Internet Speed Test

About this tool

This page measures the speed between your device and the server that hosts this website — the same server that delivered the page you are reading — and reports download throughput, latency and jitter. That single fact frames every number shown: a speed test measures one path, not your connection in the abstract, and a path to one host is a sample, not a verdict on your whole internet service. A test hosted inside your ISP's own network, or one that picks the fastest of many servers, will often show a different figure. Within those limits the result is genuinely useful. Run it a few times on the same connection and you build a baseline: is tonight slower than last week, is the guesthouse wifi worse than your phone's mobile data, did anything change after the ISP's visit. Most internet use in Cambodia happens on mobile connections — 25.3 million cellular connections were active in early 2025, around 143 percent of the population — so a quick test that runs in a phone browser matters as much as a fibre benchmark. What it cannot do is prove your ISP owes you a contractual speed, or settle which of two providers is faster. Those questions need a multi-server test repeated on a wired connection.

How it works

Press Start and two phases run in your browser, both against assets hosted on this site (speed-test-client.tsx). First, five small requests fetch a 1 KB file one after another; latency is reported as the median round trip in milliseconds, and jitter as the standard deviation of the five — how far individual pings spread around their middle. Three payloads of growing size — 1, 5 and 10 MB of incompressible random bytes generated at build time (scripts/generate-speed-test-assets.mjs) — are then downloaded in sequence. The first download acts as a warm-up and is discarded; the reported Mbps combines the 5 and 10 MB transfers as total megabits over total time (lib/speed-test/calc.ts). Each request carries a unique query string and a no-store directive so the browser asks the network rather than its own cache; the site's CDN may still serve the bytes from a nearby edge location, which mirrors how the site itself reaches you. Cancelling aborts the in-flight fetch. No upload figure is shown — a static site has no endpoint that accepts your data.

Use cases

• Building a baseline for your own line: run the same test on the same device each week, and a slow evening becomes visible instead of a feeling.
• Comparing a guesthouse or cafe wifi against your phone’s mobile data before committing to a long video call or a multi-gigabyte download.
• Getting a quick yes-or-no after an ISP visit, a new router or a plan upgrade — did the numbers actually move?
• Separating a slow connection from a slow website: if this test is fast but one site crawls, the problem is probably theirs.
• Checking how evening congestion hits a home fibre or mobile plan by repeating the test at different hours.
• Comparing what you get in a provincial town against your connection in Phnom Penh before remote work or study.
• Testing before an online class, exam or interview so a bad connection is found before the call starts, not during it.

Frequently asked questions

• Why is my result lower than the speed my ISP advertises?

  A plan’s number is a maximum to the provider’s network, not a promise for every destination. This test measures one path — your device to this site’s host — and along the way wifi distance and interference, an older router or phone, other people on the same line, and evening congestion all shave the figure. The endpoint matters too: a test hosted inside your ISP’s own network shows the best case, not the average. Test on a wired connection, at a few different hours, before drawing conclusions.

• Can this result prove my ISP owes me a faster connection?

  No — and any single-endpoint test has the same limit. A contractual complaint needs repeated tests against several independent servers, on a wired connection, ideally while nothing else uses the line. What this page gives you is a signal and a habit: dated results that show whether your everyday experience is stable or degrading. If several days look clearly wrong, that pattern — not one number — is what is worth taking to the provider.

• Why does this number differ from Speedtest by Ookla or another service?

  Because each test picks different servers. Ookla’s app tries many servers and reports the strongest — its H1 2025 report put Cambodia’s median mobile download near 60 Mbps, a best-path figure by design. This page uses one fixed endpoint: the host serving this site, behind its CDN. Neither is wrong; they answer different questions. Compare like with like — same tool, same network — before treating two numbers as a contradiction.

• Why is there no upload speed?

  A static website has no endpoint that can receive your data, so there is nothing to upload against — this tool measures download and latency only. Upload matters for video calls and sending files; Zoom, for example, works from a few megabits per second in each direction. If you need an upload figure, a multi-server test such as Ookla’s measures it — treat this page as the download-and-responsiveness check.

• Why do two runs a minute apart give different numbers?

  Because the measurement is real. The first run pays cold-start costs — connection setup shows up as a slow first ping and a slower first transfer — and traffic on your line and the network changes second to second. When this page was written, three back-to-back runs on one connection returned 30, then 252, then 239 Mbps. Repeat the test and read the middle of several results, not a single number.

• What is the difference between Mbps and MB/s?

  Mbps is megabits per second; MB/s is megabytes per second. There are 8 bits in a byte, so 100 Mbps equals about 12.5 MB/s. If a file downloads at 12.5 MB/s on a 100 Mbps plan, those are the same speed — the units, not the connection, are the usual source of confusion in speed tests. This page reports Mbps, the unit ISPs advertise.

• What do latency and jitter actually affect?

  Downloads care about bandwidth; live conversations and games care about latency and jitter. Latency is how long one small request takes to be answered — under about 50 ms feels instant on a call. Jitter is how much that time varies, and high jitter is what makes voices stutter and games stutter-step even when bandwidth looks fine. A line can test fast on Mbps and still feel bad if jitter is high.

• Does the test send my results anywhere?

  No. The figures are computed in your browser and shown on your own screen — no request carries the Mbps, latency or jitter numbers off your device. The page does record anonymous usage events: that a test was started, cancelled, completed or failed, and failures carry only the error text. Your measured results are never part of those events.

• What download speed do I need for streaming or calls?

  Rough guides from the services themselves: Netflix recommends 3 Mbps for HD, 5 Mbps for full-HD and 15 Mbps for 4K; Zoom calls work from a few megabits per second each way. Casual social video sits in the same range. A whole household sharing one line wants more headroom — but remember this test shows what reaches this site today, so leave margin above the number you actually need.

Data and accuracy

Content last reviewed 19 October 2026