Skip to main content
TEST
All testsHow tests work
TEST

Real browser measurements. No invented hardware data.

How tests workPrivacyAll tests

© 2026

Method 3.0

Scores you can understand—and challenge.

A browser score is only useful when its method is visible. Here is exactly what method 3.0 runs, how eligible measurements are combined, and where the limits are.

Calibrate, sustain, verify

Performance calibrates first, repeats long samples, checks output, and measures slowdown instead of trusting one quick burst.

Run sequentially

CPU, rendering, display, network, and audit work never run together and distort each other.

Record the context

Every saved run includes method version, browser, platform, viewport, display, and exposed capability hints.

Score model

Five reports. Eligible signals form the overall score.

Method 3.0 gives CPU and heavy rendering more weight. Performance combines single-worker, parallel-worker, memory, JSON, and DOM measurements, then applies repeatability and sustained-speed penalties. Accessibility is reported separately, private-LAN network measurements are excluded, and display cadence is excluded under automation.

DomainReference weightOverall useMeasured signals
Performance40%Included when measuredSeven sustained CPU samples plus multi-core, memory, JSON, and DOM throughput.
Gaming30%Included when measuredHeavy visible Canvas 2D loads, peak-load FPS, 1% low, and frame pacing.
Network15%Public-server runs onlyEight round trips and an adaptive uncached payload transfer.
Display10%Excluded under automationVisible browser presentation cadence and timing stability.
Accessibility5%Reported separatelyCustom audit of the current Test page and exposed preferences.

Measured phases

What actually runs—and for how long.

A Standard method 3.0 run usually takes about 30–50 seconds. Extreme intentionally runs for roughly 75–120 seconds. Device speed, frame scheduling, server distance, and transfer rate can change this.

PerformanceUsually 12–25 seconds
Calibration, seven sustained CPU samples, three parallel-worker samples, 3.75 GB of memory copies, 300,000 JSON records, and 18,000 DOM nodes.
GamingAbout 8 seconds
An unloaded baseline followed by 2,500, 8,000, and 24,000-object visible Canvas 2D loads.
DisplayAbout 4 seconds
A 600 ms warm-up followed by 3 seconds of visible requestAnimationFrame intervals.
NetworkUsually 2–10 seconds
A warm-up, eight same-origin round trips, 512 KB calibration, and an adaptive 2–16 MB transfer. Slow routes can take longer.
AccessibilityUnder 1 second
A custom DOM audit of this page's controls, names, landmarks, headings, targets, language, and exposed preferences.

Reliable comparisons

Control what you can.

  • Use the same browser window size and zoom.
  • Keep the tab visible and avoid moving the window.
  • Close heavy applications, streams, and background tabs.
  • Use the same plugged-in or battery power mode.
  • Run two or three times and compare the middle result.
Important limitations

Test measures a browser session, not the maximum native capability of a CPU, GPU, display, or internet plan. Canvas FPS is limited by browser presentation cadence. Thermal state, battery policy, extensions, server distance, VPNs, browser throttling, variable refresh rate, and background work can all change a result.

Network results describe the route to the server hosting Test; localhost and private-LAN measurements are reported but excluded from the overall score. Accessibility is a custom automated audit of the current Test page and exposed preferences. It does not certify WCAG compliance or score a person’s needs.

Ready for a controlled run?

The dashboard records method 3.0 and the browser context with every result.

Open the browser lab