A browser test measures the path visible to the browser.
Results include more than the physical component. The operating system, browser engine, permissions, driver, power mode, thermal state and competing work can all affect a measurement. A successful test confirms that the complete path worked during that run; it does not certify every part independently.
We prefer direct observations over guessed hardware details. When privacy protections hide a model name or capability, the page reports that limitation instead of attempting to fingerprint the device.
Permissions and browser events provide the signal.
The microphone meter reads the level of a live audio stream obtained after permission is granted. The webcam test displays the live camera stream and reports settings the browser exposes. Neither tool records or uploads media. Keyboard, mouse, touchscreen and gamepad pages listen only for relevant input events while the tester is active.
The dead-pixel page presents solid colors for visual inspection; it cannot inspect the panel electronically. Screen resolution uses browser-reported dimensions and pixel ratio. Refresh rate is estimated from the spacing between successive animation frames while the tab is visible.
The result combines the person and the device.
Reaction time starts from a randomized visual change and ends on the next accepted click or tap using the browser's high-resolution clock. Click speed divides accepted clicks by the five-second test window. Typing speed converts correctly entered characters into standard five-character words and normalizes them to one minute.
Display latency, refresh rate, input hardware, touch processing and browser scheduling are part of the measured chain. Compare repeated attempts on the same setup rather than treating one score as a permanent personal characteristic.
A fixed JavaScript workload in isolated workers.
The single-worker pass runs for about 900 milliseconds. The parallel pass runs for about 1,050 milliseconds using the smaller of the browser-reported logical-thread count or eight workers. Each worker repeats an inner loop containing integer multiplication, bitwise mixing and a square root, then reports completed loop iterations per second.
The displayed score is the rounded sum of single-worker iterations per second divided by 100,000 and multi-worker iterations per second divided by 250,000. It is a Test My Everything browser score, not a vendor benchmark and not a measurement of clock frequency, temperature or individual CPU instructions.
A timed WebGL rendering loop with capability checks.
The page creates a 512 × 512 off-screen WebGL 2 context when available, otherwise WebGL 1. It compiles a small vertex and fragment shader, repeatedly draws a full-canvas triangle for about 1.25 seconds and synchronizes the graphics work before counting draw calls per second.
The browser graphics score is rounded draw calls per second divided by 25. The report also shows WebGPU availability, the active WebGL version, maximum texture dimension and any renderer name the browser chooses to expose. It does not measure gaming frame rate, VRAM, ray tracing, power use or temperature.
A temporary 32 MiB file inside private browser storage.
The test creates random data, writes it in 4 MiB chunks to origin-private storage, closes the file, reads it back and requests deletion. Throughput is calculated from 32 MiB divided by the measured write or read time. Personal folders and existing files are never scanned.
Operating-system caches, browser buffering, available memory and storage quotas can make a short result differ sharply from sustained SSD performance. Use it to compare the same browser storage path under similar conditions, not to identify a drive model or validate manufacturer specifications.
Generated traffic to Cloudflare's edge network.
The network test uses Cloudflare's browser speed-test library. It sends eight latency probes, then runs staged download and upload measurements using 100 kB, 1 MB and 10 MB payloads. A complete run transfers no more than roughly 25 MB of generated test data. Result-logging endpoints are disabled.
Download and upload are reported in megabits per second. Ping is round-trip latency and jitter describes variation between latency samples. The route, Wi-Fi conditions, VPN, congestion and selected edge location can all change the result; it is not a guaranteed maximum from the internet provider.
Control the conditions before comparing two runs.
- Use the same device, browser version and test page.
- Keep the tab visible and close demanding background work.
- Use the same power mode; connect a laptop to power for maximum-performance comparisons.
- Let a hot device cool, then run at least three times and compare the middle result.
- For network tests, keep the same connection type, room position and VPN state.