Skip to content
Windows.HomesUnofficial

Laptop throttling: tell a thermal wall from a power wall

Verified 2 min read

Separate heat from firmware power limits using Windows counters, then decide whether repasting, a cooler or a different power plan can help.

Tested onPro (26H2) · Build 26300.9457 · x64
E1
Build
26300.9457
Edition
Pro (26H2)
Architecture
x64
Support
Documentation (E1)
Evidence
E1Official documentation
Verified
2026-10-29
Page type
Method

The short answer

Section titled “The short answer”

“Throttling” is two different things, and the fix differs:

  • Thermal limit — the CPU/GPU hits a temperature ceiling and reduces clocks to stay under it. Fix: cooling, airflow, repaste, lower ambient.
  • Power limit — firmware caps sustained watts regardless of temperature. Fix: a different power profile, a BIOS setting, or nothing (it is the design).

A laptop that is cool and slow is power-limited. A laptop that is hot and slow is thermally limited. Measuring both separates them.

Measure frequency and temperature together

Section titled “Measure frequency and temperature together”
Terminal window
# effective frequency versus nominal
Get-Counter '\Processor Information(_Total)\% of Maximum Frequency',
            '\Processor Information(_Total)\% Processor Performance' -MaxSamples 10 -SampleInterval 2


# per-core temperature is not exposed by Windows; use the OEM tool or a sensor library
Get-CimInstance -Namespace root\wmi -ClassName MSAcpi_ThermalZoneTemperature -ErrorAction SilentlyContinue |
  Select-Object InstanceName,CurrentTemperature

Two notes:

  • % of Maximum Frequency is relative to the nominal clock, so 100% does not mean “boosting”. A value stuck near 60% under load is a throttle.
  • MSAcpi_ThermalZoneTemperature is a firmware ACPI zone, not a core sensor. Many laptops report nothing useful there; OEM tools or third-party sensor readers are more reliable.

The experiment that answers the question

Section titled “The experiment that answers the question”
  1. Run a steady load for 10 minutes (not a 30-second burst).
  2. Record % of Maximum Frequency, package temperature, and package power every 2 seconds.
  3. Plot the three over time.

Reading the shape:

Pattern Meaning
Clocks drop as temperature rises to a ceiling thermal
Clocks drop immediately, temperature stays moderate power limit
Clocks oscillate with fan ramp thermal with hysteresis
Clocks fine on battery, drop on AC power policy profile
Clocks fine for 2 minutes then collapse small heatsink saturating (thermal, slow)

What actually helps

Section titled “What actually helps”
Cause Effective fix
Dust in heatsink clean it — often 10–20 °C
Dried thermal paste (3+ years) repaste with a quality compound
Laptop on soft surface hard surface or a stand
Power profile set to “quiet” balanced/performance profile
Firmware cap (thin-and-light design) nothing — it is the design
Undervolting (if supported) lower voltage at same clocks, less heat

Undervolting and fan-curve changes are vendor/firmware specific and increasingly locked. Do not follow a guide written for a different model — the settings often do not exist, or exist under a different name.

When it is not throttling at all

Section titled “When it is not throttling at all”
  • Battery mode: many laptops cap sustained power hard on battery, by design.
  • Single-thread workloads: they never load all cores, so package power stays low — normal.
  • Storage-bound work: CPU waits; frequency drops because there is nothing to do.
  • Windows power slider: the “best power efficiency” position caps performance deliberately.
Section titled “Related”

Lab boundary

Section titled “Lab boundary”

E1. The counters and power-state model are Microsoft-documented. The lab has no physical laptop, so no temperature or clock data is published — the method is what this page offers.

Sources