网络延迟测试工具
使用真实HTTP请求测量您的网络延迟和基本下载速度。
关于此工具
向公共CDN发送5个类似ping的请求，测量往返时间，给出最小、最大和平均延迟。还会获取较大文件以估算下载速度，并在可用时显示浏览器报告的连接信息。注意：这测量的是到外部服务器的延迟，而非您ISP的额定速度。
使用方法
- 1 第一步：点击「开始测试」开始延迟测量。
- 2 第二步：等待5个ping请求完成，查看最小/最大/平均结果。
- 3 第三步：点击「测试下载速度」估算您的下载吞吐量。
- 4 第四步：查看连接质量评级和浏览器连接信息。
What this tool actually measures
Despite the name "speed test," the most reliable number this tool gives you is latency — how long a tiny request takes to travel to a server and come back. It works by sending five real HTTP requests to a public CDN (specifically a small file hosted on unpkg.com) and timing each round trip with the browser's high-resolution performance.now() clock. Because these are genuine network requests to an external server, this tool does not run offline — disconnect your internet and every ping returns an error. That is the honest trade-off: to measure a network you have to use the network.
It is important to understand what the result is and isn't. The latency you see is the time to reach a particular CDN endpoint over the public internet, not a controlled measurement of your ISP's rated line speed. A dedicated service like a full broadband speed test spins up multiple parallel streams to a nearby server it controls; this tool makes single sequential requests to a shared third-party host. Treat the output as a quick, directional health check, not a contractual benchmark.
How the latency test works, step by step
When you click Start Latency Test, the tool runs five requests in sequence, pausing about 300 ms between them so they don't pile up. Each request carries a unique cache-busting query string (?_=timestamp_index) and is sent with cache: 'no-store', which forces the browser to actually hit the network instead of returning a cached copy that would report a misleadingly fast "0 ms." The requests use mode: 'no-cors', so the browser doesn't need permission from the server to complete the round trip — the tool only needs to know when the response arrived, not what it contained.
After all five complete, it reports three figures: the minimum (your best-case latency), the average, and the maximum. The spread between min and max is often more revealing than any single number — a tight cluster means a stable connection, while a wide gap signals jitter.
A worked example: reading your numbers
Suppose your five pings come back as 18, 22, 19, 64, and 21 milliseconds. The tool shows Min 18, Average 29, Max 64, and labels the quality based on the average. Its thresholds are concrete:
| Average latency | Rating | What it feels like |
|---|---|---|
| Under 20 ms | Excellent | Competitive gaming, video calls feel instant |
| 20–49 ms | Good | Smooth for almost everything |
| 50–99 ms | Fair | Noticeable lag in fast-paced games |
| 100 ms and up | Poor | Calls stutter, web pages feel sluggish |
In the example above the average is "Good," but that single 64 ms spike is the real story: it's jitter, the inconsistency that makes voice calls warble and game characters teleport. Run the test a few times — if the spike reappears, something on your path is intermittently congested.
The download estimate and browser connection info
The Test Download Speed button fetches the same CDN file once, measures how many bytes arrived and how long it took, and converts that to megabits per second with the formula bytes × 8 ÷ seconds ÷ 1,000,000. Because the file is small and it's a single stream over a shared host, this is explicitly a rough estimate — it will usually understate a fast connection, since a small download finishes before your link reaches full speed. For a true throughput figure, use a multi-stream broadband tester.
If your browser supports the Network Information API, the tool also displays its built-in connection hints: effectiveType (such as 4g), an estimated downlink in Mbps, the browser's own round-trip-time guess, and whether data-saver mode is on. These come straight from the browser's own estimates and appear only where the API is available — notably absent in Safari and Firefox.
Practical tips and common mistakes
- Test on Wi-Fi and cable separately. High latency on Wi-Fi but low on Ethernet points at your local wireless, not your ISP. Move closer to the router and retest.
- Close bandwidth-hungry tabs first. A running video stream or large upload inflates your latency reading. Quiet the network before measuring.
- Run it more than once. A single bad ping can be a fluke. Three runs reveal whether a high number is the norm or an outlier.
- Don't compare this to your plan's advertised speed. The download figure here is a single-stream estimate to a public CDN; mismatches with your "1 Gbps" plan are expected and not a sign something is broken.
- Geography matters. Latency partly reflects physical distance to the CDN's nearest edge server. Someone far from a data centre will see higher baselines no matter how good their connection is.
Used the right way — as a fast, in-browser pulse check on responsiveness and stability — this tool answers the question that actually matters day to day: not "how big is my pipe," but "is my connection quick and consistent right now?"