Hi everyone,
I’m investigating unusually high pch_haswell temperature readings on an HP Chromebox G1 running Ubuntu Server with MrChromebox UEFI firmware.
MrChromebox mentioned a Linux PCH temperature reading bug affecting Haswell/Broadwell in this discussion. Could this explain my readings? Is there a known cause, kernel patch, or recommended way to verify them?
Hardware and software
- Device: HP Chromebox G1 / CB1
- Board: Google ZAKO
- CPU: Intel Core i7-4600U, 2 cores / 4 threads
- RAM: 8 GB
- Storage: LITEON IT LST-16S9G-HP 16 GB SATA SSD
- Firmware:
MrChromebox-2609.3 - DMI firmware date:
10/28/2026 - OS: Fresh Ubuntu Server 24.04.4 LTS installation
- Kernel:
6.8.0-147-generic, x86-64 - Boot: UEFI, Secure Boot disabled
- PCH thermal driver:
intel_pch_thermal - PCH thermal PCI device: Intel
8086:9c24
Cold-boot readings
Approximately 17 seconds after the recorded kernel start, sensors showed:
coretemp-isa-0000
Adapter: ISA adapter
Package id 0: +30.0°C (high = +100.0°C, crit = +100.0°C)
Core 0: +29.0°C (high = +100.0°C, crit = +100.0°C)
Core 1: +29.0°C (high = +100.0°C, crit = +100.0°C)
acpitz-acpi-0
Adapter: ACPI interface
temp1: +30.0°C
pch_haswell-virtual-0
Adapter: Virtual device
temp1: +79.0°C
The chassis felt cool. I have not independently measured the PCH itself: accessing it would require removing the motherboard.
Idle observations
During a 60-second passive recording:
- CPU package: typically around 43–45°C after an initial transient
- ACPI temperature: settled around 44°C
- PCH: 95–96°C
Later, immediately before a separate short load test:
- CPU package: 46°C
- ACPI: 47°C
- PCH: 98.5°C
- Load average: approximately zero
The PCH values reported through hwmon and the thermal zone use the same underlying source, so these are not independent measurements.
Short, temperature-limited CPU tests
Two brief tests were performed using two Python CPU workers. Each had a hard maximum runtime of 20 seconds, an independent monitoring process, and additional checks in the recording script. The CPU stop threshold was 85°C. Neither test reached its full runtime.
The first test stopped after approximately 1.2 seconds when the PCH reading reached 100°C. The highest recorded CPU temperature was 63°C.
For the second test, the PCH stop threshold was explicitly raised to 110°C. It stopped after approximately 8.2 seconds.
| Sensor | Immediately before load | Recorded maximum | After 30 seconds of recovery |
|---|---|---|---|
| CPU package | 46°C | 71°C | 46°C |
| CPU core 0 | 44°C | 71°C | 44°C |
| CPU core 1 | 45°C | 68°C | 45°C |
| ACPI thermal zone | 47°C | 71°C | 47°C |
| PCH | 98.5°C | 110°C | 98.5°C |
The CPU cooled from approximately 70°C to 53°C within about one second after stopping the second test. The PCH reading declined more slowly.
The fan became audible during the second test, and the device produced noticeable warm exhaust. This confirms a physical fan response, although no RPM sensor was available.
ACPI cooling configuration
Ubuntu detects four logical ACPI fan devices, FAN0 through FAN3, using the acpi-fan driver. They are bound to the acpitz thermal zone.
The reported active trip points are:
- 50°C
- 67°C
- 75°C
- 90°C
The ACPI zone also reports a passive trip point at 90°C and a critical trip point at 98°C.
During the second test, FAN2 and FAN3 were reported active at an ACPI temperature of approximately 69°C. A lower cooling stage remained active after cooling down.
The pch_haswell zone has no fan cooling-device bindings and reports only a critical trip point at 120°C.
Logs and throttling
- No new kernel messages were recorded during the second test.
- All exposed CPU core/package thermal-throttling counters remained zero before and after testing.
- No failed systemd units were found.
thermaldis running.
Potentially relevant boot messages:
ACPI: _OSC evaluation for CPUs failed, trying _PDC
cros_ec_lpcs cros_ec_lpcs.0: EC ID not detected
thermald also reports:
sensor id 6 : No temp sysfs for reading raw temp
Config file /etc/thermald/thermal-conf.xml does not exist
Polling mode is enabled: 4
I do not know whether these messages are relevant.
Questions
- Does this match the known Linux PCH temperature-reporting issue on Haswell/Broadwell Chromeboxes?
- Is there a specific kernel issue, patch, or technical explanation for it?
- Is there a recommended read-only method to validate the PCH reading on ZAKO?
- Is this potentially related to firmware sensor initialization, or is it strictly a Linux driver issue?
The CPU/PCH difference is approximately 49–52.5°C at cold boot and idle, but it changes under load. I therefore have not applied a fixed temperature correction.
No BIOS, firmware, fan-control settings, or hardware thermal limits were changed during testing.
Thanks for any pointers!
