For an accurate VPN speed test, use the same device and test server to compare your direct connection with your VPN connection, then test again during the hours you actually use it. Comparing speed test tools and methods isn't about finding the one that produces the biggest number; it's about choosing a test that answers the right question.
Speed test tools compared: browser tests, continuous monitoring and real-world app tests
For slow downloads, start with a browser-based test that lets you choose the server. For choppy calls, check ongoing latency and the app's own stats. Different tools can produce different results on the same device without either being wrong: they may connect to different destinations or use different numbers of parallel connections.
| Method | What it measures | How to compare results |
|---|---|---|
| Speedtest browser test | Download and upload performance to a selected test server | Manually select the same server and keep the test mode consistent |
| Cloudflare speed test | Latency differences under idle and loaded conditions | Check the test endpoint details. If the entry point changes, treat results as a guide to user experience, not a fixed-endpoint comparison |
| System ping | Round-trip latency variation and missed responses when the destination responds to probes | Probe the same destination continuously and first confirm whether the probe traffic goes through the tunnel |
| In-app stats | Call latency and packet loss, playback buffering, and sustained download speed | Repeat the same task in the same app and note when stuttering occurs |
A concurrent speed test adds up the throughput of multiple connections. A single file download is also affected by server-side limits, connection behavior and routing, so the two results aren't directly comparable. System ping isn't an application performance test either: ICMP probes may be throttled or filtered, or take a different route from app traffic. A destination that doesn't respond to ping isn't necessarily dropping packets from video or web traffic.
Side-by-side testing: keep the device, network and test endpoint consistent
A direct-connection baseline is the performance of your local network with the VPN disconnected—not the speed listed on your broadband plan. Run the baseline as close as possible to the VPN test. Comparing a direct test in the morning with a VPN test in the evening isn't meaningful: household traffic and load on your provider's routes may have changed.
- Keep the device and connection type consistent. Use Ethernet if possible. If you have to use Wi-Fi, keep the device in the same place and connect to the same access point. Pause cloud sync, system updates and large file transfers, and note whether you're testing over Ethernet or Wi-Fi.
- Disconnect the VPN and choose a test server manually. Save the server name, location and test mode, and record download, upload and latency results. Run several tests and keep the lower results too—don't just pick the best-looking screenshot.
- Connect the VPN and verify the test traffic. Check the client connection log to make sure the speed test is using the selected node, not going direct because of a split-tunneling rule. Turn off automatic server selection in the test page, then rerun the test with the same server and test mode.
- Alternate tests, changing one variable at a time. When comparing nodes, don't also switch protocols or connection types. Run real-world tasks after speed tests so the test itself doesn't saturate the connection and distort a call or streaming test.
How to interpret latency, bandwidth and stability
Latency is the wait between a request and its response; round-trip latency is usually measured in milliseconds. It affects page loads, remote control and conversations on calls. If response times are good while idle but slow as soon as a download starts, compare latency under load. This is especially important when uploads saturate the connection, as interactive requests may get stuck in a queue.
Download and upload speeds describe how quickly data moved during the test, not the advertised capacity of your connection. Browser tests usually show Mbps, while download apps may show MB/s, so check the units before comparing. For video and downloads, focus on whether download speed stays high enough; for file sharing or screen sharing, check upload speed too.
Jitter and packet loss can explain why a connection feels fast most of the time but occasionally stutters. Jitter is variation in latency; packet loss means packets never arrive. An average can hide brief spikes, so check whether spikes or loss coincide with app issues. If only ping reports a problem while app stats look normal, try a responsive test destination before concluding the route is faulty.
Which metrics matter? For downloads, focus on sustained throughput. For calls and remote control, prioritize latency under load, variation and packet loss reported by the app. A high peak on a speed test page can't replace checking an actual call or an uninterrupted stream.
How to test during the evening peak
Run a test when your home network is relatively idle, then repeat it during the hours you normally have calls, watch video or share the connection with your household. At each time, first test with as little background traffic as possible, then resume normal household use and check how apps perform. The first test helps compare routes; the second shows whether calls stay smooth while, for example, someone is backing up photos.
Record the date and time, connection type, node, test server, tool and what you experienced in the app. To summarize repeat tests under the same conditions, use the median as a typical result and note slowdowns, spikes and interruptions separately. If performance drops occasionally, test again on another day at a similar time. If the issue persists, compare Ethernet with Wi-Fi, different nodes and different destinations. A single result describes one route at one moment; it isn't a promise of all-day performance.
Speed tests look fine, but apps stutter: find the bottleneck
- ✅ Both direct and VPN connections are slow, but both improve over Ethernet: check Wi-Fi signal, interference or access point load first.
- ✅ The same test endpoint works normally without the VPN, but one node's route repeatedly slows down: keep the protocol and test time consistent, try another node, and investigate that route first.
- ✅ Performance is fine while idle, but calls stutter as soon as an upload starts: pause background uploads and check latency under load to see if local connection queueing is involved.
- ✅ Browser speed tests look normal, but one app stutters: check whether it uses the same node, then review its in-app stats. A test server and an app service in the same region may still take different routes.
If every node is slow, there's no need to keep switching to exits in the same region. Start with a direct-connection baseline on the same device, then see if another device shows the same issue. If only one device is affected, check background tasks and client routing rules. For further troubleshooting, see the network troubleshooting guide. Before switching exits, check the server list, then retest under the same conditions.