Technically honest, by design

Know what a result can tell you.

Browser tools can reveal useful observations. They cannot certify every part of your hardware.

Different evidence, clearly labeled

Automatically verified means a particular browser operation was observed. User verified means you confirmed a visible, audible or tactile observation. Informational values describe exposed state. Approximate values are estimates with sampling and browser limitations. Unsupported, permission denied, skipped and failed-to-run states are distinct from a hardware finding.

Timing is a browser observation

Animation-frame cadence is not certified panel refresh. Pointer-event rate is not USB polling. Key-event timing is not physical switch latency. Operating system load, coalescing, scheduling and privacy protections affect samples.

Klyvero Benchmark v1

Workloads measure browser execution, workers, typed-array operations, drawing, rendering and optional graphics backends. The protocol uses three one-second warmups and seven scored one-second samples with 250 ms rest gaps. Raw samples, median and dispersion are reported. Background or interrupted runs are invalidated, not silently scored.

Workload source hashes and method version identify comparable runs. WebGL rendering and WebGPU compute are different workloads. Scores, percentiles and rankings remain unavailable until real reference calibration is approved. No equivalence with Geekbench, Cinebench, 3DMark or native CPU/GPU diagnostics is claimed.

No whole-device pass

A successful browser operation is not proof that hardware is fault-free. Guided summaries preserve observations, user reports, skipped and unavailable checks rather than calculating a device-health percentage.