Use this guide when Shadowrocket is connected but webpages load slowly, videos buffer, or downloads are unstable. It assumes you already have your own provider and subscription. Start by checking your local network, then use Global Routing to isolate rule-related issues, and finally review protocol settings, connection quality, and server load. Change one condition at a time so a temporary improvement isn't mistaken for a fix.
1. Distinguish Latency, Throughput, and Connection Failures
The latency shown in Home is not the same as download speed. Latency measures how long a probe takes; actual downloads are also affected by packet loss, bandwidth, congestion, server-side limits, and the destination site's response. A low latency reading for a server only means that the specific probe was fast at that moment—it does not guarantee faster sustained transfers. Results from different testing methods shouldn't be ranked side by side.
First, describe exactly what “slow” means: does a link take a long time to respond, or does the page appear right away but images load slowly? Are all apps affected, or just one site? Is it always slow, or only in the evening? Requests in Shadowrocket pass through the following stages, and delays at any point can make a webpage seem unresponsive.
Keep a simple log of the network type, time, selected server, Global Routing mode, test destination, time to first response, and download speed. Use consistent units: 1 MB/s = 8 Mbps. MB/s from a download manager and Mbps used for bandwidth are not the same unit. Repeat the test three times to observe variation; three tests alone cannot establish long-term connection quality.
2. Local Network: Check Your Direct-Connection Baseline First
Pause cloud backups, system updates, and other large transfers. On the same device, access a resource that normally works over a direct connection and whose usual performance you know. Don't use a destination that requires a proxy to determine whether your direct connection is working; a failure to load it wouldn't help diagnose the issue.
- Test with Shadowrocket disconnected. Go to Home, turn off the connection, and confirm that the system VPN is disconnected. If it reconnects automatically, check Settings → On Demand, temporarily disable the setting that triggers the connection, and note its original state.
- Keep the test destination the same. On the same Wi-Fi network, reopen the same page or download the same file. Note whether it still lags. Don't compare a webpage in one round with video playback quality in another.
- Switch networks. If available, repeat the test over cellular. On a Mac or another desktop device, you can also compare with a wired connection. Testing uses data, so check your plan first.
- Reconnect and test again. Keep the same server and destination, and switch only between Wi-Fi and another network. See whether the issue follows the network you're using.
| What you observe | Check first | Next step |
|---|---|---|
| Still slow while disconnected | Wi-Fi signal, router congestion, local bandwidth | Move closer to the router and pause high-bandwidth activity on the same network |
| Wi-Fi is slow, but cellular works normally | The Wi-Fi connection and its route to the server | Retest with the same server; don't change protocol settings yet |
| Direct connection works, but multiple destinations are slow when connected | Proxy route, server, or protocol settings | Compare results using Global Routing |
| Only one destination is slow | Destination service, matched rule, and exit route | Compare with a similar test destination |
Public Wi-Fi may require you to sign in before you can use it. Being connected to Wi-Fi doesn't necessarily mean you have internet access. Complete the network login first, then start Shadowrocket. Switching servers repeatedly won't fix a local network that hasn't been authenticated.
3. Global Routing: Isolate Rules from Proxy Traffic
Check the current mode in Home → Global Routing. Config applies the rules in the active configuration. Proxy sends requests normally controlled by proxy routing through the selected proxy. Direct provides a direct-connection comparison. Scene follows its scene settings, so don't treat it as a fixed Proxy or Direct mode while troubleshooting. System traffic, local network exclusions, and other settings may also affect routing; the mode name alone doesn't tell you how a specific request was handled.
- Note the original mode, active configuration, and selected server so you can restore them when troubleshooting is done.
- Choose one destination, switch to Proxy, and start a new request using the same server. Stop any existing download and start it again so an old connection doesn't continue on its previous route.
- Switch to Config and test again. If Proxy works but Config doesn't, check rule order, the selected member of any policy group, and the policy actually matched by the destination domain.
- Use Direct for comparison only with a destination that normally works directly. Restore your usual configuration after the test; don't leave a test setting in place as a permanent fix.
This minimal rule snippet illustrates rule order; it isn't a complete file to replace your existing configuration with. PROXY must refer to a valid proxy policy or policy group in your configuration. If your configuration uses a different policy name, check its actual definition.
[Rule]
DOMAIN-SUFFIX,example.com,PROXY
IP-CIDR,192.168.0.0/16,DIRECT,no-resolve
GEOIP,CN,DIRECT
FINAL,PROXY
DOMAIN-SUFFIX matches a domain and its subdomains, IP-CIDR matches an IP range, GEOIP matches based on an IP database, and FINAL catches requests that weren't matched earlier. Rules are generally evaluated in order, so a broad rule placed too early may change the intended route. The private ranges in this example are only a subset; they aren't a complete LAN configuration.
If Proxy works, check which policy Config matches
Keep the current server, return to Config, and check the rule and policy group member actually used for the destination request. Changing protocols repeatedly at this point would muddle rule issues with connection issues.
4. Protocol Settings: A Working Connection Doesn't Guarantee Every Transfer Works
Settings for Shadowsocks, VMess, VLESS, Trojan, Hysteria2, and WireGuard are not interchangeable. No protocol has a consistently faster ranking independent of network conditions. Get subscription links and connection settings from your own provider. If your provider manages the settings through a subscription, first check whether they've changed the port or transport settings, then update your existing subscription. Don't assemble configuration fields based on guesswork.
| What to check | Fields that must match | Common symptoms |
|---|---|---|
| Server connection | Address, Port, authentication details | Can't connect, repeated reconnections, or authentication failure |
| TLS transport | Server Name, certificate, and device time required by the server | Handshake failure or noticeably slow connection setup |
| WebSocket transport | Path, Host, and TLS settings specified by the server | Basic port is reachable, but proxy requests fail |
| Hysteria2 | UDP reachability, authentication, bandwidth settings provided by the server | Works on one network but times out on another |
| WireGuard | Key pairing, Endpoint, Address, AllowedIPs, MTU | Handshake or routing problems; some large transfers stall |
For example, if your provider specifies port 443, that doesn't mean TCP 443 and UDP 443 are interchangeable. Some networks allow TCP but restrict UDP. In that case, a UDP-based connection may fail while another connection using TCP works. This reflects differences in the network and transport; it doesn't mean one protocol is always faster.
A larger MTU isn't always better. Consider testing a different value, following your provider's advice, only when small requests work but large files or some pages stall and other causes have been ruled out. Record the original value, change one setting at a time, and retest. Don't change MTU, DNS, and protocol settings together. If TLS validation fails, check the system time and Server Name first; disabling certificate verification is not a speed optimization.
Error: connection refused
Cause and fix: The connection was actively refused, possibly because nothing is listening on the port or a policy rejected it. Check Address and Port, then ask your provider whether that endpoint is operating normally.
Error: i/o timeout
Cause and fix: The operation didn't finish within the allowed time. This message alone can't distinguish packet loss from an unreachable destination or a busy server. Keep the server fixed, switch networks, and record when the error occurred and which connection stage was in progress.
Error: TLS handshake timeout
Cause and fix: The TLS handshake didn't complete in time, possibly due to packet loss along the route or a slow server response. Check the device time and TLS settings, then compare using another network. A timeout doesn't necessarily mean the certificate is invalid.
5. Connection Quality and Server Load: Narrow It Down with Cross-Tests
Connection quality describes the route from your device to the server and then to the destination service. Server load depends on the server's resource usage and scheduling at that moment. It's difficult to distinguish the two from a single client-side speed test. Cross-tests are more useful: keep the server fixed and change networks, then keep the network fixed and try another server from your existing subscription, using the same test destination.
- Keep the server fixed and change networks: If performance remains poor on just one Wi-Fi network, first consider the route from that network to the server rather than assuming the server is overloaded.
- Keep the network fixed and change servers: If just one server remains slow while other servers you already have work normally, check that server's entry and exit routes and server status.
- Keep conditions fixed and change the time: Compare different times under similar conditions. If things repeatedly slow down in the evening, note the pattern; provider-side information is still needed to distinguish congestion from server load.
- Keep the server fixed and change destinations: If just one file source or website is slow, consider destination-side speed limits, peering on the exit route, and content delivery paths.
A sudden, highly variable increase in latency may indicate queuing or an unstable route, but it isn't direct evidence of server load. ICMP probes may also be rate-limited or ignored by intermediate devices. If you can use system network tools, consider probe packet loss alongside real traffic; don't assume a device is faulty just because an intermediate hop doesn't respond.
Share repeatable comparison results with your provider
Summarize results from “same server, different networks” and “same network, different servers,” including the time, time zone, destination, and exact error text. This is more useful for distinguishing route issues from server issues than a single screenshot showing a low speed.
Remove sensitive information before sharing test resultsDon't share full subscription links, passwords, keys, or authentication identifiers publicly. Include only the time and relevant symptoms in your report. Send sensitive details such as server addresses through the private support channel specified by your provider.
6. FAQs and Checks to Restore Your Settings
Why are downloads slow even when latency is low?
A brief probe doesn't measure sustained throughput. Keep the server and download destination fixed, observe the transfer for a while, and note whether the speed drops or stalls. Then compare using another network. Don't run multiple speed tests at once; they'll compete for bandwidth.
Why is Proxy fast but Config slow, and what should I change?
In Config, check which configuration is active, the rule for the destination domain, and the policy group member actually selected. The two modes may use different servers. Use the same outbound server in both modes first, then decide whether the DOMAIN-SUFFIX or FINAL policy needs adjustment.
Will changing DNS improve download speeds?
DNS mainly affects name resolution and can indirectly change content delivery, but it can't fix every transfer bottleneck. If the delay happens only before a page opens, check DNS-related logs first. If the connection is established but remains slow, keep investigating the route and server instead of repeatedly changing DNS.
It's still slow after updating my subscription. Should I delete and reinstall the app?
First confirm that Home is using an updated server, and check Global Routing and Config. Updating a subscription won't fix Wi-Fi congestion or speed limits at a destination website. Keep your configuration and diagnostic notes, and complete the cross-tests before deciding what to do next so you don't lose useful evidence.
When you're done troubleshooting, restore your original Global Routing mode and, if needed, Settings → On Demand. Review any settings changed temporarily and remove rules added just for testing. If the cause is still unclear, provide reproducible steps: network type, time, which existing server, destination, mode, and the exact error text. “Connected but slow” isn't enough detail.
- Confirm the local network baseline is normal and no large transfers are running in parallel.
- Use the same destination before and after testing, with consistent units and measurement methods.
- Identify the actual outbound route behind the difference between Proxy and Config; don't rely on the mode names alone.
- Confirm protocol settings came from your provider and revert any unsuccessful test changes.
- Save a redacted test log to compare if the same issue occurs again.
Get Shadowrocket and the Getting Started Guide
Get Shadowrocket for iPhone and iPad on the App Store for a one-time payment of $2.99. The developer is Shadow Launch Technology Limited; check the App Store listing for system requirements. The Mac (Apple Silicon and Intel), Windows, Android, and Linux versions are free. The app price does not include a service plan; get a subscription from your own provider.