If Shadowrocket connects but runs slowly, keep the test target fixed and check your local network, protocol settings, route, and server load in that order. This guide is for users with their own provider and subscription who can connect but experience slow page loads, inconsistent downloads, or video buffering. Comparing results can help distinguish local network issues, incorrect connection settings, routing mismatches, and congestion on the provider’s side—without changing everything at once.
1. Distinguish latency, throughput, and connection failures
Latency is the time a probe takes to make a round trip. Throughput is the amount of data transferred over a period of time. The latency test in Home reflects reachability and response time for the selected test method; it is not a direct measure of download speed. Different methods may test different parts of the connection, so make sure you use the same method before comparing results.
If a page is slow to load the first time but images load normally afterward, check DNS, connection setup, and the TLS handshake first. If pages load quickly but large files download slowly, look at bandwidth, packet loss, congestion, and server load. If only one site is slow, the issue may be with the destination or the route from the server’s exit to that site.
- Choose a website you can access normally and a fixed-size download file you’re authorized to access.
- Record the network type, server name, Global Routing mode, test method, and time of day.
- Repeat each test three times. Record page load time, sustained download speed, and failures; don’t keep only the fastest result.
- Pause system updates, cloud sync, and other bandwidth-heavy tasks. Change only one condition at a time during testing.
The number of repetitions suggested here is a testing procedure, not a performance claim. Download tools report MB/s, while advertised connection speeds use Mbps; 1 MB/s is about 8 Mbps. Actual throughput is also affected by protocol overhead.
2. Local network: establish a baseline with direct access and another network
First, check whether your network connection is working normally. Choose a destination that is normally reachable directly, temporarily set Global Routing to Direct, and test page loading and downloads. Then restore the original mode and test again. Direct routes traffic outside the proxy, but it does not disconnect the system VPN tunnel. Also run a comparison after disconnecting Shadowrocket.
- Disconnect in Home and test a website you know is reachable directly. If it’s still slow, check your Wi-Fi signal, router, and ISP connection first.
- Move closer to the router and test again. If performance improves nearby but drops farther away, address Wi-Fi coverage before changing protocol settings.
- If your plan allows, switch between Wi-Fi and cellular while keeping the destination and server the same. Cellular data charges may apply.
- Check whether another VPN, system proxy, or network filter is active. Temporarily use only the network path needed for this test, then restore your original settings afterward.
If both direct and proxied traffic are slow on the same Wi-Fi but work normally on cellular, the evidence points more toward the local network. If direct access works but the same server is slow on both networks, continue checking the protocol and the route on the provider’s side. These comparisons help narrow down the cause; one successful or failed test is not enough to determine responsibility.
3. Protocol settings: confirm the connection works before comparing speed
A server appearing in the Home list doesn’t mean its settings are complete or that the server is still active. Get subscription links from your provider. To update a subscription, use the update option on its entry in Home, then confirm which server is selected. Updates may overwrite manual changes, so save the original settings first and keep temporary tests out of your everyday configuration.
| Protocol | Check first | Common misdiagnosis |
|---|---|---|
| Shadowsocks | Server, port, encryption method, password, and any additional settings required by your provider | Mistaking a mismatched password or encryption method for a slow route |
| VMess / VLESS | UUID, transport type, and the related TLS, SNI, Host, and Path settings | Copying only the address and port while missing transport settings |
| Trojan | Password, port, TLS, and server name | Mistaking a failed TLS handshake for insufficient download bandwidth |
| Hysteria2 | Authentication, TLS, server port, and UDP reachability on the current network | Assuming a QUIC-based protocol works on every network |
| WireGuard | Keys, peer address, tunnel addresses, and routing ranges | Confirming the tunnel is up without checking whether destination traffic goes through it |
When checking the type in Home → Add Server, use the complete settings supplied by your provider. Don’t change VMess directly to VLESS, or change a server’s port to 443 just because HTTPS commonly uses that port. The protocol, authentication, and transport must match the server. For TLS errors, check the system time, server name, and certificate status; don’t hide the problem by disabling certificate verification.
Error: connection refused
Cause: The destination host or an intermediary actively refused the connection, often because nothing is listening on that port or access is denied.
Fix: Check the server and port provided by your provider, then ask them to check the listener and access restrictions. Repeated speed tests won’t fix this.
Error: i/o timeout
Cause: A network operation did not finish in time. Possible causes include an unreachable destination, packet loss, filtering, or no response from the server.
Fix: Keep the server the same and switch networks, then compare it with another server you already have. Record which combinations time out.
These are error messages that may appear in network logs; the exact message depends on where the failure occurred. A timeout alone doesn’t prove the server is overloaded, and a refused connection doesn’t prove your local bandwidth is insufficient. Resolve reproducible connection errors before assessing sustained download speed.
4. Rules and DNS: verify the route your traffic actually takes
Turning on the connection doesn’t mean every request goes through the selected server. In Global Routing, Proxy sends traffic through the proxy, Direct bypasses it, Config follows the active configuration, and Scene uses the configured scene policy. During troubleshooting, note the current mode; don’t assume where traffic goes based only on the VPN indicator.
In Config, rules are generally evaluated in order, and the matching rule’s policy is applied. The example below demonstrates syntax and ordering only; it isn’t a complete configuration to use as-is. 192.0.2.0/24 is a documentation example subnet.
[Rule]
DOMAIN-SUFFIX,example.com,DIRECT
IP-CIDR,192.0.2.0/24,DIRECT,no-resolve
GEOIP,CN,DIRECT
FINAL,PROXY
DOMAIN-SUFFIXmatches domain suffixes. Check whether an earlier rule is sending the destination to an unintended policy.IP-CIDRmatches an IP range. In this example,no-resolvemeans the rule won’t trigger an additional domain lookup.GEOIPrelies on IP geolocation data. A country match does not guarantee route quality.FINALhandles traffic that didn’t match earlier rules. Place it appropriately as a fallback and check that its policy is the one you intend.
If the same destination works in Proxy but not in Config, check the available request records for the destination, matched rule, and final policy, then return to Config and adjust the specific rule. Proxy is a temporary comparison to narrow down the issue; it doesn’t mean all requests should stay proxied.
If only the first visit takes a long time, check which DNS servers the current configuration uses and how DNS requests are routed. Don’t change DNS, servers, and protocol all at once; even if performance improves, you won’t know which change helped. A first request after clearing the cache differs from one using cached data, so record those results separately.
5. Routes and server load: compare across times and servers
Route quality depends on the path from your ISP to the server’s entry point, internal forwarding within the server, and the route from its exit to the destination. Server load depends on processing capacity and shared bandwidth. The client usually can’t distinguish these causes from a single latency reading; compare networks, times of day, and servers you already have.
| Repeatable pattern | Likely cause | Next step |
|---|---|---|
| The same server is slow on only one network | The network connection or the route from that ISP | Keep the server and retest the same destination on another network |
| Several existing servers slow down at the same time of day | Congestion on a shared link, or a common local bottleneck | Also test a direct-connection baseline at the same time |
| Only one server is slow; other existing servers work normally | That server, its dedicated route, or its exit path | Send your provider the time period and comparison results |
| Websites load normally, but one download source is slow | A speed limit at the destination or the route from the server’s exit to it | Compare with another download source you’re authorized to access |
Bottom line: identify the failing combination, then report it to your provider
If only one existing server remains slow on the same network and destination, keep your other settings unchanged and test that server again at a different time. Don’t rebuild the entire configuration based on one slow test.
Low latency and slow downloads can occur at the same time. Small probes may complete normally while sustained transfers are affected by congestion, packet loss, bandwidth allocation, or destination-side limits. Hysteria2 uses QUIC, but the current network may restrict UDP. TCP-based transports can also suffer substantial throughput drops when packets are lost. No protocol is consistently fastest on every network; compare connections your provider offers with correctly configured settings.
- Give your provider the test time and time zone, network type, server identifier, destination type, and steps to reproduce the issue.
- Include both a working and a failing combination—for example, “the same server is slow on Wi-Fi but works normally on cellular.” This is more useful than simply reporting “slow speeds.”
- Before sharing logs, remove subscription links, authentication details, UUIDs, keys, and personal browsing records. Keep the error text and only the necessary time information.
6. FAQs: automatic connections, subscriptions, and speed tests
Why does video keep buffering when latency is low?
First, use the same server to reach another destination and determine whether the issue affects one site or overall throughput. Then check the active Global Routing policy. A quick response to a small probe doesn’t guarantee enough bandwidth for continuous video, and the probe and video traffic may take different routes to their destinations.
Why does it reconnect as soon as I disconnect, preventing a direct-connection test?
Go to Settings → On Demand and check the automatic connection conditions. Note the original settings, temporarily disable the relevant triggers, and run the disconnection comparison. Restore the settings afterward. On Demand controls when a connection starts; it doesn’t increase server bandwidth.
It’s still slow after updating the subscription. Should I delete all the servers?
First confirm the subscription updated successfully, then make sure the selected entry in Home is the updated server. A subscription update only retrieves configuration published by your provider; it won’t resolve congestion. Don’t delete everything—keep a known-working configuration for comparison.
Can I increase concurrent connections and turn up all the other settings?
Keep the default settings first. Concurrent tests compete for bandwidth, and excessive load may make your device heat up. Change a setting only if your provider specifically recommends it. Record the original value and retest against the same destination. Changing several things at once makes it harder to identify the cause.
7. Restore your settings and save your results
Once you’ve identified a likely cause, there’s no need to keep changing other layers. Restore any temporary changes to Global Routing, On Demand, and network settings, then test the destinations you use every day. If the issue happens only at certain times, keep those results and avoid comparing them directly with tests from other time periods.
- Direct access is also slow: Check Wi-Fi, your router, background traffic, or your ISP connection first.
- Connection errors: Check the protocol, authentication, port, and transport settings, and confirm service status with your provider.
- Proxy works but Config doesn’t: Check the matched rule, policy name, and DNS route.
- A specific server or time period is slow: Organize your comparisons and ask your provider to check the route and server load.
A one-time client purchase doesn’t include a network plan. Getting the client won’t change your provider’s bandwidth, routes, or server load. See this site’s download page and Getting Started guide for Shadowrocket download options and setup steps. For everyday troubleshooting, keep a repeatable baseline.