What the browser can actually tell you about your hardware (and what it can't)

Building browser-based hardware diagnostics led to understanding where web platform capabilities are genuine and where they mislead. The refresh rate signal is limited to requestAnimationFrame, with no direct screen.refreshRate. The only way to estimate it is by timing requestAnimationFrame callbacks and calculating median frame delta.

There are two pitfalls: averaging can be affected by a single dropped frame, and rAF is throttled in background tabs, rendering tab-invisibility measurements meaningless. Screen dimensions have four different answers, all valid. screen.width, window.innerWidth, window.devicePixelRatio, and screen.availWidth measure distinct aspects, with screen.width * devicePixelRatio representing the actual native panel resolution.

However, this does not reveal true hardware resolution. Keyboard event.code vs event.key comparison shows event.key being layout-dependent while event.code corresponds to physical position. Certain keys, like PrintScreen, lack JS detection, and Meta combinations might be obscured by the OS. N-key rollover testing exhibits a Set size growth pattern when simultaneously held keys are monitored.

Pointer events unify mouse, touch, and pen inputs, with pointerId facilitating multi-touch tracking. Measuring the gap between consecutive pointerdown events on the same button can detect worn-switch double-click faults. PreventDefault() on pointerdown halts compatibility mouse events, potentially breaking code expecting them. Gamepad API is polling-based and dormant until user interaction, with axis movement not being event-driven.

This causes a one-hour confusion regarding resting axis values indicating potential orientation defects. Media device labels are gated behind permission, revealing empty strings before getUserMedia() success. For microphone level metering, AnalyserNode with getByteTimeDomainData and RMS computation is more stable than frequency data.

Dead pixels cannot be detected programmatically, necessitating visual inspection through Fullscreen API and solid-color viewport filling. Despite these limitations, these APIs offer relative measurements, permission-protected labels, and input-protected device lists, serving as the right trade-off for fingerprinting resistance. The full set of demos can be found here for further comparison.