
cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.)
cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.)
Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
\fB PKG_%\fP percent of the interval that RAPL throttling was active on the Package.
\fB RAM_%\fP percent of the interval that RAPL throttling was active on DRAM.
2010-10-23 11:53:03 +08:00
.fi
.PP
.SH EXAMPLE
2015-02-10 14:56:38 +08:00
Without any parameters, turbostat displays statistics ever 5 seconds.
2010-10-23 11:53:03 +08:00
(override interval with "-i sec" option, or specify a command
for turbostat to fork).
2015-02-23 13:34:57 +08:00
.nf
[root@hsw]# ./turbostat
CPU Avg_MHz %Busy Bzy_MHz TSC_MHz
- 488 12.51 3898 3498
0 0 0.01 3885 3498
4 3897 99.99 3898 3498
1 0 0.00 3861 3498
5 0 0.00 3882 3498
2 1 0.02 3894 3498
6 2 0.06 3898 3498
3 0 0.00 3849 3498
7 0 0.00 3877 3498
.fi
.SH DEBUG EXAMPLE
The "--debug" option prints additional system information before measurements:
2010-10-23 11:53:03 +08:00
2012-02-07 07:37:16 +08:00
The first row of statistics is a summary for the entire system.
tools/power turbostat: v3.0: monitor Watts and Temperature
Show power in Watts and temperature in Celsius
when hardware support is present.
Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL
(Run-Time-Average-Power-Limiting). Per the Intel SDM
(Intel® 64 and IA-32 Architectures Software Developer Manual)
RAPL provides hardware energy counters and power control MSRs
(Model Specific Registers). RAPL MSRs are designed primarily
as a method to implement power capping. However, they are useful
for monitoring system power whether or not power capping is used.
In addition, Turbostat now shows temperature from DTS
(Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware,
if present.
As before, turbostat reads MSRs, and never writes MSRs.
New columns are present in turbostat output:
The Pkg_W column shows Watts for each package (socket) in the system.
On multi-socket systems, the system summary on the 1st row shows the sum
for all sockets together.
The Cor_W column shows Watts due to processors cores.
Note that Core_W is included in Pkg_W.
The optional GFX_W column shows Watts due to the graphics "un-core".
Note that GFX_W is included in Pkg_W.
The optional RAM_W column on server processors shows Watts due to DRAM DIMMS.
As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W.
The optional PKG_% and RAM_% columns on server processors shows the % of time
in the measurement interval that RAPL power limiting is in effect on the
package and on DRAM.
Note that the RAPL energy counters have some limitations.
First, hardware updates the counters about once every milli-second.
This is fine for typical turbostat measurement intervals > 1 sec.
However, when turbostat is used to measure events that approach
1ms, the counters are less useful.
Second, the 32-bit energy counters are subject to wrapping.
For example, a counter incrementing 15 micro-Joule units
on a 130 Watt TDP server processor could (in theory)
roll over in about 9 minutes. Turbostat detects and handles
up to 1 counter overflow per measurement interval.
But when the measurement interval exceeds the guaranteed
counter range, we can't detect if more than 1 overflow occured.
So in this case turbostat indicates that the results are
in question by replacing the fractional part of the Watts
in the output with "**":
Pkg_W Cor_W GFX_W
3** 0** 0**
Third, the RAPL counters are energy (Joule) counters -- they sum up
weighted events in the package to estimate energy consumed. They are
not analong power (Watt) meters. In practice, they tend to under-count
because they don't cover every possible use of energy in the package.
The accuracy of the RAPL counters will vary between product generations,
and between SKU's in the same product generation, and with temperature.
turbostat's -v (verbose) option now displays more power and thermal configuration
information -- as shown on the turbostat.8 manual page.
For example, it now displays the Package and DRAM Thermal Design Power (TDP):
cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.)
cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.)
cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.)
cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.)
Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
For residency % columns, the summary is a weighted average.
For Temperature columns, the summary is the column maximum.
For Watts columns, the summary is a system total.
2010-10-23 11:53:03 +08:00
Subsequent rows show per-CPU statistics.
.nf
2015-02-23 13:34:57 +08:00
turbostat version 4.1 10-Feb, 2015 - Len Brown <lenb@kernel.org>
CPUID(0): GenuineIntel 13 CPUID levels; family:model:stepping 0x6:3c:3 (6:60:3)
tools/power turbostat: v3.0: monitor Watts and Temperature
Show power in Watts and temperature in Celsius
when hardware support is present.
Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL
(Run-Time-Average-Power-Limiting). Per the Intel SDM
(Intel® 64 and IA-32 Architectures Software Developer Manual)
RAPL provides hardware energy counters and power control MSRs
(Model Specific Registers). RAPL MSRs are designed primarily
as a method to implement power capping. However, they are useful
for monitoring system power whether or not power capping is used.
In addition, Turbostat now shows temperature from DTS
(Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware,
if present.
As before, turbostat reads MSRs, and never writes MSRs.
New columns are present in turbostat output:
The Pkg_W column shows Watts for each package (socket) in the system.
On multi-socket systems, the system summary on the 1st row shows the sum
for all sockets together.
The Cor_W column shows Watts due to processors cores.
Note that Core_W is included in Pkg_W.
The optional GFX_W column shows Watts due to the graphics "un-core".
Note that GFX_W is included in Pkg_W.
The optional RAM_W column on server processors shows Watts due to DRAM DIMMS.
As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W.
The optional PKG_% and RAM_% columns on server processors shows the % of time
in the measurement interval that RAPL power limiting is in effect on the
package and on DRAM.
Note that the RAPL energy counters have some limitations.
First, hardware updates the counters about once every milli-second.
This is fine for typical turbostat measurement intervals > 1 sec.
However, when turbostat is used to measure events that approach
1ms, the counters are less useful.
Second, the 32-bit energy counters are subject to wrapping.
For example, a counter incrementing 15 micro-Joule units
on a 130 Watt TDP server processor could (in theory)
roll over in about 9 minutes. Turbostat detects and handles
up to 1 counter overflow per measurement interval.
But when the measurement interval exceeds the guaranteed
counter range, we can't detect if more than 1 overflow occured.
So in this case turbostat indicates that the results are
in question by replacing the fractional part of the Watts
in the output with "**":
Pkg_W Cor_W GFX_W
3** 0** 0**
Third, the RAPL counters are energy (Joule) counters -- they sum up
weighted events in the package to estimate energy consumed. They are
not analong power (Watt) meters. In practice, they tend to under-count
because they don't cover every possible use of energy in the package.
The accuracy of the RAPL counters will vary between product generations,
and between SKU's in the same product generation, and with temperature.
turbostat's -v (verbose) option now displays more power and thermal configuration
information -- as shown on the turbostat.8 manual page.
For example, it now displays the Package and DRAM Thermal Design Power (TDP):
cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.)
cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.)
cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.)
cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.)
Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
CPUID(6): APERF, DTS, PTM, EPB
2015-02-23 13:34:57 +08:00
RAPL: 3121 sec. Joule Counter Range, at 84 Watts
cpu0: MSR_NHM_PLATFORM_INFO: 0x80838f3012300
8 * 100 = 800 MHz max efficiency
tools/power turbostat: v3.0: monitor Watts and Temperature
Show power in Watts and temperature in Celsius
when hardware support is present.
Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL
(Run-Time-Average-Power-Limiting). Per the Intel SDM
(Intel® 64 and IA-32 Architectures Software Developer Manual)
RAPL provides hardware energy counters and power control MSRs
(Model Specific Registers). RAPL MSRs are designed primarily
as a method to implement power capping. However, they are useful
for monitoring system power whether or not power capping is used.
In addition, Turbostat now shows temperature from DTS
(Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware,
if present.
As before, turbostat reads MSRs, and never writes MSRs.
New columns are present in turbostat output:
The Pkg_W column shows Watts for each package (socket) in the system.
On multi-socket systems, the system summary on the 1st row shows the sum
for all sockets together.
The Cor_W column shows Watts due to processors cores.
Note that Core_W is included in Pkg_W.
The optional GFX_W column shows Watts due to the graphics "un-core".
Note that GFX_W is included in Pkg_W.
The optional RAM_W column on server processors shows Watts due to DRAM DIMMS.
As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W.
The optional PKG_% and RAM_% columns on server processors shows the % of time
in the measurement interval that RAPL power limiting is in effect on the
package and on DRAM.
Note that the RAPL energy counters have some limitations.
First, hardware updates the counters about once every milli-second.
This is fine for typical turbostat measurement intervals > 1 sec.
However, when turbostat is used to measure events that approach
1ms, the counters are less useful.
Second, the 32-bit energy counters are subject to wrapping.
For example, a counter incrementing 15 micro-Joule units
on a 130 Watt TDP server processor could (in theory)
roll over in about 9 minutes. Turbostat detects and handles
up to 1 counter overflow per measurement interval.
But when the measurement interval exceeds the guaranteed
counter range, we can't detect if more than 1 overflow occured.
So in this case turbostat indicates that the results are
in question by replacing the fractional part of the Watts
in the output with "**":
Pkg_W Cor_W GFX_W
3** 0** 0**
Third, the RAPL counters are energy (Joule) counters -- they sum up
weighted events in the package to estimate energy consumed. They are
not analong power (Watt) meters. In practice, they tend to under-count
because they don't cover every possible use of energy in the package.
The accuracy of the RAPL counters will vary between product generations,
and between SKU's in the same product generation, and with temperature.
turbostat's -v (verbose) option now displays more power and thermal configuration
information -- as shown on the turbostat.8 manual page.
For example, it now displays the Package and DRAM Thermal Design Power (TDP):
cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.)
cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.)
cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.)
cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.)
Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
35 * 100 = 3500 MHz TSC frequency
2015-02-23 13:34:57 +08:00
cpu0: MSR_IA32_POWER_CTL: 0x0004005d (C1E auto-promotion: DISabled)
cpu0: MSR_NHM_SNB_PKG_CST_CFG_CTL: 0x1e000400 (UNdemote-C3, UNdemote-C1, demote-C3, demote-C1, UNlocked: pkg-cstate-limit=0: pc0)
tools/power turbostat: v3.0: monitor Watts and Temperature
Show power in Watts and temperature in Celsius
when hardware support is present.
Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL
(Run-Time-Average-Power-Limiting). Per the Intel SDM
(Intel® 64 and IA-32 Architectures Software Developer Manual)
RAPL provides hardware energy counters and power control MSRs
(Model Specific Registers). RAPL MSRs are designed primarily
as a method to implement power capping. However, they are useful
for monitoring system power whether or not power capping is used.
In addition, Turbostat now shows temperature from DTS
(Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware,
if present.
As before, turbostat reads MSRs, and never writes MSRs.
New columns are present in turbostat output:
The Pkg_W column shows Watts for each package (socket) in the system.
On multi-socket systems, the system summary on the 1st row shows the sum
for all sockets together.
The Cor_W column shows Watts due to processors cores.
Note that Core_W is included in Pkg_W.
The optional GFX_W column shows Watts due to the graphics "un-core".
Note that GFX_W is included in Pkg_W.
The optional RAM_W column on server processors shows Watts due to DRAM DIMMS.
As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W.
The optional PKG_% and RAM_% columns on server processors shows the % of time
in the measurement interval that RAPL power limiting is in effect on the
package and on DRAM.
Note that the RAPL energy counters have some limitations.
First, hardware updates the counters about once every milli-second.
This is fine for typical turbostat measurement intervals > 1 sec.
However, when turbostat is used to measure events that approach
1ms, the counters are less useful.
Second, the 32-bit energy counters are subject to wrapping.
For example, a counter incrementing 15 micro-Joule units
on a 130 Watt TDP server processor could (in theory)
roll over in about 9 minutes. Turbostat detects and handles
up to 1 counter overflow per measurement interval.
But when the measurement interval exceeds the guaranteed
counter range, we can't detect if more than 1 overflow occured.
So in this case turbostat indicates that the results are
in question by replacing the fractional part of the Watts
in the output with "**":
Pkg_W Cor_W GFX_W
3** 0** 0**
Third, the RAPL counters are energy (Joule) counters -- they sum up
weighted events in the package to estimate energy consumed. They are
not analong power (Watt) meters. In practice, they tend to under-count
because they don't cover every possible use of energy in the package.
The accuracy of the RAPL counters will vary between product generations,
and between SKU's in the same product generation, and with temperature.
turbostat's -v (verbose) option now displays more power and thermal configuration
information -- as shown on the turbostat.8 manual page.
For example, it now displays the Package and DRAM Thermal Design Power (TDP):
cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.)
cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.)
cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.)
cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.)
Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
cpu0: MSR_NHM_TURBO_RATIO_LIMIT: 0x25262727
37 * 100 = 3700 MHz max turbo 4 active cores
38 * 100 = 3800 MHz max turbo 3 active cores
39 * 100 = 3900 MHz max turbo 2 active cores
39 * 100 = 3900 MHz max turbo 1 active cores
cpu0: MSR_IA32_ENERGY_PERF_BIAS: 0x00000006 (balanced)
2015-02-23 13:34:57 +08:00
cpu0: MSR_CORE_PERF_LIMIT_REASONS, 0x31200000 (Active: ) (Logged: Auto-HWP, Amps, MultiCoreTurbo, Transitions, )
cpu0: MSR_GFX_PERF_LIMIT_REASONS, 0x00000000 (Active: ) (Logged: )
cpu0: MSR_RING_PERF_LIMIT_REASONS, 0x0d000000 (Active: ) (Logged: Amps, PkgPwrL1, PkgPwrL2, )
cpu0: MSR_RAPL_POWER_UNIT: 0x000a0e03 (0.125000 Watts, 0.000061 Joules, 0.000977 sec.)
cpu0: MSR_PKG_POWER_INFO: 0x000002a0 (84 W TDP, RAPL 0 - 0 W, 0.000000 sec.)
cpu0: MSR_PKG_POWER_LIMIT: 0x428348001a82a0 (UNlocked)
cpu0: PKG Limit #1: ENabled (84.000000 Watts, 8.000000 sec, clamp DISabled)
cpu0: PKG Limit #2: ENabled (105.000000 Watts, 0.002441* sec, clamp DISabled)
tools/power turbostat: v3.0: monitor Watts and Temperature
Show power in Watts and temperature in Celsius
when hardware support is present.
Intel's Sandy Bridge and Ivy Bridge processor generations support RAPL
(Run-Time-Average-Power-Limiting). Per the Intel SDM
(Intel® 64 and IA-32 Architectures Software Developer Manual)
RAPL provides hardware energy counters and power control MSRs
(Model Specific Registers). RAPL MSRs are designed primarily
as a method to implement power capping. However, they are useful
for monitoring system power whether or not power capping is used.
In addition, Turbostat now shows temperature from DTS
(Digital Thermal Sensor) and PTM (Package Thermal Monitor) hardware,
if present.
As before, turbostat reads MSRs, and never writes MSRs.
New columns are present in turbostat output:
The Pkg_W column shows Watts for each package (socket) in the system.
On multi-socket systems, the system summary on the 1st row shows the sum
for all sockets together.
The Cor_W column shows Watts due to processors cores.
Note that Core_W is included in Pkg_W.
The optional GFX_W column shows Watts due to the graphics "un-core".
Note that GFX_W is included in Pkg_W.
The optional RAM_W column on server processors shows Watts due to DRAM DIMMS.
As DRAM DIMMs are outside the processor package, RAM_W is not included in Pkg_W.
The optional PKG_% and RAM_% columns on server processors shows the % of time
in the measurement interval that RAPL power limiting is in effect on the
package and on DRAM.
Note that the RAPL energy counters have some limitations.
First, hardware updates the counters about once every milli-second.
This is fine for typical turbostat measurement intervals > 1 sec.
However, when turbostat is used to measure events that approach
1ms, the counters are less useful.
Second, the 32-bit energy counters are subject to wrapping.
For example, a counter incrementing 15 micro-Joule units
on a 130 Watt TDP server processor could (in theory)
roll over in about 9 minutes. Turbostat detects and handles
up to 1 counter overflow per measurement interval.
But when the measurement interval exceeds the guaranteed
counter range, we can't detect if more than 1 overflow occured.
So in this case turbostat indicates that the results are
in question by replacing the fractional part of the Watts
in the output with "**":
Pkg_W Cor_W GFX_W
3** 0** 0**
Third, the RAPL counters are energy (Joule) counters -- they sum up
weighted events in the package to estimate energy consumed. They are
not analong power (Watt) meters. In practice, they tend to under-count
because they don't cover every possible use of energy in the package.
The accuracy of the RAPL counters will vary between product generations,
and between SKU's in the same product generation, and with temperature.
turbostat's -v (verbose) option now displays more power and thermal configuration
information -- as shown on the turbostat.8 manual page.
For example, it now displays the Package and DRAM Thermal Design Power (TDP):
cpu0: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.)
cpu0: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.)
cpu8: MSR_PKG_POWER_INFO: 0x2f064001980410 (130 W TDP, RAPL 51 - 200 W, 0.045898 sec.)
cpu8: MSR_DRAM_POWER_INFO,: 0x28025800780118 (35 W TDP, RAPL 15 - 75 W, 0.039062 sec.)
Signed-off-by: Len Brown <len.brown@intel.com>
2012-11-08 13:48:57 +08:00
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)
2015-02-23 13:34:57 +08:00
cpu0: MSR_IA32_TEMPERATURE_TARGET: 0x00641400 (100 C)
cpu0: MSR_IA32_PACKAGE_THERM_STATUS: 0x88340800 (48 C)
cpu0: MSR_IA32_THERM_STATUS: 0x88340000 (48 C +/- 1)
cpu1: MSR_IA32_THERM_STATUS: 0x88440000 (32 C +/- 1)
cpu2: MSR_IA32_THERM_STATUS: 0x88450000 (31 C +/- 1)
cpu3: MSR_IA32_THERM_STATUS: 0x88490000 (27 C +/- 1)
Core CPU Avg_MHz %Busy Bzy_MHz TSC_MHz SMI CPU%c1 CPU%c3 CPU%c6 CPU%c7 CoreTmp PkgTmp PkgWatt CorWatt GFXWatt
- - 493 12.64 3898 3498 0 12.64 0.00 0.00 74.72 47 47 21.62 13.74 0.00
0 0 4 0.11 3894 3498 0 99.89 0.00 0.00 0.00 47 47 21.62 13.74 0.00
0 4 3897 99.98 3898 3498 0 0.02
1 1 7 0.17 3887 3498 0 0.04 0.00 0.00 99.79 32
1 5 0 0.00 3885 3498 0 0.21
2 2 29 0.76 3895 3498 0 0.10 0.01 0.01 99.13 32
2 6 2 0.06 3896 3498 0 0.80
3 3 1 0.02 3832 3498 0 0.03 0.00 0.00 99.95 28
3 7 0 0.00 3879 3498 0 0.04
^C
2010-10-23 11:53:03 +08:00
.fi
The \fB max efficiency\fP frequency, a.k.a. Low Frequency Mode, is the frequency
2015-01-23 14:33:58 +08:00
available at the minimum package voltage. The \fB TSC frequency\fP is the base
frequency of the processor -- this should match the brand string
in /proc/cpuinfo. This base frequency
2010-10-23 11:53:03 +08:00
should be sustainable on all CPUs indefinitely, given nominal power and cooling.
The remaining rows show what maximum turbo frequency is possible
2015-01-23 14:33:58 +08:00
depending on the number of idle cores. Note that not all information is
available on all processors.
2015-02-23 13:34:57 +08:00
.PP
The --debug option adds additional columns to the measurement ouput, including CPU idle power-state residency processor temperature sensor readinds.
See the field definitions above.
2010-10-23 11:53:03 +08:00
.SH FORK EXAMPLE
If turbostat is invoked with a command, it will fork that command
and output the statistics gathered when the command exits.
eg. Here a cycle soaker is run on 1 CPU (see %c0) for a few seconds
until ^C while the other CPUs are mostly idle:
.nf
2015-02-23 13:34:57 +08:00
root@hsw: turbostat cat /dev/zero > /dev/null
2012-02-07 07:37:16 +08:00
^C
2015-02-23 13:34:57 +08:00
CPU Avg_MHz %Busy Bzy_MHz TSC_MHz
- 482 12.51 3854 3498
0 0 0.01 1960 3498
4 0 0.00 2128 3498
1 0 0.00 3003 3498
5 3854 99.98 3855 3498
2 0 0.01 3504 3498
6 3 0.08 3884 3498
3 0 0.00 2553 3498
7 0 0.00 2126 3498
10.783983 sec
2014-02-06 13:55:19 +08:00
2010-10-23 11:53:03 +08:00
.fi
2015-02-23 13:34:57 +08:00
Above the cycle soaker drives cpu5 up its 3.9 GHz turbo limit.
The first row shows the average MHz and %Busy across all the processors in the system.
2010-10-23 11:53:03 +08:00
2014-02-06 13:55:19 +08:00
Note that the Avg_MHz column reflects the total number of cycles executed
divided by the measurement interval. If the %Busy column is 100%,
then the processor was running at that speed the entire interval.
The Avg_MHz multiplied by the %Busy results in the Bzy_MHz --
which is the average frequency while the processor was executing --
not including any non-busy idle time.
2010-10-23 11:53:03 +08:00
.SH NOTES
.B "turbostat "
must be run as root.
2015-01-23 14:33:58 +08:00
Alternatively, non-root users can be enabled to run turbostat this way:
# setcap cap_sys_rawio=ep ./turbostat
# chmod +r /dev/cpu/*/msr
2010-10-23 11:53:03 +08:00
.B "turbostat "
reads hardware counters, but doesn't write them.
So it will not interfere with the OS or other programs, including
multiple invocations of itself.
\fB turbostat \fP
may work poorly on Linux-2.6.20 through 2.6.29,
2015-01-23 14:33:58 +08:00
as \fB acpi-cpufreq \fP periodically cleared the APERF and MPERF MSRs
2010-10-23 11:53:03 +08:00
in those kernels.
2015-01-23 14:33:58 +08:00
AVG_MHz = APERF_delta/measurement_interval. This is the actual
number of elapsed cycles divided by the entire sample interval --
2015-02-10 14:56:38 +08:00
including idle time. Note that this calculation is resilient
2015-01-23 14:33:58 +08:00
to systems lacking a non-stop TSC.
TSC_MHz = TSC_delta/measurement_interval.
On a system with an invariant TSC, this value will be constant
and will closely match the base frequency value shown
in the brand string in /proc/cpuinfo. On a system where
the TSC stops in idle, TSC_MHz will drop
below the processor's base frequency.
%Busy = MPERF_delta/TSC_delta
Bzy_MHz = TSC_delta/APERF_delta/MPERF_delta/measurement_interval
Note that these calculations depend on TSC_delta, so they
are not reliable during intervals when TSC_MHz is not running at the base frequency.
Turbostat data collection is not atomic.
Extremely short measurement intervals (much less than 1 second),
or system activity that prevents turbostat from being able
to run on all CPUS to quickly collect data, will result in
inconsistent results.
2012-09-22 11:45:46 +08:00
2010-10-23 11:53:03 +08:00
The APERF, MPERF MSRs are defined to count non-halted cycles.
Although it is not guaranteed by the architecture, turbostat assumes
that they count at TSC rate, which is true on all processors tested to date.
.SH REFERENCES
"Intel® Turbo Boost Technology
in Intel® Core™ Microarchitecture (Nehalem) Based Processors"
http://download.intel.com/design/processor/applnots/320354.pdf
"Intel® 64 and IA-32 Architectures Software Developer's Manual
Volume 3B: System Programming Guide"
http://www.intel.com/products/processor/manuals/
.SH FILES
.ta
.nf
/dev/cpu/*/msr
.fi
.SH "SEE ALSO"
msr(4), vmstat(8)
.PP
2012-02-07 07:37:16 +08:00
.SH AUTHOR
2010-10-23 11:53:03 +08:00
.nf
Written by Len Brown <len.brown@intel.com>
