In Shadowrocket, seeing Connected only means the connection switch is on; it does not guarantee fast access to every destination. This guide assumes you already have your own service provider and subscription. Establish a baseline with the same destination, then check your local network, protocol settings, route, and node in turn. Finally, use rules and logs to verify the results. Change one variable at a time to distinguish configuration errors from network congestion or remote speed limits.
1. Distinguish latency, throughput, and actual routing
“Slow” can mean several things: a long wait for a webpage to open, a low sustained download speed, or intermittent pauses during video or calls. These point to connection setup time, sustained throughput, and jitter or packet loss, respectively. The latency test on Home does not replace an actual download test. A test returning a result also does not prove that the full route to a website is working normally.
First, go to Home → Global Routing and confirm the current mode. Config routes traffic according to the active configuration and matching rules; Proxy sends traffic handled by the tunnel through the proxy; Direct bypasses the proxy; Scene selects behavior according to configured scene conditions. During troubleshooting, don't leave Scene unverified: switching networks could also change the egress path.
- Choose one test destination you’re authorized to access. Record page load time, sustained transfer speed, and any interruptions.
- Keep the node and network the same. Observe for about 30 seconds and repeat three times. Don’t compare results from different websites directly.
- Note the units. A download speed of 5 MB/s is about 40 Mbps. This is a unit conversion example, not a speed guarantee.
- Temporarily use Proxy to rule out differences in rule-based routing. When finished, restore your usual Config or other mode.
2. Local network: rule out Wi-Fi issues and background traffic first
When local access is unstable, switching remote nodes may only change the symptoms temporarily. Pause cloud sync, system updates, and large file transfers, then retest the same destination. Wi-Fi signal bars show only part of the wireless picture; they say nothing conclusive about router upload capacity, your ISP’s upstream connection, or interference on the same channel.
- Direct baseline: Temporarily select Global Routing → Direct and visit a commonly accessible website. If it’s still slow, check your local connection first. This test alone can’t determine whether a destination that normally uses a proxy is reachable.
- Switch networks: Keep the node unchanged and test once on Wi-Fi and once on cellular. Cellular uses mobile data, so check your plan allowance first.
- Isolate Wi-Fi variables: Move closer to the router and test available 5 GHz and 2.4 GHz networks separately. Don’t change the Wi-Fi band and node in the same test.
- Check for overlapping network controls: Make sure no other VPN or network filter is changing the route at the same time. On managed devices, follow your organization’s policies and don’t remove management settings without authorization.
If Wi-Fi is slow but cellular works normally, the issue is related to the local network or its route to the server; that alone doesn’t prove the router is faulty. Try another Wi-Fi network. If multiple devices show the same problem on the original network, check router load and broadband service first.
3. Protocol settings: check the server requirements before considering overhead
Return to Home and open the current node’s edit screen. Check each field against the original parameters from your service provider. The protocol, server address, port, credentials, transport, and TLS settings must match one another. Port 443 is commonly used for TLS, but the port number alone doesn’t prove that a service uses TLS or guarantee a faster connection.
| Protocol or setting | Check first | What it can—and can’t—tell you about speed |
|---|---|---|
| Shadowsocks | Encryption method, password, port, and any plugin parameters required by the server | Mismatched settings usually prevent a connection. Latency alone doesn’t reveal encryption overhead. |
| VMess / VLESS | UUID, transport, TLS, Host, and path | VLESS does not provide payload encryption by itself. Use the secure transport specified by the server; don’t interchange fields. |
| Trojan | Password, TLS, SNI, and certificate name | If the handshake fails, correct the domain and certificate settings. Don’t disable certificate verification just to get a connection. |
| Hysteria2 | Authentication, TLS, UDP reachability, and bandwidth-related parameters | Based on QUIC; may not work on networks that restrict UDP. Follow the server’s instructions for bandwidth parameters. |
| WireGuard | Key, Endpoint, and Allowed IPs | Usually relies on UDP. Establishing the tunnel does not mean that routing or DNS is correct for every destination. |
If small pages load but large downloads or some websites keep stalling, consider MTU and path compatibility. Don’t apply an MTU value found online to every node. Only adjust it separately if the current protocol and interface offer the setting and you can preserve the original value; follow your provider’s guidance and retest.
Bottom line: change only the protocol when comparing protocols
Compare protocols only when the server offers equivalent endpoints for each. If you also change the region, node, and protocol, you can’t attribute the speed difference to the protocol alone.
4. Route: use time-of-day and network comparisons to spot congestion
A route is the path traffic takes between your local network and a remote service. A node is the remote endpoint that runs the proxy service. They are not the same thing. Nodes with different names may still use the same ISP interconnection path, and a region in a node’s name does not prove its actual route or available bandwidth.
| Comparison | What you observe | Next step |
|---|---|---|
| Same node, different networks | Slow on Wi-Fi, normal on cellular | Check your local broadband and its route to the remote endpoint before changing protocol fields. |
| Same network, different times | Repeated slowdowns in the evening, with normal speeds at other times | Record when it happens and ask your service provider about route load. |
| Same node, different destinations | Only one website or download source is slow | Check the destination website, the route from the remote egress to it, and content delivery differences. |
| Same network, different existing routes | Only one route remains slow | Keep your comparison results and ask your service provider to investigate that route. |
Low latency does not mean high throughput. A connection may handle a small request quickly when idle, yet slow down during sustained transfers because of congestion, retransmissions, or speed limits. If you use system network diagnostics to check packet loss, note that intermediate devices may limit probe responses. A single hop failing to respond while the final destination works does not prove that the hop is dropping forwarded packets.
5. Nodes: distinguish server load, speed limits, and connection errors
Once you’ve mostly ruled out the local network and protocol settings, check the status of the relevant node in your existing subscription, your account’s data allowance, and any service limits. Keep a copy of the current settings, then update the subscription and check whether the server address or port changed. Updating a subscription refreshes configuration; it won’t fix local wireless interference or increase your account’s available bandwidth.
The examples below are underlying error messages that may appear in connection logs. The exact wording varies by protocol and operating system. Use the stage at which an error occurs to guide diagnosis; not every timeout means the node is congested.
Error: connection refused
Cause: The destination port has no matching service, or the connection was actively refused.
Fix: Check the server address and port, then update your existing subscription. If the problem persists, ask your service provider to confirm that the service is listening.
Error: i/o timeout
Cause: A network read or write did not complete in time. The route may be blocked, packets may be lost, or the server may not be responding.
Fix: Keep the node unchanged and retest on a different network, then compare it with another node you already have. Don’t change authentication settings without evidence.
Error: TLS handshake timeout
Cause: The TLS handshake did not complete in time. The route may be having issues, or the connection may be going to a server port that doesn’t match the settings.
Fix: Check the TLS, SNI, and port combination, confirm the device’s clock is correct, and leave certificate verification enabled.
- If all destinations consistently top out at about the same speed, check for bandwidth limits on your account or server, and rule out a limit on your local broadband connection.
- If speeds drop only during concurrent downloads, reduce the number of simultaneous connections and see whether single-connection speed and errors improve.
- If one node remains problematic while other nodes in the same subscription work normally, give your service provider the test time, network type, and error details.
- Before sharing logs, redact subscription tokens, passwords, UUIDs, keys, and personal address information. Never post a complete subscription URL publicly.
6. Restore Config and check whether rules or DNS changed the route
If things work normally in Proxy but remain slow in Config, check the rules before switching nodes again. Open Config and confirm which configuration file is active, then check the policy matching the destination. Rules generally match from top to bottom, stopping at the first match. An overly broad Direct rule or a different proxy policy may send the same website along another route.
DOMAIN-SUFFIX,example.com,PROXY
IP-CIDR,192.168.0.0/16,DIRECT
GEOIP,CN,DIRECT
FINAL,PROXY
This rule excerpt illustrates syntax and order; it is not a complete configuration to replace yours with. DOMAIN-SUFFIX routes example.com and its subdomains through PROXY; IP-CIDR routes the specified private IP range through DIRECT; GEOIP matches based on the IP database; FINAL handles connections not matched by earlier rules. Your configuration may also use custom policy groups, so check the selection within those groups too.
- Restore your usual Config mode and start a new connection to the same destination. Don’t reuse an existing connection, which can affect the comparison.
- Check the connection log for the destination, matched rule, and actual policy. Confirm that traffic wasn’t unexpectedly sent Direct or routed through a different node.
- If the first visit is slow but later visits are faster, check DNS lookup time and caching. Don’t replace all DNS settings based on this symptom alone.
- Keep a copy of the original configuration. Change only one rule or DNS setting at a time, then repeat the same three tests.
7. Common questions and troubleshooting notes
Why are downloads still slow when latency is only a few dozen milliseconds?
Latency measures the round-trip time for a small request; it doesn’t indicate sustained transfer capacity. Keep the node and destination unchanged, record download performance for about 30 seconds, then compare on a different network. Also check whether the speed is shown in Mbps or MB/s.
Will switching to Hysteria2 always make it faster?
No. First confirm that your server supports Hysteria2 and that your current network allows UDP, then test with the parameters provided. If the handshake cannot complete, check reachability before changing bandwidth parameters.
Will updating my subscription restore speed?
Refreshing the configuration may help if an old address or port is no longer valid. After updating, check which node is selected so you don’t mistake the effect of switching nodes for a speed improvement caused by the update.
How can I test Direct if the connection reconnects automatically?
First, note your original Settings → On Demand settings. Temporarily turn off On Demand and run the test. When finished, restore the setting and confirm that Global Routing is back in its original mode.
Your troubleshooting notes should include the network type, test time and time zone, node identifier, protocol, Global Routing mode, destination, all three results, and any error details. If you can’t identify the cause, keeping these notes is more useful than changing several switches at once. This site does not provide, sell, or recommend nodes or subscriptions. Ask your own service provider about server-side issues.