At a Glance

This guide is for readers who have imported their own server details or subscription and need to choose among existing entries. Use latency tests to rule out entries that repeatedly time out, then keep the network and routing mode consistent while comparing how the same page loads. Finally, retest on both Wi-Fi and cellular and keep entries that suit your everyday needs, plus a backup. This site does not provide, sell, or recommend servers or subscriptions.

What Latency in Milliseconds Tells You

The Connectivity Test reading is the time taken for a probe, not a download speed. What it measures depends on the probe method: a TCP connection test only confirms that the test could connect to the specified address and port at that moment; it does not verify proxy authentication. An HTTP or HTTPS request sent through the proxy may also be affected by the proxy handshake, destination lookup, TLS handshake, and test server response. Avoid ranking results from different probe methods or test targets together.

1,000 ms
1 second of elapsed time, not a speed
443
The default HTTPS port, not necessarily the proxy server port
2 legs
Proxy traffic involves at least two legs: device to proxy, then proxy to destination

For example, a one-off 40 ms probe may succeed while the page still stalls as images load. Pages often make multiple requests, sometimes to different domains. A probe neither captures the time for every request nor shows how much data the connection can sustain. A timeout means the probe did not finish within the allotted time; it is not 0 ms, and a single timeout does not prove the server is down. If results vary from fast to slow across repeated tests, note the variability and number of failures rather than recording only the fastest result.

Test in Home and verify real-world loading

The steps below use the iPhone / iPad interface. Choose two or three candidates from your imported list and keep the same device, network, and test window. If a large download or system sync is using the network, wait until it finishes before testing.

  1. Keep test conditions consistent

    Open Home, check Global Routing, and note the current mode. If you use Settings → On Demand, stay on the same network during testing to avoid connection changes triggered by switching between Wi-Fi and cellular.

  2. Run a connectivity test

    In Home, tap Connectivity Test. When the results appear, record each candidate's time or timeout. Repeat twice using the same method and note both successful readings and failures. If you change the test method or target, take new samples for every entry instead of mixing them with the old results.

  3. Switch between entries one at a time

    In Home, select a candidate and make sure the connection toggle is on. Once the connection is stable, open the same page you use every day. Before switching to the next candidate, confirm which entry is selected. Keep Global Routing unchanged and make sure the test request actually uses the selected proxy; otherwise, you may be measuring a direct connection or a different policy.

  4. Check that the page finishes loading

    Record when the page becomes usable, whether all images load, and whether a refresh produces an error. Alternate the order—A then B, and B then A—when opening the same page, keeping cache conditions as consistent as possible. If the second load is noticeably faster, consider caching before attributing the improvement to the selected entry.

  5. Keep a usable backup

    Prefer entries that load the entire page reliably across multiple tests and keep long-lived connections stable. Repeat the comparison on another network you use regularly. If the results differ, keep separate choices for each network; one entry does not have to rank first in every situation.

Choosing a Location: Compare the Actual Route on Your Network

With other conditions equal, a shorter distance can reduce propagation time. But a location label in the list does not reveal the actual route between carriers, evening congestion, or the path from the proxy to the destination site. Test both nearby and farther-away locations, then compare them on the same page. If a nearby location has a faster probe but often stalls partway through loading, while a farther one is slightly slower but consistently finishes, choose based on the real-world result.

Retest the same entries on each network

Regular Wi-Fi
  • Keep the device in the same place and open the same page using the same entries.
  • If problems happen mostly in the evening, repeat the test then and note whether failures cluster at a particular time.
  • If several entries slow down at once, use a page that normally connects directly on that network as a baseline to check whether the local network is also fluctuating.
Cellular
  • Make sure the device is off Wi-Fi, then repeat the steps in a location with a stable signal.
  • Test while stationary first. Then, if it matters for your use, check whether the connection drops while moving. This keeps network handoffs from skewing the same set of readings.
  • A brief page load is enough to check for interruptions; there is no need to repeatedly test with large downloads.

Record Wi-Fi and cellular results separately. Use location to narrow the candidates, then choose based on repeated failures and real-world loading.

Protocol Types: Check How Each Entry Actually Transmits Data

A protocol name alone cannot predict speed. How Shadowsocks, VMess, VLESS, and Trojan perform depends on the entry's transport, server processing, and current route; they should not all be assumed to use TCP. Hysteria2 is based on QUIC, while WireGuard uses UDP. If the current network restricts UDP, these entries may work on one network but time out during the handshake on another. Conversely, UDP availability does not mean every entry using it will be faster.

When comparing protocols, use existing entries with similar service conditions and the same location where possible. If the location, server, and protocol all change, you can only conclude that one entry performed better in that test—not that the protocol caused the difference. Don't change ports or transport settings just to make entries look alike; the client entry must match the server configuration.

Example: Recording Success Rates

Entry A
42 ms, timeout, 46 ms
Entry B
81 ms、84 ms、79 ms
What to do
Open the same page with each entry and check that both the first screen and images load
Retest order
Retest B first, then check whether A's timeout happens again

These figures illustrate how to record results; they are not measured performance or pass/fail thresholds. A succeeded twice and was faster, but it also failed once—don't leave that out.

Troubleshooting UDP Differences

Check the protocol
Confirm whether the existing entry uses Hysteria2 or WireGuard
Keep the settings unchanged
Check the address and port against the existing server details; don't change them temporarily
Retest across networks
Compare the same entry's connection and loading on Wi-Fi and cellular
Know what the results can tell you
Failure on only one network points to a difference in the route or network conditions

Results across networks alone cannot prove that UDP is restricted. Also check for address resolution failures, server availability, and mismatched entry settings.

Low Latency but Pages Won't Load? Check Global Routing

If a candidate tests quickly but a page still won't load, first check whether the page request is using the entry you're comparing. A Connectivity Test result does not show which route a website takes under your everyday configuration. If you need to temporarily change modes, note the original setting first and test only pages where changing the route is appropriate. Restore the setting when you're done.

Config

Recommended

Rules are matched against the current configuration. A request may use DIRECT, PROXY, a specified policy group, or be blocked by REJECT. The route selected by a rule may not use the entry currently selected in Home.

Use this to check whether pages load completely under your everyday configuration.

Proxy

Route traffic handled by the proxy through the proxy mode, without relying on per-rule routing in Config. If the same page works here but fails after switching back to Config, check the rules and policy groups before changing servers based only on latency.

Use this for a brief comparison between Config and the proxy route.

Direct

Send traffic without using a proxy server. Check whether a page that normally works directly on your current network is accessible as a baseline. A destination that requires a proxy failing here does not indicate a problem with your local network.

Use this to check your local network and establish a baseline with pages that connect directly.

If only one page fails, check whether its main domain and static asset domains match different rules. For example, DOMAIN-SUFFIX covers the specified domain suffix, but an image hosted on another domain may use a different policy. If none of the earlier rules in the configuration match, check where FINAL points. If every candidate still fails after switching to Proxy, while the direct-connection baseline works, check whether the address and port in the subscription are still valid, whether authentication succeeds, and whether the problem occurs only on a particular network. A lower millisecond reading cannot replace these checks.