Measure for 5 minutes whether your connection is truly stable. Reveals dropouts and periodic slowdowns that a short test misses.
A short speed test measures maximum bandwidth, but tells you nothing about how stable your connection is. PulseSpeed pings your line 300 times over 5 minutes and shows every variation live. This reveals micro-dropouts, jitter spikes and temporary slowdowns that cause stuttering video calls or laggy gaming.
An average ping says little; the spread around it says everything. The thresholds below are the values we apply when assessing a connection.
| Metric | Good | Acceptable | Problem | Why it matters |
|---|---|---|---|---|
| Range (best vs worst ping) | under 20 ms | 20 to 60 ms | over 60 ms | The key figure on this page. A wide range means your line changes behaviour markedly under varying conditions. |
| Jitter | under 5 ms | 5 to 15 ms | over 15 ms | Determines whether calls and games feel smooth; more important than average ping. |
| Packet loss | 0 percent | up to 0.5 percent | over 1 percent | Every lost packet is data that must be resent, or in calls is gone for good. |
| Spikes above 2x median | 0 to 2 | 3 to 10 | more than 10 | Isolated outliers are normal; repeated spikes indicate structural congestion or WiFi interference. |
Because the problems people actually experience rarely occur within 20 seconds. Micro-dropouts, a router revisiting its WiFi channel, a neighbour starting to stream, or congestion on the local loop are all events on a timescale of minutes. A short test measures your peak speed; this test measures whether you hold on to it. With 300 measurements at one per second, a single hiccup is clearly distinguishable from a pattern.
Do not look at average ping but at spread. A line averaging 30 ms that stays between 28 and 34 ms is excellent. A line with the same average that swings between 8 and 190 ms is poor, despite the identical average. As a rule of thumb: a range under 20 ms and jitter under 5 ms with no packet loss is a healthy connection.
A single spike is almost always harmless: a background process, a WiFi scan, or a brief queue somewhere along the path. Patterns are what matter. Regular spikes at a fixed interval often point to a device on your own network. Spikes coinciding with housemates' activity point to bufferbloat. Gradual degradation over the test points to congestion or thermal problems in equipment.
Between roughly 7 and 11 PM most people use the internet at once. On shared infrastructure — more so on cable than on fibre — that can measurably worsen latency and stability. Always measure at several times before drawing conclusions, and specifically keep the evening measurement if you want to support a complaint.
Yes, and of all the tests on this site this is the most persuasive, because a single snapshot is easy to dismiss and a series of 300 measurements is not. The condition is that you measure over an ethernet cable: with WiFi your ISP can legitimately argue the problem is inside your home. Export the result and repeat the measurement at different times of day.
No. The test sends one small probe per second and therefore uses a negligible amount of bandwidth — considerably less than an average web page. You can keep working during the test, though for a clean measurement it is better to avoid heavy downloads, unless you specifically want to see what those do to your stability.