Key takeaways
Record a no-VPN baseline first
Testing only after the tunnel is enabled removes your reference point. Before connecting, open several ordinary sites, check the application you care about, and record approximate latency. On mobile data, keep the device in the same location during comparison when possible.
The baseline is not about achieving a maximum benchmark. It tells you which symptoms already existed before the VPN and which appeared after the route changed.
Confirm that the route actually changed
After connecting, verify the public IP and repeat a small set of independent requests. Check DNS next. If the public IP changes but one application still behaves as if it were outside the tunnel, inspect split-tunnel and bypass rules.
One failed website is not enough to declare the VPN broken because the remote service can fail independently. Use several unrelated checks.
Measure latency, loss, and throughput as separate properties
Latency, packet loss, and bandwidth answer different questions. High throughput does not compensate for unstable loss in voice traffic, and low latency does not guarantee enough capacity for video. Repeat measurements and use a range instead of selecting one best result.
INFOCROSS project measurements show 60-70 Mbps for VLESS and 70-80 Mbps for AmneziaWG 3.1. Average server latency is 150-400 ms depending on region and operator.
Test reconnect and network handover
Briefly interrupt connectivity, restore it, and observe whether the tunnel returns without rebuilding the profile. On a phone, switch between Wi-Fi and mobile data as a separate test. A stable client should either recover predictably or expose a clear state that requires reconnection.
A failure that appears only after handover is a distinct diagnostic problem and should not be mixed with throughput results.
Keep a reproducible record
Record the date, connection type, VPN profile, several measurements, and the applications actually tested. Avoid treating one carrier name as a permanent property of the result because routing can change. The record is valuable as a baseline for your own future comparisons after a configuration or network change.
Continue reading