TCPing checks whether the server’s TCP port accepts a connection. A proxy latency test checks whether a request can reach a specific website through the proxy and get a response. A download test measures how much data that route can transfer over time. If the latency shown in your server list doesn’t match your experience, first use proxy latency tests to rule out servers that fail repeatedly, then retest for your intended use, such as browsing or downloads.
TCPing: Tests the server endpoint, not proxy egress
TCPing tries to connect to the address and TCP port configured for the server. It primarily measures how long it takes to establish a TCP connection. It is not an ICMP probe like the system ping command. If the server address is a hostname, it must also be resolved before connecting; whether DNS lookup time is included depends on how the client measures elapsed time.
- A result is shown: The probe connected to the TCP port at that address. This does not confirm that VMess or VLESS authentication succeeded, or that the transport, security layer, or server-side forwarding is configured correctly.
- The endpoint uses a relay: If the server address points to an access layer, TCPing reaches that layer. It does not measure the connection from the access layer to the backend or from the backend to the destination website.
- No result or a timeout: First check whether the server uses TCP transport. TCPing cannot determine whether an endpoint that only supports UDP is working. For TCP servers, verify the address and port in your subscription, then cross-check with a proxy latency test.
TCPing is useful for checking whether a server endpoint is reachable, but not for ranking every server overall. Two servers can have similar connection times to their endpoints and completely different routes from their exits to the same website.
Proxy latency: Measures response time to a specific website
A proxy latency test sends a real request to the test website through the proxy core. The diagram below shows a typical route through proxy egress. Unlike TCPing, it also checks the proxy transport, protocol exchange, and access to the destination—but it does not evaluate every website.
If the test opens a new connection, the result may include connection setup, the handshake, and time waiting for the website to respond. Connection reuse, DNS caching, and the points where timing starts and stops can also affect the result. Many latency tests only need a small HTTP response; they do not download the full page and do not represent the total time a browser needs to load images, scripts, and resources from other domains.
Low TCPing, but proxy latency test times out
- Confirmed
- TCP connection to the server endpoint
- Not confirmed
- Proxy handshake and access to the destination website
- First
- Test a known-working server with the same test website
- Then
- If only the original server fails, check the core logs for handshake or egress errors
A reachable port doesn’t mean the configuration is correct. Don’t change the port based on TCPing alone.
Proxy latency test succeeds, but common websites fail
- Confirmed
- The test website responded this time
- Not confirmed
- The proxy endpoint used by the app and its route to that website
- First
- Select the server and open a commonly used website in the app
- Then
- Check whether the app uses the system proxy or TUN, then review routing rules for the destination domain
A successful built-in client test does not prove that your browser or other apps are using the proxy.
Download tests: Measure throughput, not page load time
A download test transfers data from a test resource through the proxy to your device, then divides the amount transferred by the elapsed time. The result depends on the test server, your local network, the server’s egress, and how long the transfer runs. With a very small file, connection setup and the initial transfer can account for much of the test time. For sustained throughput, the speed later in a long download is more representative.
These figures only illustrate unit conversion; they are not test results. When comparing client and router readings, distinguish uppercase B (bytes) from lowercase b (bits). If the interface uses MiB/s, convert using binary units rather than comparing it directly with MB/s. Download speed also does not indicate upload speed; test uploads separately with the app you actually use.
How to choose: Rule out repeated failures, then compare by use case
Prioritize servers that pass proxy latency tests consistently. For browsing, check response times and page loads on websites you use regularly. For large downloads, test each server using the same resource and compare sustained speeds. Don’t replace a stable connection just to chase a one-off peak.
v2rayN and v2rayNG: How to run comparable tests
In v2rayN, select a server in the list, right-click, and choose the test named “Tcping,” “Real connection latency,” or “Speed.” Options such as “Multiple selection” may vary by version. In v2rayNG, open the menu at the top right of the configuration list and choose “Test real connection for all configurations” to check whether proxy requests succeed. These latency results are not download speeds.
- Keep test conditions consistent.Compare on the same device and network, and pause system updates, cloud sync, and other downloads. Use the same test website for latency checks and the same resource for downloads; changing the test target changes the route being measured.
- Record failures before comparing times.Run the proxy latency test three times for each candidate and note whether each attempt succeeds or times out. If all three succeed, sort the times and use the middle result as a reference. If any attempt times out, track the failures separately instead of letting the fastest successful result hide the problem.
- Test downloads one server at a time.Narrow down the candidates first, then test the selected servers in turn. Simultaneous downloads compete for the same access bandwidth and can lower each server’s measured speed. If two servers are close, alternate tests to avoid mistaking network fluctuations at different times for a difference between routes.
- Finish by checking the apps you actually use.Switch to a candidate server, then open a familiar website or download service. Note the time and connection type. If the problem only occurs in the evening, retest around the same time under the same conditions rather than drawing conclusions from a single daytime result.
V2Ray Server Speed Test FAQs
Why do the same subscription servers test so differently on a PC and an Android device?
First make sure both devices have the same configuration selected; don’t rely on server labels alone. Connect both devices to the same Wi-Fi network and test the same website at around the same time. Even under similar network conditions, clients may use different cores, probes, and timing methods. Focus first on whether each device can succeed consistently.
A proxy latency test works, but download tests always show zero. What should I check first?
On the same device, use a server known to download successfully to test the same resource. Make sure it actually returns a file rather than an error or login page. If that works, try downloading a small file through the original server and check the core logs for connection resets or read timeouts.
Speed tests work, but my browser can’t use the server. Should I reimport the configuration?
Don’t reimport it yet. If you use the v2rayN system proxy, check that “Set system proxy” in the tray menu is set to “Auto configure system proxy,” then make sure your browser follows the system proxy. If TUN is handling traffic, troubleshoot in that mode and check whether routing rules send the destination domain’s requests through a direct connection.
The same server is fast one time and times out the next. Should I delete it?
Test the original server and a known-good comparison server in turn against the same website. If both fail, check your local connection and the test website first. If only the original server fails repeatedly, retest it for your actual use case. If it remains unstable, remove it from your regular candidates; that makes it easier to revisit later than deleting it after a single timeout.