Back to learning center
Methodology
8 min read

Thermal Throttling: How Browser GPU Benchmarks Drift

Learn how thermal throttling affects browser GPU benchmarks. Covers heat, boost clocks, power profiles, sustained WebGL workloads, and clear result labels.

Published 2026-05-07/Updated 2026-05-09

What you'll learn

  • How thermal throttling appears in FPS and frame-time graphs.
  • Why volume shaders are useful for sustained browser GPU stress tests.
  • How cold, warmed, and sustained benchmark labels prevent misleading comparisons.
  • Which notes make a thermal run reproducible.
  • Why browsers usually infer thermal behavior instead of reading GPU sensors directly.

Quick definition

Thermal throttling happens when a GPU or mobile SoC reduces clock speed to stay within heat and power limits. In browser GPU benchmarks, it often appears as falling FPS, wider frame-time spikes, or a sustained score that is lower than the first cold run.

What does thermal throttling look like in a browser GPU benchmark?

Thermal throttling shows up as a downward FPS trend or increasingly uneven frame times during a sustained run. The shader did not change; the device reduced clocks to stay within its thermal and power limits.

A browser benchmark can reveal this even without direct temperature sensors. If a cold run starts high and a later run settles lower under the same preset, sustained performance is the more honest number.

Why do volume shaders reveal thermal throttling?

Volume shaders keep many pixels busy with repeated math. Cloud and fractal presets are especially useful because they combine high fill rate with loops, noise, and iterative kernels.

That makes them different from a short UI benchmark. They can behave more like a continuous graphics workload, which is why power mode, cooling, and room temperature matter.

How should sustained thermal runs be reported?

Use separate labels for quick check, warmed run, and sustained run. Do not mix a 15-second burst with a 10-minute thermal run in the same ranking.

Report average FPS, 1% low, P95 frame time, run duration, and whether the device was plugged in. Those notes make the number more useful than a bare score.

Repeatable workflow

01Take a cold baseline

Run the selected preset after the device has been idle and record the early FPS and frame-time range.

02Hold the same workload

Keep the same preset and render size running long enough to warm the device.

03Watch the trend

A sustained drop in FPS or a wider frame-time range suggests thermal or power throttling.

04Label the result

Mark the score as quick, warmed, or sustained so it is not compared against a different test type.

Comparison notes

QuestionVolume Shader BMOther approach
Quick checkUseful for compatibility and rough shader throughput.Native benchmarks may include short graphics tests and separate stress tests.
Sustained runUseful for detecting drift in FPS and frame time under WebGL load.Dedicated stress tools often expose temperature, clocks, and fan data directly.
Result labelMust include duration and thermal state to be meaningful.Usually standardized by the benchmark suite or stress-test mode.

Key takeaways

  1. 01 Thermal throttling is visible as drift, not only as one bad frame.
  2. 02 A shader-heavy browser workload can expose sustained performance without native telemetry.
  3. 03 Quick scores and sustained scores should not share the same ranking without labels.
  4. 04 Run duration, power mode, and device temperature context are part of the result.

FAQ

Does Volume Shader BM read my GPU temperature?

No. Browser APIs generally do not expose detailed temperature sensors. The benchmark infers sustained behavior from FPS and frame-time trends.

How long should a thermal run be?

For quick screening, a few minutes can show obvious drift. For stronger evidence, use a longer fixed duration and label it clearly.

Why is a sustained score lower than a quick score?

Boost clocks can be higher at the start of a cold run. Sustained scores reflect the device after heat and power limits settle.

Related guides

External references