Updated July 2026 · 6 min read
Troubleshoot slow connections
A node can answer a latency test in 40 ms and still crawl at under a megabyte a second. Work out which kind of slow you have first, then decide whether to change lines or look at your own machine.
Low latency is not high throughput
The number on the test page is the round trip of a single handshake with a test address. It tells you whether the path is clear and how far it detours, and nothing at all about how much bandwidth the exit has or how many people share it.
So a 40 ms node doing 2 MB/s while a 180 ms one does 20 MB/s is completely normal. Pick download nodes by measured throughput and save the latency figure for gaming and calls, where the round trip is what you actually feel.
Two kinds of slow, two different fixes
In the first, pages snap open and video starts quickly, but large downloads crawl. The path is fine and the exit bandwidth is being shared out, so the answer is a different line, not another restart of the client.
In the second, everything hesitates, pages sit there without a first byte, and the latency test swings wildly or times out. That looks like packet loss along the path or trouble on your own network, which is what the next section is for.
Re-running speed tests during peak hours is the least useful thing you can do: each run measures the congestion of that one second, the numbers contradict each other, and you still cannot tell which line to pick.
Rule out your own end first
Glance at the Connections page and see whether some process is moving a lot of data. Cloud storage sync, system updates, torrent clients and game launchers downloading in the background are the usual suspects: they eat the bandwidth and the browser gets the blame.
Suspect the Wi-Fi as well. The 2.4 GHz band is crowded in apartment blocks, and switching to 5 GHz or plugging in a cable for one test settles the question. TUN mode adds a little overhead of its own, so testing on the plain system proxy with TUN off isolates that part.
Peak hours and a test worth comparing
From eight in the evening until midnight the shared bandwidth is fought over hardest, and none of that has anything to do with your settings. Most subscriptions carry several lines in the same region, and moving to another one beats testing the same node over and over.
To compare fairly, hold everything else still: same time window, same download source, same file, change only the node, run each one three times and take the middle result. A speed test site one time and a download manager the next gives you numbers that cannot be lined up.
What is the project?
Clash Verge Rev is a Tauri-based GUI for Mihomo with profile management, system proxy, TUN mode, rule editing and WebDAV backups.