SpeedArcBETA
Ready
LET’S SEE WHAT YOU’VE GOT
DownloadDown—Mbps
UploadUp—Mbps
Latency
—ms
Jitter
—ms

Activities

Run a test

Latency under load

While downloading—
While uploading—
Added delay—

Speed over the test

Download Upload
Waiting for your testMbps
IPChecking…
ISPChecking…
ServerCloudflare
EdgeChecking…

Did it actually help?

01 · Measure a baseline02 · Make one change03 · Repeat and compare

Complete a test, then use it as your baseline.

Keep the device, browser, test depth, and other conditions the same. Record any connection or VPN changes in your note.

History

Start with a measurement

We’ll report what we observed and suggest a useful next step.

NO RESULT YET

    A little context. A lot more clarity.

    Download

    How fast data reaches your device. More speed means smoother streaming and quicker downloads.

    Upload

    How fast you send data. Think video calls, sharing large files, and backing up your work.

    Latency & jitter

    How quickly your connection responds, and how consistent it is. Lower is better for both.

    Methodology, privacy, and limits

    SpeedArc method 2.0 uses @cloudflare/speedtest 1.14.1. Sequential transfers increase in size and may stop early when the engine has enough transfer duration. The headline bandwidth is the 90th percentile of eligible transfer rates, not a guaranteed maximum or a sustained multi-connection capacity test. Idle and loaded latency use resource timing with server-time correction; jitter is the mean absolute difference between consecutive latency samples. Largest-payload sample ranges are descriptive quartiles, not confidence intervals.

    “Usable evidence” requires a completed, visible, uninterrupted run, all core metrics, at least ten idle samples, at least three loaded samples per direction, and three largest-payload samples per direction with an interquartile spread no greater than 25% of their median. This is an evidence check, not an accuracy certification. The diagnostic suggestion flags an added delay above 40 ms only with at least three relevant loaded-latency samples; it cannot locate the cause.

    Before-and-after comparisons require the same method, engine, profile, and destination hostname, completed visible tests, and no more than 24 hours between tests. One pair cannot establish causation. Cloudflare’s physical edge and route may change between runs. Comparison arrows show arithmetic differences, not statistical significance.

    Test traffic goes directly to Cloudflare, which receives your IP address. The ISP label comes from Cloudflare’s network metadata and may identify the network owner rather than your retail ISP. SpeedArc disables the engine’s optional result-logging endpoints. Results, raw samples, method settings, and your notes stay in memory unless you opt into local storage or export. No account or SpeedArc results server is involved. Payload estimates exclude network overhead and retries. Cloudflare privacy policy

    Activity fit is a SpeedArc estimate from this test, not a service test or a guarantee. Basic browsing reaches its top tier at 50 Mbps based on FCC web-loading observations. Streaming tiers follow YouTube playback guidance. Video-call bandwidth and latency limits follow Zoom bandwidth guidance and Zoom call statistics guidance. Gaming latency bands use EA guidance. Intermediate 1–5 cutoffs and conservative latency caps are our estimates. This test measures a Cloudflare route, not a game or video-service server, and does not measure packet loss, Wi-Fi reliability, device performance, or sustained playback.