ApexMark › Methodology
How ApexMark measures performance
A browser benchmark cannot simply trust the clock. This page explains the rules that keep ApexMark results comparable.
Run the benchmarkVersioned protocols
Every workload has a version and a source hash. If the measuring code changes, results move to a new scoreboard rather than being mixed with older ones.
Verified work
The CPU test checks completed ray-traced image sections and its cache lookup chain. The memory test verifies copied data and dependent lookups. The GPU test checks both the visual resolution ladder and its WebGPU compute profile. Incomplete or unverified work cannot earn a comparable category score.
A frozen baseline
Each scored category is measured against the frozen ApexMark 1.0 baseline of 1,000 points. The overall score combines CPU, GPU and memory; browser storage and internet are reported separately. Different measurement protocols keep separate scoreboards, so their points are not directly interchangeable.
Honest limits
Browser storage includes managed caches and is not a raw SSD benchmark. Results depend on browser, driver, power mode and temperature, and ApexMark is experimental. For the closest comparison, use the same protocol, browser family and version under similar conditions.