So, I am currently working on a Linux tool+daemon to replace auto-cpufreq with a more efficient, tunable and observable alternative.

And obviously, to tune a CPU well, knowing how hot it is is a requirement.

Then, how do you actually see how hot your CPU is in Linux? Using /sys/class/hwmon of course.

This path contains hardware monitoring devices. In the laptop I'm testing this on, hwmon4 under this directory corresponds to the CPU.

Let's run a tree:

/sys/class/hwmon/hwmon4 -> ../../devices/platform/coretemp.0/hwmon/hwmon4
├── device -> ../../../coretemp.0
│   ├── driver_override
│   ├── hwmon
│   ├── modalias
│   ├── power
│   ├── subsystem -> ../../../bus/platform
│   └── uevent
├── name
├── power
│   ├── autosuspend_delay_ms
│   ├── control
│   ├── runtime_active_time
│   ├── runtime_status
│   └── runtime_suspended_time
├── subsystem -> ../../../../../class/hwmon  [recursive, not followed]
├── temp10_crit
├── temp10_crit_alarm
├── temp10_input
├── temp10_label
├── temp10_max
├── temp14_crit
├── temp14_crit_alarm
├── temp14_input
├── temp14_label
├── temp14_max
├── temp18_crit
├── temp18_crit_alarm
├── temp18_input
├── temp18_label
├── temp18_max
├── temp1_crit
├── temp1_crit_alarm
├── temp1_input
├── temp1_label
├── temp1_max
├── temp22_crit
├── temp22_crit_alarm
├── temp22_input
├── temp22_label
├── temp22_max
├── temp26_crit
├── temp26_crit_alarm
├── temp26_input
├── temp26_label
├── temp26_max
├── temp2_crit
├── temp2_crit_alarm
├── temp2_input
├── temp2_label
├── temp2_max
├── temp30_crit
├── temp30_crit_alarm
├── temp30_input
├── temp30_label
├── temp30_max
├── temp34_crit
├── temp34_crit_alarm
├── temp34_input
├── temp34_label
├── temp34_max
├── temp35_crit
├── temp35_crit_alarm
├── temp35_input
├── temp35_label
├── temp35_max
├── temp36_crit
├── temp36_crit_alarm
├── temp36_input
├── temp36_label
├── temp36_max
├── temp37_crit
├── temp37_crit_alarm
├── temp37_input
├── temp37_label
├── temp37_max
├── temp38_crit
├── temp38_crit_alarm
├── temp38_input
├── temp38_label
├── temp38_max
├── temp39_crit
├── temp39_crit_alarm
├── temp39_input
├── temp39_label
├── temp39_max
├── temp40_crit
├── temp40_crit_alarm
├── temp40_input
├── temp40_label
├── temp40_max
├── temp41_crit
├── temp41_crit_alarm
├── temp41_input
├── temp41_label
├── temp41_max
├── temp42_crit
├── temp42_crit_alarm
├── temp42_input
├── temp42_label
├── temp42_max
├── temp43_crit
├── temp43_crit_alarm
├── temp43_input
├── temp43_label
├── temp43_max
├── temp44_crit
├── temp44_crit_alarm
├── temp44_input
├── temp44_label
├── temp44_max
├── temp45_crit
├── temp45_crit_alarm
├── temp45_input
├── temp45_label
├── temp45_max
├── temp46_crit
├── temp46_crit_alarm
├── temp46_input
├── temp46_label
├── temp46_max
├── temp47_crit
├── temp47_crit_alarm
├── temp47_input
├── temp47_label
├── temp47_max
├── temp48_crit
├── temp48_crit_alarm
├── temp48_input
├── temp48_label
├── temp48_max
├── temp49_crit
├── temp49_crit_alarm
├── temp49_input
├── temp49_label
├── temp49_max
├── temp6_crit
├── temp6_crit_alarm
├── temp6_input
├── temp6_label
├── temp6_max
└── uevent

Let's cat all the _label files:

/sys/class/hwmon/hwmon4/temp1_label:
Package id 0
/sys/class/hwmon/hwmon4/temp2_label:
Core 0
/sys/class/hwmon/hwmon4/temp6_label:
Core 4
/sys/class/hwmon/hwmon4/temp10_label:
Core 8
/sys/class/hwmon/hwmon4/temp14_label:
Core 12
/sys/class/hwmon/hwmon4/temp18_label:
Core 16
/sys/class/hwmon/hwmon4/temp22_label:
Core 20
/sys/class/hwmon/hwmon4/temp26_label:
Core 24
/sys/class/hwmon/hwmon4/temp30_label:
Core 28
/sys/class/hwmon/hwmon4/temp34_label:
Core 32
/sys/class/hwmon/hwmon4/temp35_label:
Core 33
/sys/class/hwmon/hwmon4/temp36_label:
Core 34
/sys/class/hwmon/hwmon4/temp37_label:
Core 35
/sys/class/hwmon/hwmon4/temp38_label:
Core 36
/sys/class/hwmon/hwmon4/temp39_label:
Core 37
/sys/class/hwmon/hwmon4/temp40_label:
Core 38
/sys/class/hwmon/hwmon4/temp41_label:
Core 39
/sys/class/hwmon/hwmon4/temp42_label:
Core 40
/sys/class/hwmon/hwmon4/temp43_label:
Core 41
/sys/class/hwmon/hwmon4/temp44_label:
Core 42
/sys/class/hwmon/hwmon4/temp45_label:
Core 43
/sys/class/hwmon/hwmon4/temp46_label:
Core 44
/sys/class/hwmon/hwmon4/temp47_label:
Core 45
/sys/class/hwmon/hwmon4/temp48_label:
Core 46
/sys/class/hwmon/hwmon4/temp49_label:
Core 47

Notice something? I do:

  • temp1_label is Package id 0 - what is that?
  • The core numbers make no sense. This device only has 32 cores, there shouldn't be any gaps within numbers and the numbers shouldn't go that high.

The explanation for the first point is simple, looking at the kernel coretemp.c implementation, we can see that it is the temperature of the CPU as a whole:

static ssize_t show_label(
  struct device *dev,
  struct device_attribute *devattr,
  char *buf
) {
  struct platform_data *pdata = dev_get_drvdata(dev);
  struct temp_data *tdata = container_of(devattr, struct temp_data, sd_attrs[ATTR_LABEL]);

  if (is_pkg_temp_data(tdata))
    return sprintf(buf, "Package id %u\n", pdata->pkg_id);

  return sprintf(buf, "Core %u\n", tdata->cpu_core_id);
}

That leaves us the second question. Why do we have CPUs 1, 2, 6, 10, 14, 18, 22, 26, 30, 34-49, instead of the expected 0-31?

It turns out that CPU It's common for chip manufacturers to disable faulty or degraded cores before shipping. Or if a lower tier SKU^1 is selling more, cores of higher tier SKUs are disabled to match expectations.

This process is called binning, and it exists to not waste silicon. Since silicon manufacturing is not perfect, some CPUs are more faulty, thus slower than others. The process determines how faulty a CPU is and sorts them into "bins". This is also why lower tier CPUs who are almost exactly the same as higher tier CPUs exist.

The likelihood of faulty silicon also increases with the smaller the architecture size gets (the Apple M4 is 4m, which is crazy), so this method of recycling worse chips is becoming much more valuable by the day.

So, in summary the 32 core CPU I was testing this on was most likely just the 64 core version with a bunch of the cores disabled.

  1. Stock Keeping Unit..?