Real download and upload measurements. Precision readings in brass, copper and a little neon.
Internet speedtest
The instrument room
1,000 Mbps range · engages automatically above 1,000 Mbps
Download
—
—Mbps
Upload
—
—Mbps
HTTP latency
—
—ms
IDLE HTTP LATENCY TRACE
Waiting for idle latency samples
Added delay
—
—ms
The pressure experiment
Bufferbloat check
Runs with your speedtest
Download: — · Upload: —
Waiting for idle and loaded samples.
Extra HTTP delay while your connection is busy. The instrument shows the worse of the download and upload median increases; lower is better. The balloon exaggerates small increases for visibility; the numbers are measured.
How to read this check
Each increase is the loaded median minus the idle median, floored at zero. At least 5 idle and 3 samples in each loaded phase are required for a final result. Low: under 30 ms; moderate: under 60 ms; high: under 200 ms; very high: 200 ms or more. These are our descriptive bands, not a Waveform grade.
This bounded speedtest may not fully saturate a fast link, so added delay can be underestimated. HTTP server time, Wi-Fi and other traffic can also affect it. For a dedicated comparison, try Waveform’s bufferbloat test.
Ready when you are. No speedtest traffic until you press Start.
Idle HTTP jitter
— ms
HTTP latency during download
— ms
HTTP latency during upload
— ms
Test network: Cloudflare. Starting a test sends data directly between your browser and Cloudflare and may use up to about 400 MB, plus protocol overhead. Cloudflare collects completed results for aggregated connection-quality insights. About Cloudflare's test. Avoid testing on a metered connection unless you are happy with that data use.
Real readings. Your connection.
The test bypasses Togethia's hosting connection. Upload uses four concurrent transfers. Results reflect your current route to Cloudflare, Wi-Fi, VPNs and other traffic. Different test servers and methods, including Speedtest.net, can give different speeds.
Let the instruments settle.
The needles update as samples finish. Overdrive selects a logarithmic 10 Gbps range for speed dials and flask fills. Latency and added-delay scales remain linear. Stop leaves partial readings; Test again starts fresh.
Look beyond the big numbers.
HTTP latency includes the request and server response time; it is not ICMP ping. Jitter is the variation between these same samples. Loaded latency shows how response time changes while downloading or uploading. A dash means a measurement is unavailable.