TCP ping checks a node’s port, real connection latency measures a request sent through the proxy, and a download speed test measures throughput. For readers testing nodes or troubleshooting timeouts in v2rayN or v2rayNG: first identify which stage succeeds, then test the service you actually need, and compare latency or speed based on your use case.
What Ping, TCP Ping, and Real Connection Tests Measure
First, clarify what “ping” means. The system ping command measures ICMP echo round-trip time. A server or network device along the route may ignore ICMP while still accepting application connections normally. TCP ping in a client attempts to connect to the specified TCP port at the node address. Although both results are shown in milliseconds, they are different kinds of probes. The comparison below groups system ping and TCP ping as pre-connection reachability checks.
Compare by test scope
Ping and TCP Ping: Address and Port
Suppose the node is configured to use port 443. A successful TCP ping only means the client established a connection to that TCP port at the address. It does not check the TLS certificate, proxy authentication, transport path, or whether the server can reach the destination website. If the node uses a non-TCP transport, testing a TCP port with the same number says even less about whether that transport works.
Best for: An initial check that an address or TCP port is reachable.
Real Connection Latency: A Proxied Request
RecommendedThe client uses its core to send an HTTP or HTTPS request to a test address, making this more representative of real-world access than a port-only check. Results may be affected by connection reuse, DNS resolution, handshakes, routing, and the test site’s response. The timing endpoint and success criteria vary by client version, so the result should not be treated as the time it takes to load a complete webpage.
Best for: A first-pass check of nodes that can complete a proxied request.
Download Speed Test: Sustained Transfer
The client retrieves data over the test route and measures speed over a set interval. Results depend on the test server’s rate limits, connection count, test duration, and whether the client reports an average or peak. Download speed does not predict upload speed or replace latency measurements for short requests.
Best for: Comparing large-file download performance under the same test conditions.
Why Low Latency, Slow Downloads, and Timeouts Can Coexist
The figures below are hypothetical examples to illustrate how test results relate; they are not measurements from a specific route. Suppose a node uses TCP port 443, and the real connection probe reaches the proxy egress. During a 10-second download test, the client then receives 20 MB of data. Even when all three results come from the same node, they describe different stages.
- Fast port, slow request: The node may accept a TCP connection quickly, while the proxy handshake, the route from the server to the test site, or the test site’s response is slow. Don’t treat
280 − 42 = 238 msas kernel overhead: the two figures are not timings for separate stages of the same request. - Fast request, slow download: A small response transfers little data. Sustained downloads are more likely to expose congestion during peak hours, test server rate limits, or fluctuations in your local Wi-Fi. Check the units before comparing:
MB/smeans megabytes per second, whileMb/smeans megabits per second. - Fast test, but one site still times out: The speed test address and the actual destination may match different routing rules. Even when both use the proxy, the server may take different routes to the two destinations. Being able to download one test file does not mean another domain is reachable.
How to compare: Verify the destination first, then rank by use case
If both nodes can reach the service you need, compare the median latency and failure rate for short requests to websites and APIs. For large downloads, compare average speed over a fixed interval. Don’t reject a node that hasn’t had a request test based on one low TCP ping result.
Retesting Individual Nodes in v2rayN and v2rayNG
Test one node at a time, pause background downloads, and keep the network and test address the same. In the v2rayN 7.x desktop app, right-click a node to find “Test server latency (Tcping),” “Test server real connection latency,” and speed test options; menu labels may differ between builds. Note the exact version and test name, and don’t mistake an older result in the table for a test you just ran.
Retest on Both Platforms: Match the Target Before Comparing Results
v2rayN desktop
- Right-click the selected node to run TCP ping, then run the real connection latency test on that same node. Don’t run download speed tests on a whole group of nodes at once; they can compete for bandwidth.
- Switch to the node and use an app configured to send traffic through the client proxy to access the destination you actually need. If the test succeeds but the app fails, check whether the app uses the system proxy or a manual proxy, and which outbound route handles the destination.
v2rayNG on Android
- In the node list, open the “⋮” menu in the top-right corner and look for a real connection test. If there’s an option to test the selected configuration individually, use it to make the corresponding log entries easier to identify.
- When testing actual access, make sure the target app is included in the current per-app proxy settings. Don’t switch back and forth between Wi-Fi and mobile data during a retest. If you change networks, record a separate set of results.
For cross-platform comparisons, use the same node, test address, and test type. If the timing methods aren’t comparable, compare success or failure only—not small differences in milliseconds.
- Run the real connection test three times in a row and record each result. For example,
210 ms / 240 ms / timeoutmeans two successes and one failure. Don’t record a timeout as zero or report only the fastest result, 210 ms. - Use the same test file and duration for available nodes, and rotate the test order. If a tool reports only peak speed, label it as a peak and don’t compare it directly with another tool’s average.
- Finally, open the website or service you actually use and repeat an action that confirms the connection works. If the test site loads but the destination keeps failing, troubleshoot the actual destination separately instead of rerunning the speed test.
Timeouts and -1: Troubleshoot by Failure Stage
-1 is not negative latency. It usually means the client didn’t get a valid test result, not that a specific error code occurred. A timeout only means the test didn’t finish before the deadline. First identify which test failed, then use client logs from the same time to distinguish DNS resolution, connection attempts, handshakes, and request responses. Increasing the timeout can help determine whether a request is simply slow, but it won’t fix certificate or configuration errors.
Four Common Results
System ping fails, but the node still loads websites. Should I switch nodes?
Don’t switch nodes based on ping alone. Note the system ping target, then run TCP ping against the address and actual TCP port configured for the node, and finally access the real destination. If the latter two repeatedly succeed, a lack of ICMP replies is not evidence that the service is down.
TCP ping returns a result, but real connection tests keep showing -1. What should I check first?
Retest only this node and check the first relevant error in the logs. For certificate validation errors, check the device’s clock, the server certificate, and the server name in the node configuration. For transport handshake or authentication failures, verify the protocol parameters and transport settings against the configuration source. Don’t disable certificate validation just to get a latency reading.
All nodes suddenly time out. Does that mean every node in my subscription is down?
First rule out issues shared by all nodes: confirm your local network is working, the core can start, and the logs show no local port binding errors. Then select a previously working node and check whether the test domain resolves. If the client lets you change the test address, retest separately with another known-accessible HTTPS address. If only the old address fails, check the test site or its route first.
The real connection test passes, so why won’t my browser load the site?
Check whether the probe and browser request use the same outbound route. Search the logs by the time of the failure and the destination domain. If there’s no matching request, first check whether the browser is using the proxy. If a request appears but uses direct access or a different outbound route, review the routing rules. Even when both use the same outbound route, check the destination domain’s resolution and response separately.