For a meaningful VPN speed comparison, keep your device, network, and test targets consistent before comparing routes. Peak download speed only tells you how fast data transferred that time. It doesn’t show whether pages load promptly, video plays without interruptions, or work calls stay stable. Real-world speed is performance you can reproduce with the apps you use, at the times you usually use them.

Establish a baseline for your local network

Cross-border traffic travels through a chain of devices and networks: your device, Wi-Fi or Ethernet, home router, local ISP, VPN route, and destination service. Congestion at any point can affect the result. If you skip local checks and switch nodes right away, Wi-Fi interference may look like a route problem.

Pause cloud sync, system updates, and other bandwidth-heavy tasks on your devices, then test the local network with the proxy turned off. Use Ethernet if possible. If you’re on Wi-Fi, keep your device in the same place and note the signal strength. Don’t switch networks during testing, or you’ll be comparing different conditions.

  • ✅ Keep the device, client version, and connection type consistent. Plug in the device and turn off power-saving limits.
  • ✅ Pause background downloads and check that the router isn’t applying speed limits or traffic controls.
  • ✅ Record local speed test results with the proxy off as a baseline for your internet connection.
  • ✅ Update the subscription, then select a route manually so automatic selection won’t switch routes during testing.
  • ❌ Don’t rank peak results taken on different devices, test servers, or at different times.

A good local speed test only indicates that your local connection has no obvious bottleneck at that moment. It doesn’t mean the international route is equally clear. Conversely, if local results fluctuate significantly, check the device and router first rather than repeatedly testing and using up subscription traffic.

How to choose speed test tools and metrics

Browser-based speed tests are useful for measuring throughput, client probes can quickly screen nodes, and real-app tests show how things feel in everyday use. They measure different things, so results may not match. A tool’s automatically selected server may also change with your exit region. For route comparisons, choose the test target manually and keep it fixed.

Metric or tool What it can tell you What it can’t tell you on its own
Browser throughput test Download and upload performance to a specific test server That other websites will have the same speeds
Client URL probe Response time when accessing a specific address through a node That the probe result is the pure network round-trip latency for the entire route
Ongoing latency and packet loss monitoring Fluctuations, timeouts, and response consistency along a specific path That lost probe packets mean application data will be lost at the same rate
Real file upload or download Transfer performance for a specific service, account, and route combination That the subscription service must be the bottleneck
Video and work apps Startup, buffering, interaction, and sustained-use experience That smooth playback from a cached stream reflects first-time access speed

Latency measures how long a response takes; jitter measures how much that response time varies; packet loss tracks whether probes receive replies consistently. Download speed matters for receiving large files, while upload speed affects cloud storage, calls, and attachments. For remote desktops, steady responses are often more important than occasional download speed peaks.

Keep the units when recording results. Speed test tools often use Mbps, while file managers commonly show MB/s; don’t compare just the numbers. Also distinguish idle latency from latency under load: if interaction slows during a large transfer, your local upload may be saturated or queues may be building up.

Make sure test traffic actually uses the selected route

In split-tunneling mode, rules may send a speed test site over a direct connection. Even if the client says it’s connected, the browser test may be measuring your local network. Check the client’s connection logs first to see which rule matched the test domain and which outbound route it used.

To compare routes, you can temporarily switch to a global proxy policy if you understand what traffic it affects, then restore your usual split-tunneling settings afterward. A system proxy and a global policy aren’t the same thing: the system proxy is how apps connect to a proxy, while the global policy determines how traffic that reaches the client is routed. Apps that ignore the system proxy may still need the client’s TUN mode to handle their traffic.

Browsers on Windows and macOS usually follow the system proxy, but command-line tools and some apps may need separate configuration. On mobile, coverage depends on the client, system permissions, and per-app rules. Don’t rely on the status bar icon alone; app connection logs and checking the exit region are more useful.

Check DNS separately, too. If DNS queries take a different path from app connections, that can affect resolution and apparent location—but a resolver outside the exit region alone isn’t proof of a leak. Check for conflicts between the client’s DNS policy, encrypted DNS in your browser, and system settings. Don’t disable protections just to get a better-looking latency result.

Repeat speed tests under consistent conditions at your usual times

Test at the times you actually use the network, such as during work and busy evenings. There’s no need to seek out the quietest hours to prove a route is fast, or to treat a one-off outage as typical performance. Repeat tests at similar times and keep a record of unusual results; that’s more useful than picking the best-looking screenshot.

  1. Note the setup. Record the date, time, connection type, device, client version, proxy mode, and node name. Don’t record the full subscription URL.
  2. Verify the route. Select a node manually, check the connection logs, and confirm that the test domain uses the expected outbound route.
  3. Keep the target fixed. Use the same test tool and server, keep settings such as the number of connections consistent, and don’t run other speed tests at the same time.
  4. Watch the whole test. Record download and upload speeds, latency changes, pauses, and timeouts—not just the final peak.
  5. Retest with another route. Keep all other conditions the same. If results differ significantly, switch the test order and check again.
  6. Test your everyday apps. Open the sites, video services, or work apps you normally use and see whether the speed test results hold up.

Multi-connection tests can saturate a link, while a single file transfer may not reach the same speeds. Both are useful, but record them separately. Cached pages, files, and video segments can make performance look better than it is. When retesting, consider browser and server caches, and whether the resources themselves have changed.

How to interpret route and protocol differences

A direct route usually means the client connects straight to an overseas entry point. A relayed route passes through an intermediate entry point before forwarding traffic to the exit. IEPL describes a dedicated transport segment within a route. These terms describe how a route is organized; they don’t mean the whole path has no shared resources or guarantee speeds for any particular destination.

Shadowsocks, VMess, Trojan, and VLESS use different protocols from Hysteria2 and TUIC. The latter two are based on QUIC and depend on UDP conditions. Different networks may handle UDP differently, and results also depend on device performance, client implementation, congestion control, and server configuration. Protocol names aren’t a speed ranking.

If you change the node, protocol, entry, and exit at the same time, all you can conclude is that “this configuration works better in the current environment.” You can’t attribute the difference to the protocol alone. To compare routes, keep the protocol and target as consistent as possible; to compare protocols, keep the route and other settings similar. Note any variables you can’t control.

Low latency in a node list doesn’t necessarily mean a destination service will respond faster. Traffic still travels from the route’s exit to the destination website, which may also assign different content delivery servers based on the exit region. For route details, see the route list. The nodes currently available to you are shown in the client’s subscription list.

Turn test results into actionable conclusions

If every node is slow, recheck your local baseline, background tasks, and whether the client is handling traffic as expected. If only one website is slow, test other destinations and check for regional restrictions, account issues, or server load on that site. If the same node fluctuates only during your usual busy evening hours, keep consecutive results from similar times instead of relying on a single speed test screenshot.

If a speed test is fast but video stutters, check whether the test and video domains use the same outbound route, then look at buffering, player diagnostics, and the content delivery path. If downloads are fast but calls keep dropping, focus on upload speed, latency variation, and the route used by the app. A browser download test alone can’t rule out call quality problems.

Choosing a route: Favor routes that deliver repeatable results with the apps you use at your usual times. For large file transfers, compare sustained throughput; for calls and remote access, focus on consistent response times. Speed comparisons between routes are only useful when the test conditions match.

When submitting troubleshooting details, include the time, connection type, client version, node name, destination domain, and redacted error messages. Don’t share subscription URLs, account credentials, or full request headers. If you still can’t pinpoint the issue, continue with the troubleshooting guide or submit your records through a support ticket in the dashboard. Clear test conditions are more useful for diagnosing a problem than an isolated speed figure.