turbostat.8

  1.TH TURBOSTAT 8
  2.SH NAME
  3turbostat \- Report processor frequency and idle statistics
  4.SH SYNOPSIS
  5.ft B
  6.B turbostat
  7.RB [ Options ]
  8.RB command
  9.br
 10.B turbostat
 11.RB [ Options ]
 12.RB [ "\--interval seconds" ]
 13.SH DESCRIPTION
 14\fBturbostat \fP reports processor topology, frequency,
 15idle power-state statistics, temperature and power on X86 processors.
 16There are two ways to invoke turbostat.
 17The first method is to supply a
 18\fBcommand\fP, which is forked and statistics are printed
 19in one-shot upon its completion.
 20The second method is to omit the command,
 21and turbostat displays statistics every 5 seconds interval.
 22The 5-second interval can be changed using the --interval option.
 23.PP
 24Some information is not available on older processors.
 25.SS Options
 26Options can be specified with a single or double '-', and only as much of the option
 27name as necessary to disambiguate it from others is necessary.  Note that options are case-sensitive.
 28.PP
 29\fB--add attributes\fP add column with counter having specified 'attributes'.  The 'location' attribute is required, all others are optional.
 30.nf
 31	location: {\fBmsrDDD\fP | \fBmsr0xXXX\fP | \fB/sys/path...\fP | \fBperf/<device>/<event>\fP}
 32		msrDDD is a decimal offset, eg. msr16
 33		msr0xXXX is a hex offset, eg. msr0x10
 34		/sys/path... is an absolute path to a sysfs attribute
 35		<device> is a perf device from /sys/bus/event_source/devices/<device> eg. cstate_core
 36			On Intel hybrid platforms, instead of one "cpu" perf device there are two, "cpu_core" and "cpu_atom" devices for P and E cores respectively.
 37			Turbostat, in this case, allow user to use "cpu" device and will automatically detect the type of a CPU and translate it to "cpu_core" and "cpu_atom" accordingly.
 38			For a complete example see "ADD PERF COUNTER EXAMPLE #2 (using virtual "cpu" device)".
 39		<event> is a perf event for given device from /sys/bus/event_source/devices/<device>/events/<event> eg. c1-residency
 40			perf/cstate_core/c1-residency would then use /sys/bus/event_source/devices/cstate_core/events/c1-residency
 41
 42	scope: {\fBcpu\fP | \fBcore\fP | \fBpackage\fP}
 43		sample and print the counter for every cpu, core, or package.
 44		default: cpu
 45
 46	size: {\fBu32\fP | \fBu64\fP }
 47		MSRs are read as 64-bits, u32 truncates the displayed value to 32-bits.
 48		default: u64
 49
 50	format: {\fBraw\fP | \fBdelta\fP | \fBpercent\fP | \fBaverage\fP}
 51		'raw' shows the MSR contents in hex.
 52		'delta' shows the difference in values during the measurement interval.
 53		'percent' shows the delta as a percentage of the cycles elapsed.
 54		'average' similar to raw, but also averaged for node/package summaries (or when using -S).
 55		default: delta
 56
 57	name: "name_string"
 58		Any string that does not match a key-word above is used
 59		as the column header.
 60.fi
 61.PP
 62\fB--add pmt,[attr_name=attr_value, ...]\fP add column with a PMT (Intel Platform Monitoring Technology) counter in a similar way to --add option above, but require PMT metadata to be supplied to correctly read and display the counter. The metadata can be found in the Intel PMT XML files, hosted at https://github.com/intel/Intel-PMT. For a complete example see "ADD PMT COUNTER EXAMPLE".
 63.nf
 64	name="name_string"
 65		For column header.
 66
 67	type={\fBraw\fP}
 68		'raw' shows the counter contents in hex.
 69		default: raw
 70
 71	format={\fBraw\fP | \fBdelta\fP}
 72		'raw' shows the counter contents in hex.
 73		'delta' shows the difference in values during the measurement interval.
 74		default: raw
 75
 76	domain={\fBcpu%u\fP | \fBcore%u\fP | \fBpackage%u\fP}
 77		'cpu' per cpu/thread counter.
 78		'core' per core counter.
 79		'package' per package counter.
 80		'%u' denotes id of the domain that the counter is associated with. For example core4 would mean that the counter is associated with core number 4.
 81
 82	offset=\fB%u\fP
 83		'%u' offset within the PMT MMIO region.
 84
 85	lsb=\fB%u\fP
 86		'%u' least significant bit within the 64 bit value read from 'offset'. Together with 'msb', used to form a read mask.
 87
 88	msb=\fB%u\fP
 89		'%u' most significant bit within the 64 bit value read from 'offset'. Together with 'lsb', used to form a read mask.
 90
 91	guid=\fB%x\fP
 92		'%x' hex identifier of the PMT MMIO region.
 93.fi
 94.PP
 95\fB--cpu cpu-set\fP limit output to system summary plus the specified cpu-set.  If cpu-set is the string "core", then the system summary plus the first CPU in each core are printed -- eg. subsequent HT siblings are not printed.  Or if cpu-set is the string "package", then the system summary plus the first CPU in each package is printed.  Otherwise, the system summary plus the specified set of CPUs are printed.  The cpu-set is ordered from low to high, comma delimited with ".." and "-" permitted to denote a range. eg. 1,2,8,14..17,21-44
 96.PP
 97\fB--hide column\fP do not show the specified built-in columns.  May be invoked multiple times, or with a comma-separated list of column names.
 98.PP
 99\fB--enable column\fP show the specified built-in columns, which are otherwise disabled, by default.  Currently the only built-in counters disabled by default are "usec", "Time_Of_Day_Seconds", "APIC" and "X2APIC".
100The column name "all" can be used to enable all disabled-by-default built-in counters.
101.PP
102\fB--show column\fP show only the specified built-in columns.  May be invoked multiple times, or with a comma-separated list of column names.
103.PP
104\fB--show CATEGORY --hide CATEGORY\fP  Show and hide also accept a comma-separated-list of CATEGORIES of columns: "all", "topology", "idle", "frequency", "power", "cpuidle", "hwidle", "swidle", "cache", "llc", "other".  "idle" (enabled by default), includes "hwidle" and "pct_idle".  "cpuidle" (default disabled) includes cpuidle software invocation counters.  "swidle" includes "cpuidle" plus "pct_idle".  "hwidle" includes only hardware based idle residency counters.  Older versions of turbostat used the term "sysfs" for what is now "swidle".
105.PP
106\fB--Dump\fP displays the raw counter values.
107.PP
108\fB--quiet\fP Do not decode and print the system configuration header information.
109.PP
110\fB--no-msr\fP Disable all the uses of the MSR driver.
111.PP
112\fB--no-perf\fP Disable all the uses of the perf API.
113.PP
114\fB--interval seconds\fP overrides the default 5.0 second measurement interval.
115.PP
116\fB--num_iterations num\fP number of the measurement iterations.
117.PP
118\fB--out output_file\fP turbostat output is written to the specified output_file.
119The file is truncated if it already exists, and it is created if it does not exist.
120.PP
121\fB--help\fP displays usage for the most common parameters.
122.PP
123\fB--Joules\fP displays energy in Joules, rather than dividing Joules by time to print power in Watts.
124.PP
125\fB--list\fP display column header names available for use by --show and --hide, then exit.
126.PP
127\fB--Summary\fP limits output to a 1-line System Summary for each interval.
128.PP
129\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.
130.PP
131\fB--version\fP displays the version.
132.PP
133The \fBcommand\fP parameter forks \fBcommand\fP, and upon its exit,
134displays the statistics gathered since it was forked.
135.PP
136.SH ROW DESCRIPTIONS
137The system configuration dump (if --quiet is not used) is followed by statistics.  The first row of the statistics labels the content of each column (below).  The second row of statistics is the system summary line.  The system summary line has a '-' in the columns for the Package, Core, and CPU.  The contents of the system summary line depends on the type of column.  Columns that count items (eg. IRQ) show the sum across all CPUs in the system.  Columns that show a percentage show the average across all CPUs in the system.  Columns that dump raw MSR values simply show 0 in the summary.  After the system summary row, each row describes a specific Package/Core/CPU.  Note that if the --cpu parameter is used to limit which specific CPUs are displayed, turbostat will still collect statistics for all CPUs in the system and will still show the system summary for all CPUs in the system.
138.SH COLUMN DESCRIPTIONS
139.PP
140\fBusec\fP For each CPU, the number of microseconds elapsed during counter collection, including thread migration -- if any.  This counter is disabled by default, and is enabled with "--enable usec", or --debug.  On the summary row, usec refers to the total elapsed time to snapshot the procfs/sysfs and collect the counters on all cpus.
141.PP
142\fBTime_Of_Day_Seconds\fP For each CPU, the gettimeofday(2) value (seconds.subsec since Epoch) when the counters ending the measurement interval were collected.  This column is disabled by default, and can be enabled with "--enable Time_Of_Day_Seconds" or "--debug".  On the summary row, Time_Of_Day_Seconds refers to the timestamp following collection of counters on the last CPU.
143.PP
144\fBCore\fP processor core number.  Note that multiple CPUs per core indicate support for Intel(R) Hyper-Threading Technology (HT).
145.PP
146\fBCPU\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.
147.PP
148\fBPackage\fP processor package number -- not present on systems with a single processor package.
149.PP
150\fBAvg_MHz\fP number of cycles executed divided by time elapsed.  Note that this includes idle-time when 0 instructions are executed.
151.PP
152\fBBusy%\fP percent of the measurement interval that the CPU executes instructions, aka. % of time in "C0" state.
153.PP
154\fBBzy_MHz\fP average clock rate while the CPU was not idle (ie. in "c0" state).
155.PP
156\fBTSC_MHz\fP average MHz that the TSC ran during the entire interval.
157.PP
158\fBIRQ\fP The number of interrupts serviced by that CPU during the measurement interval.  The system total line is the sum of interrupts serviced across all CPUs.  turbostat parses /proc/interrupts to generate this summary.
159.PP
160\fBSMI\fP The number of System Management Interrupts  serviced CPU during the measurement interval.  While this counter is actually per-CPU, SMI are triggered on all processors, so the number should be the same for all CPUs.
161.PP
162\fBLLCkRPS\fP Last Level Cache Thousands of References Per Second.  For CPUs with an L3 LLC, this is the number of references that CPU made to the L3 (and the number of misses that CPU made to it's L2).  For CPUs with an L2 LLC, this is the number of references to the L2 (and the number of misses to the CPU's L1).  The system summary row shows the sum for all CPUs.  In both cases, the value displayed is the actual value divided by 1000 in the interest of usually fitting into 8 columns.
163.PP
164\fBLLC%hit\fP Last Level Cache Hit Rate %.  Hit Rate Percent = 100.0 * (References - Misses)/References.  The system summary row shows the weighted average for all CPUs (100.0 * (Sum_References - Sum_Misses)/Sum_References).
165.PP
166\fBC1, C2, C3...\fP The number times Linux requested the C1, C2, C3 idle state during the measurement interval.  The system summary line shows the sum for all CPUs.  These are C-state names as exported in /sys/devices/system/cpu/cpu*/cpuidle/state*/name.  While their names are generic, their attributes are processor specific. They the system description section of output shows what MWAIT sub-states they are mapped to on each system.  These counters are in the "cpuidle" group, which is disabled, by default.
167.PP
168\fBC1+, C2+, C3+...\fP The idle governor idle state misprediction statistics. Inidcates the number times Linux requested the C1, C2, C3 idle state during the measurement interval, but should have requested a deeper idle state (if it exists and enabled). These statistics come from the /sys/devices/system/cpu/cpu*/cpuidle/state*/below file.  These counters are in the "cpuidle" group, which is disabled, by default.
169.PP
170\fBC1-, C2-, C3-...\fP The idle governor idle state misprediction statistics. Inidcates the number times Linux requested the C1, C2, C3 idle state during the measurement interval, but should have requested a shallower idle state (if it exists and enabled). These statistics come from the /sys/devices/system/cpu/cpu*/cpuidle/state*/above file.  These counters are in the "cpuidle" group, which is disabled, by default.
171.PP
172\fBC1%, C2%, C3%\fP The residency percentage that Linux requested C1, C2, C3....  The system summary is the average of all CPUs in the system.  Note that these are software, reflecting what was requested.  The hardware counters reflect what was actually achieved.  These counters are in the "pct_idle" group, which is enabled by default.
173.PP
174\fBCPU%c1, CPU%c3, CPU%c6, CPU%c7\fP show the percentage residency in hardware core idle states.  These numbers are from hardware residency counters and are in the "hwidle" group, which is enabled, by default.
175.PP
176\fBCoreTmp\fP Degrees Celsius reported by the per-core Digital Thermal Sensor.
177.PP
178\fBPkgTmp\fP Degrees Celsius reported by the per-package Package Thermal Monitor.
179.PP
180\fBCoreThr\fP Core Thermal Throttling events during the measurement interval.  Note that events since boot can be find in /sys/devices/system/cpu/cpu*/thermal_throttle/*
181.PP
182\fBGFX%rc6\fP The percentage of time the GPU is in the "render C6" state, rc6, during the measurement interval. From /sys/class/drm/card0/power/rc6_residency_ms or /sys/class/drm/card0/gt/gt0/rc6_residency_ms or /sys/class/drm/card0/device/tile0/gtN/gtidle/idle_residency_ms depending on the graphics driver being used.
183.PP
184\fBGFXMHz\fP Instantaneous snapshot of what sysfs presents at the end of the measurement interval. From /sys/class/graphics/fb0/device/drm/card0/gt_cur_freq_mhz or /sys/class/drm/card0/gt_cur_freq_mhz or /sys/class/drm/card0/gt/gt0/rps_cur_freq_mhz or /sys/class/drm/card0/device/tile0/gtN/freq0/cur_freq depending on the graphics driver being used.
185.PP
186\fBGFXAMHz\fP Instantaneous snapshot of what sysfs presents at the end of the measurement interval. From /sys/class/graphics/fb0/device/drm/card0/gt_act_freq_mhz or /sys/class/drm/card0/gt_act_freq_mhz or /sys/class/drm/card0/gt/gt0/rps_act_freq_mhz or /sys/class/drm/card0/device/tile0/gtN/freq0/act_freq depending on the graphics driver being used.
187.PP
188\fBSAM%mc6\fP The percentage of time the SA Media is in the "module C6" state, mc6, during the measurement interval. From /sys/class/drm/card0/gt/gt1/rc6_residency_ms or /sys/class/drm/card0/device/tile0/gtN/gtidle/idle_residency_ms depending on the graphics driver being used.
189.PP
190\fBSAMMHz\fP Instantaneous snapshot of what sysfs presents at the end of the measurement interval. From /sys/class/drm/card0/gt/gt1/rps_cur_freq_mhz or /sys/class/drm/card0/device/tile0/gtN/freq0/cur_freq depending on the graphics driver being used.
191.PP
192\fBSAMAMHz\fP Instantaneous snapshot of what sysfs presents at the end of the measurement interval. From /sys/class/drm/card0/gt/gt1/rps_act_freq_mhz or /sys/class/drm/card0/device/tile0/gtN/freq0/act_freq depending on the graphics driver being used.
193.PP
194\fBTotl%C0\fP Weighted percentage of time that CPUs are busy. If N CPUs are busy during an interval, the percentage is N * 100%.
195.PP
196\fBAny%C0\fP Percentage of time that at least one CPU is busy.
197.PP
198\fBGFX%C0\fP Percentage of time that at least one GFX compute engine is busy.
199.PP
200\fBCPUGFX%\fP Percentage of time that at least one CPU is busy at the same time as at least one Graphics compute enginer is busy.
201.PP
202\fBPkg%pc2, Pkg%pc3, Pkg%pc6, Pkg%pc7\fP percentage residency in hardware package idle states.  These numbers are from hardware residency counters.
203.PP
204\fBPkgWatt\fP Watts consumed by the whole package.
205.PP
206\fBCorWatt\fP Watts consumed by the core part of the package.
207.PP
208\fBGFXWatt\fP Watts consumed by the Graphics part of the package -- available only on client processors.
209.PP
210\fBRAMWatt\fP Watts consumed by the DRAM DIMMS -- available only on server processors.
211.PP
212\fBSysWatt\fP Watts consumed by the whole platform (RAPL PSYS).
213.PP
214\fBPKG_%\fP percent of the interval that RAPL throttling was active on the Package.  Note that the system summary is the sum of the package throttling time, and thus may be higher than 100% on a multi-package system.  Note that the meaning of this field is model specific.  For example, some hardware increments this counter when RAPL responds to thermal limits, but does not increment this counter when RAPL responds to power limits.  Comparing PkgWatt and PkgTmp to system limits is necessary.
215.PP
216\fBRAM_%\fP percent of the interval that RAPL throttling was active on DRAM.
217.PP
218\fBUncMHz\fP per-package uncore MHz, instantaneous sample.
219.PP
220\fBUMHz1.0\fP per-package uncore MHz for pm_domain=1 and fabric_cluster=0, instantaneous sample.  System summary is the average of all packages.
221Intel Granite Rapids systems use pm_domains 0-2 for CPUs, and 3-4 for IO, with cluster always 0.
222For the "--show" and "--hide" options, use "UncMHz" to operate on all UMHz*.* as a group.
223.SH TOO MUCH INFORMATION EXAMPLE
224By default, turbostat dumps all possible information -- a system configuration header, followed by columns for all counters.
225This is ideal for remote debugging, use the "--out" option to save everything to a text file, and get that file to the expert helping you debug.
226.PP
227When you are not interested in all that information, and there are several ways to see only what you want.  First the "--quiet" option will skip the configuration information, and turbostat will show only the counter columns.  Second, you can reduce the columns with the "--hide" and "--show" options.  If you use the "--show" option, then turbostat will show only the columns you list.  If you use the "--hide" option, turbostat will show all columns, except the ones you list.
228.PP
229To find out what columns are available for --show and --hide, the "--list" option is available.  Usually, the CATEGORY names above are used to refer to groups of counters.  Also, for convenience, the special string "sysfs" can be used to refer to all of the sysfs C-state counters at once:
230.PP
231.nf
232sudo ./turbostat --show sysfs --quiet sleep 10
23310.003837 sec
234	C1	C1E	C3	C6	C7s	C1%	C1E%	C3%	C6%	C7s%
235	4	21	2	2	459	0.14	0.82	0.00	0.00	98.93
236	1	17	2	2	130	0.00	0.02	0.00	0.00	99.80
237	0	0	0	0	31	0.00	0.00	0.00	0.00	99.95
238	2	1	0	0	52	1.14	6.49	0.00	0.00	92.21
239	1	2	0	0	52	0.00	0.08	0.00	0.00	99.86
240	0	0	0	0	71	0.00	0.00	0.00	0.00	99.89
241	0	0	0	0	25	0.00	0.00	0.00	0.00	99.96
242	0	0	0	0	74	0.00	0.00	0.00	0.00	99.94
243	0	1	0	0	24	0.00	0.00	0.00	0.00	99.84
244.fi
245.PP
246.SH ONE SHOT COMMAND EXAMPLE
247If turbostat is invoked with a command, it will fork that command
248and output the statistics gathered after the command exits.
249In this case, turbostat output goes to stderr, by default.
250Output can instead be saved to a file using the --out option.
251In this example, the "sleep 10" command is forked, and turbostat waits for it to complete before saving all statistics into "ts.out".  Note that "sleep 10" is not part of turbostat, but is simply an example of a command that turbostat can fork.  The "ts.out" file is what you want to edit in a very wide window, paste into a spreadsheet, or attach to a bugzilla entry.
252
253.nf
254[root@hsw]# ./turbostat -o ts.out sleep 10
255[root@hsw]#
256.fi
257
258.SH PERIODIC INTERVAL EXAMPLE
259Without a command to fork, turbostat displays statistics ever 5 seconds.
260Periodic output goes to stdout, by default, unless --out is used to specify an output file.
261The 5-second interval can be changed with the "-i sec" option.
262.nf
263sudo turbostat --quiet --show CPU,frequency
264	Core	CPU	Avg_MHz	Busy%	Bzy_MHz	TSC_MHz	CPU%c7	UncMhz
265	-	-	524	12.48	4198	3096	74.53	3800
266	0	0	4	0.09	4081	3096	98.88	3800
267	0	4	1	0.02	4063	3096
268	1	1	2	0.06	4063	3096	99.60
269	1	5	2	0.05	4070	3096
270	2	2	4178	99.52	4199	3096	0.00
271	2	6	3	0.08	4159	3096
272	3	3	1	0.04	4046	3096	99.66
273	3	7	0	0.01	3989	3096
274	Core	CPU	Avg_MHz	Busy%	Bzy_MHz	TSC_MHz	CPU%c7	UncMhz
275	-	-	525	12.52	4198	3096	74.54	3800
276	0	0	4	0.10	4051	3096	99.49	3800
277	0	4	2	0.04	3993	3096
278	1	1	3	0.07	4054	3096	99.56
279	1	5	4	0.10	4018	3096
280	2	2	4178	99.51	4199	3096	0.00
281	2	6	4	0.09	4143	3096
282	3	3	2	0.06	4026	3096	99.10
283	3	7	7	0.17	4074	3096
284.fi
285This example also shows the use of the --show option to show only the desired columns.
286
287.SH SYSTEM CONFIGURATION INFORMATION EXAMPLE
288
289By default, turbostat always dumps system configuration information
290before taking measurements.  In the example above, "--quiet" is used
291to suppress that output.  Here is an example of the configuration information:
292.nf
293turbostat version 2022.04.16 - Len Brown <lenb@kernel.org>
294Kernel command line: BOOT_IMAGE=/boot/vmlinuz-5.18.0-rc6-00001-ge6891250e3b5 ...
295CPUID(0): GenuineIntel 0x16 CPUID levels
296CPUID(1): family:model:stepping 0x6:9e:9 (6:158:9) microcode 0xea
297CPUID(0x80000000): max_extended_levels: 0x80000008
298CPUID(1): SSE3 MONITOR - EIST TM2 TSC MSR ACPI-TM HT TM
299CPUID(6): APERF, TURBO, DTS, PTM, HWP, HWPnotify, HWPwindow, HWPepp, No-HWPpkg, EPB
300cpu7: MSR_IA32_MISC_ENABLE: 0x00850089 (TCC EIST MWAIT PREFETCH TURBO)
301CPUID(7): SGX
302cpu7: MSR_IA32_FEATURE_CONTROL: 0x00000005 (Locked )
303CPUID(0x15): eax_crystal: 2 ebx_tsc: 258 ecx_crystal_hz: 0
304TSC: 3096 MHz (24000000 Hz * 258 / 2 / 1000000)
305CPUID(0x16): base_mhz: 3100 max_mhz: 4200 bus_mhz: 100
306cpu7: MSR_MISC_PWR_MGMT: 0x00401cc0 (ENable-EIST_Coordination DISable-EPB DISable-OOB)
307RAPL: 5825 sec. Joule Counter Range, at 45 Watts
308cpu7: MSR_PLATFORM_INFO: 0x80839f1011f00
3098 * 100.0 = 800.0 MHz max efficiency frequency
31031 * 100.0 = 3100.0 MHz base frequency
311cpu7: MSR_IA32_POWER_CTL: 0x002c005d (C1E auto-promotion: DISabled)
312cpu7: MSR_TURBO_RATIO_LIMIT: 0x2728292a
31339 * 100.0 = 3900.0 MHz max turbo 4 active cores
31440 * 100.0 = 4000.0 MHz max turbo 3 active cores
31541 * 100.0 = 4100.0 MHz max turbo 2 active cores
31642 * 100.0 = 4200.0 MHz max turbo 1 active cores
317cpu7: MSR_CONFIG_TDP_NOMINAL: 0x0000001f (base_ratio=31)
318cpu7: MSR_CONFIG_TDP_LEVEL_1: 0x00000000 ()
319cpu7: MSR_CONFIG_TDP_LEVEL_2: 0x00000000 ()
320cpu7: MSR_CONFIG_TDP_CONTROL: 0x80000000 ( lock=1)
321cpu7: MSR_TURBO_ACTIVATION_RATIO: 0x00000000 (MAX_NON_TURBO_RATIO=0 lock=0)
322cpu7: MSR_PKG_CST_CONFIG_CONTROL: 0x1e008008 (UNdemote-C3, UNdemote-C1, demote-C3, demote-C1, locked, pkg-cstate-limit=8 (unlimited))
323Uncore Frequency pkg0 die0: 800 - 3900 MHz (800 - 3900 MHz)
324/dev/cpu_dma_latency: 2000000000 usec (default)
325current_driver: intel_idle
326current_governor: menu
327current_governor_ro: menu
328cpu7: POLL: CPUIDLE CORE POLL IDLE
329cpu7: C1: MWAIT 0x00
330cpu7: C1E: MWAIT 0x01
331cpu7: C3: MWAIT 0x10
332cpu7: C6: MWAIT 0x20
333cpu7: C7s: MWAIT 0x33
334cpu7: C8: MWAIT 0x40
335cpu7: C9: MWAIT 0x50
336cpu7: C10: MWAIT 0x60
337cpu7: cpufreq driver: intel_pstate
338cpu7: cpufreq governor: performance
339cpufreq intel_pstate no_turbo: 0
340cpu7: MSR_MISC_FEATURE_CONTROL: 0x00000000 (L2-Prefetch L2-Prefetch-pair L1-Prefetch L1-IP-Prefetch)
341cpu0: MSR_PM_ENABLE: 0x00000001 (HWP)
342cpu0: MSR_HWP_CAPABILITIES: 0x01101f53 (high 83 guar 31 eff 16 low 1)
343cpu0: MSR_HWP_REQUEST: 0x00005353 (min 83 max 83 des 0 epp 0x0 window 0x0 pkg 0x0)
344cpu0: MSR_HWP_INTERRUPT: 0x00000001 (EN_Guaranteed_Perf_Change, Dis_Excursion_Min)
345cpu0: MSR_HWP_STATUS: 0x00000004 (No-Guaranteed_Perf_Change, No-Excursion_Min)
346cpu0: EPB: 6 (balanced)
347cpu0: MSR_RAPL_POWER_UNIT: 0x000a0e03 (0.125000 Watts, 0.000061 Joules, 0.000977 sec.)
348cpu0: MSR_PKG_POWER_INFO: 0x00000168 (45 W TDP, RAPL 0 - 0 W, 0.000000 sec.)
349cpu0: MSR_PKG_POWER_LIMIT: 0x42820800218208 (UNlocked)
350cpu0: PKG Limit #1: ENabled (65.000 Watts, 64.000000 sec, clamp ENabled)
351cpu0: PKG Limit #2: ENabled (65.000 Watts, 0.002441* sec, clamp DISabled)
352cpu0: MSR_VR_CURRENT_CONFIG: 0x00000000
353cpu0: PKG Limit #4: 0.000000 Watts (UNlocked)
354cpu0: MSR_DRAM_POWER_LIMIT: 0x5400de00000000 (UNlocked)
355cpu0: DRAM Limit: DISabled (0.000 Watts, 0.000977 sec, clamp DISabled)
356cpu0: MSR_PP0_POLICY: 0
357cpu0: MSR_PP0_POWER_LIMIT: 0x00000000 (UNlocked)
358cpu0: Cores Limit: DISabled (0.000 Watts, 0.000977 sec, clamp DISabled)
359cpu0: MSR_PP1_POLICY: 0
360cpu0: MSR_PP1_POWER_LIMIT: 0x00000000 (UNlocked)
361cpu0: GFX Limit: DISabled (0.000 Watts, 0.000977 sec, clamp DISabled)
362cpu0: MSR_IA32_TEMPERATURE_TARGET: 0x00640000 (100 C) (100 default - 0 offset)
363cpu0: MSR_IA32_PACKAGE_THERM_STATUS: 0x88200800 (68 C)
364cpu0: MSR_IA32_PACKAGE_THERM_INTERRUPT: 0x00000003 (100 C, 100 C)
365cpu7: MSR_PKGC3_IRTL: 0x0000884e (valid, 79872 ns)
366cpu7: MSR_PKGC6_IRTL: 0x00008876 (valid, 120832 ns)
367cpu7: MSR_PKGC7_IRTL: 0x00008894 (valid, 151552 ns)
368cpu7: MSR_PKGC8_IRTL: 0x000088fa (valid, 256000 ns)
369cpu7: MSR_PKGC9_IRTL: 0x0000894c (valid, 339968 ns)
370cpu7: MSR_PKGC10_IRTL: 0x00008bf2 (valid, 1034240 ns)
371.fi
372.PP
373The \fBmax efficiency\fP frequency, a.k.a. Low Frequency Mode, is the frequency
374available at the minimum package voltage.  The \fBTSC frequency\fP is the base
375frequency of the processor -- this should match the brand string
376in /proc/cpuinfo.  This base frequency
377should be sustainable on all CPUs indefinitely, given nominal power and cooling.
378The remaining rows show what maximum turbo frequency is possible
379depending on the number of idle cores.  Note that not all information is
380available on all processors.
381.SH ADD COUNTER EXAMPLE
382Here we limit turbostat to showing just the CPU number for cpu0 - cpu3.
383We add a counter showing the 32-bit raw value of MSR 0x199 (MSR_IA32_PERF_CTL),
384labeling it with the column header, "PRF_CTRL", and display it only once,
385after the conclusion of a 0.1 second sleep.
386.nf
387sudo ./turbostat --quiet --cpu 0-3 --show CPU --add msr0x199,u32,raw,PRF_CTRL sleep .1
3880.101604 sec
389CPU	  PRF_CTRL
390-	0x00000000
3910	0x00000c00
3921	0x00000800
3932	0x00000a00
3943	0x00000800
395
396.fi
397
398.SH ADD PERF COUNTER EXAMPLE
399Here we limit turbostat to showing just the CPU number for cpu0 - cpu3.
400We add a counter showing time spent in C1 core cstate,
401labeling it with the column header, "pCPU%c1", and display it only once,
402after the conclusion of 0.1 second sleep.
403We also show CPU%c1 built-in counter that should show similar values.
404.nf
405sudo ./turbostat --quiet --cpu 0-3 --show CPU,CPU%c1 --add perf/cstate_core/c1-residency,cpu,delta,percent,pCPU%c1 sleep .1
4060.102448 sec
407CPU     pCPU%c1 CPU%c1
408-       34.89   34.89
4090       45.99   45.99
4101       45.94   45.94
4112       23.83   23.83
4123       23.84   23.84
413
414.fi
415
416.SH ADD PERF COUNTER EXAMPLE #2 (using virtual cpu device)
417Here we run on hybrid, Meteor Lake platform.
418We limit turbostat to show output for just cpu0 (pcore) and cpu4 (ecore).
419We add a counter showing number of L3 cache misses, using virtual "cpu" device,
420labeling it with the column header, "VCMISS".
421We add a counter showing number of L3 cache misses, using virtual "cpu_core" device,
422labeling it with the column header, "PCMISS". This will fail on ecore cpu4.
423We add a counter showing number of L3 cache misses, using virtual "cpu_atom" device,
424labeling it with the column header, "ECMISS". This will fail on pcore cpu0.
425We display it only once, after the conclusion of 0.1 second sleep.
426.nf
427sudo ./turbostat --quiet --cpu 0,4 --show CPU --add perf/cpu/cache-misses,cpu,delta,VCMISS --add perf/cpu_core/cache-misses,cpu,delta,PCMISS --add perf/cpu_atom/cache-misses,cpu,delta,ECMISS sleep 5
428turbostat: added_perf_counters_init_: perf/cpu_atom/cache-misses: failed to open counter on cpu0
429turbostat: added_perf_counters_init_: perf/cpu_core/cache-misses: failed to open counter on cpu4
4305.001207 sec
431CPU	ECMISS		PCMISS		VCMISS
432-	41586506	46291219	87877749
4334	83173012	0		83173040
4340	0		92582439	92582458
435.fi
436
437.SH ADD PMT COUNTER EXAMPLE
438Here we limit turbostat to showing just the CPU number 0.
439We add two counters, showing crystal clock count and the DC6 residency.
440All the parameters passed are based on the metadata found in the PMT XML files.
441
442For the crystal clock count, we
443label it with the column header, "XTAL",
444we set the type to 'raw', to read the number of clock ticks in hex,
445we set the format to 'delta', to display the difference in ticks during the measurement interval,
446we set the domain to 'package0', to collect it and associate it with the whole package number 0,
447we set the offset to '0', which is a offset of the counter within the PMT MMIO region,
448we set the lsb and msb to cover all 64 bits of the read 64 bit value,
449and finally we set the guid to '0x1a067102', that identifies the PMT MMIO region to which the 'offset' is applied to read the counter value.
450
451For the DC6 residency counter, we
452label it with the column header, "Die%c6",
453we set the type to 'txtal_time', to obtain the percent residency value
454we set the format to 'delta', to display the difference in ticks during the measurement interval,
455we set the domain to 'package0', to collect it and associate it with the whole package number 0,
456we set the offset to '0', which is a offset of the counter within the PMT MMIO region,
457we set the lsb and msb to cover all 64 bits of the read 64 bit value,
458and finally we set the guid to '0x1a067102', that identifies the PMT MMIO region to which the 'offset' is applied to read the counter value.
459
460.nf
461sudo ./turbostat --quiet --cpu 0 --show CPU --add pmt,name=XTAL,type=raw,format=delta,domain=package0,offset=0,lsb=0,msb=63,guid=0x1a067102 --add pmt,name=Die%c6,type=txtal_time,format=delta,domain=package0,offset=120,lsb=0,msb=63,guid=0x1a067102
4620.104352 sec
463CPU                   XTAL      Die%c6
464-       0x0000006d4d957ca7      0.00
4650       0x0000006d4d957ca7      0.00
4660.102448 sec
467.fi
468
469.SH INPUT
470
471For interval-mode, turbostat will immediately end the current interval
472when it sees a newline on standard input.
473turbostat will then start the next interval.
474Control-C will be send a SIGINT to turbostat,
475which will immediately abort the program with no further processing.
476.SH SIGNALS
477
478SIGINT will interrupt interval-mode.
479The end-of-interval data will be collected and displayed before turbostat exits.
480
481SIGUSR1 will end current interval,
482end-of-interval data will be collected and displayed before turbostat
483starts a new interval.
484.SH NOTES
485
486.B "turbostat "
487must be run as root.
488Alternatively, non-root users can be enabled to run turbostat this way:
489
490# setcap cap_sys_admin,cap_sys_rawio,cap_sys_nice=+ep path/to/turbostat
491
492# chmod +r /dev/cpu/*/msr
493
494# chmod +r /dev/cpu_dma_latency
495
496.B "turbostat "
497reads hardware counters, but doesn't write them.
498So it will not interfere with the OS or other programs, including
499multiple invocations of itself.
500
501\fBturbostat \fP
502may work poorly on Linux-2.6.20 through 2.6.29,
503as \fBacpi-cpufreq \fPperiodically cleared the APERF and MPERF MSRs
504in those kernels.
505
506AVG_MHz = APERF_delta/measurement_interval.  This is the actual
507number of elapsed cycles divided by the entire sample interval --
508including idle time.  Note that this calculation is resilient
509to systems lacking a non-stop TSC.
510
511TSC_MHz = TSC_delta/measurement_interval.
512On a system with an invariant TSC, this value will be constant
513and will closely match the base frequency value shown
514in the brand string in /proc/cpuinfo.  On a system where
515the TSC stops in idle, TSC_MHz will drop
516below the processor's base frequency.
517
518Busy% = MPERF_delta/TSC_delta
519
520Bzy_MHz = TSC_delta*APERF_delta/MPERF_delta/measurement_interval
521
522Note that these calculations depend on TSC_delta, so they
523are not reliable during intervals when TSC_MHz is not running at the base frequency.
524
525Turbostat data collection is not atomic.
526Extremely short measurement intervals (much less than 1 second),
527or system activity that prevents turbostat from being able
528to run on all CPUS to quickly collect data, will result in
529inconsistent results.
530
531The APERF, MPERF MSRs are defined to count non-halted cycles.
532Although it is not guaranteed by the architecture, turbostat assumes
533that they count at TSC rate, which is true on all processors tested to date.
534
535.SH REFERENCES
536Volume 3B: System Programming Guide"
537https://www.intel.com/products/processor/manuals/
538
539.SH RUN THE LATEST VERSION
540If turbostat complains that it doesn't recognize your processor,
541please try the latest version.
542
543The latest version of turbostat does not require the latest version of the Linux kernel.
544However, some features, such as perf(1) counters, do require kernel support.
545
546The latest turbostat release is available in the upstream Linux Kernel source tree.
547eg. "git pull https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git"
548and run make in tools/power/x86/turbostat/.
549
550n.b. "make install" will update your system manually, but a distro update may subsequently downgrade your turbostat to an older version.
551For this reason, manually installing to /usr/local/bin may be what you want.
552
553Note that turbostat/Makefile has a "make snapshot" target, which will create a tar file
554that can build without a local kernel source tree.
555
556If the upstream version isn't new enough, the development tree can be found here:
557"git pull https://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux.git turbostat"
558
559If the development tree doesn't work, please contact the author via chat,
560or via email with the word "turbostat" on the Subject line.
561
562.SH FILES
563.ta
564.nf
565/sys/bus/event_source/devices/
566/dev/cpu/*/msr
567/sys/class/intel_pmt/
568/sys/devices/system/cpu/
569.fi
570
571.SH "SEE ALSO"
572perf(1), msr(4), vmstat(8)
573.PP
574.SH AUTHOR
575.nf
576Written by Len Brown <len.brown@intel.com>