Connection Lab.
Speed & loaded latency.
Measure HTTP transfer speed and response timing to this site. See what changes during downloads and uploads.
Every move has a return trip.
Compare baseline response timing with timing under load. A rise can help explain lag during downloads or uploads. This does not test your game server.
Each bar is an observed HTTP response. Height uses a relative scale; dots show recent probes, not individual network packets.
- 1Baseline HTTP response
- 2Download + HTTP response
- 3Upload + HTTP response
- 4Post-transfer HTTP response
How this loaded response test works
Connection Lab is a free, user-initiated browser test. It measures HTTP throughput and response timing between your device and Internet Analysis. Use it to compare the same connection before and during transfers, then repeat after changing one condition.
- Establish a baseline. Eight small HTTP probes run before transfers. At least two successful responses are required.
- Measure during downloads and uploads. Small probes run alongside each bounded transfer. Results show the mean response time in each window.
- Check after the transfer. Fourteen additional probes measure response timing once the transfers finish.
- Compare, then investigate. Repeat with the same device and endpoint. Test Wi-Fi versus Ethernet or a quieter time of day.
Read the numbers correctly
- Throughput (Mbps)
- Completed payload bytes × 8 ÷ elapsed seconds ÷ 1,000,000. HTTP and server overhead affect the result.
- Loaded response change (ms)
- Mean HTTP response during a transfer minus the baseline mean. A negative difference is possible.
- Response variation (ms)
- Mean absolute difference between consecutive successful response timings; not RTP jitter.
- Unavailable
- Fewer than two successful probes in a window. Fast connections may reach the byte cap before enough probes complete.
Connection Lab questions
What does Connection Lab measure?
Connection Lab measures HTTP download and upload throughput to this site, baseline HTTP response time, response variation, and response time during transfers. It compares the loaded and baseline means. It does not test your game server or streaming service.
Is loaded latency the same as bufferbloat?
No. A rise in response time during transfers is an observation, not a diagnosis. Queues, Wi-Fi contention, browser scheduling and server load can affect HTTP timing. This bounded test does not prove that your access link was saturated or that bufferbloat caused a delay.
How do I compare Wi-Fi and Ethernet?
Run a test and label its connection. Switch the same device to Ethernet or Wi-Fi, select the new label, and run again. Your two successful results appear side by side while the page stays open. Repeat the comparison with other traffic paused. Labels are not detected automatically.
Does this test measure packet loss?
No. A failed HTTP request is not a network packet-loss measurement. The results show successful and failed HTTP probe counts separately. Unavailable values mean there were too few successful probes to calculate that result.
How much data does it use?
The transfer test targets approximately 64 MiB downloaded and 64 MiB uploaded at most, plus protocol overhead and small HTTP probes. Each transfer phase stops at its byte cap or after approximately five seconds, subject to request timeouts. The overall test stops after 60 seconds. Avoid it on a limited data plan unless you accept that usage.
Are my results public?
Connection Lab keeps its results in the current page session. It does not submit them to public measurement reports. Reloading or leaving the page clears this comparison. Requests still reach the test server and remain subject to the site privacy policy.
Choose your next check
Connection Lab focuses on within-test timing and a two-run comparison. For a different method or a deeper investigation, use the relevant tool.
Method references: Internet Analysis methodology and MDN: browser timing with performance.now(). See our privacy policy for request handling.