Starlink Speed and Latency, Explained

Two households with identical hardware, in the same country, can see very different Starlink performance. Understanding why means separating three layers that get lumped together as 'speed'.

Layer one: the physics floor

There is a minimum latency you cannot go below, set by the distance to the satellite and the speed of light. At 550 km straight overhead, the round trip to the satellite and back is about 3.7 ms. Add the hop down to a ground station and the terrestrial path onward and you get a floor of roughly 20 ms for a well-placed user near a gateway.

That floor rises when the satellite is low on your horizon rather than overhead, because the slant range is longer — a satellite at 25° elevation is well over twice as far away as one directly above you. It rises again if your traffic has to cross several satellites over laser links before reaching a ground station.

In practice Starlink users typically measure 25–60 ms. The spread within that range is mostly geometry and routing, and it changes minute to minute as the constellation moves.

• Satellite directly overhead at 550 km: about 3.7 ms round trip through space
• Satellite low on the horizon: several times that, from the longer slant path
• Typical measured Starlink ping: 25–60 ms including ground routing
• Geostationary comparison: 600 ms or more, permanently

Layer two: the shared-capacity ceiling

Starlink is a shared medium. The satellites over your region have a fixed total throughput, and everyone beneath them draws from the same pool. Your speed at any moment is roughly the available capacity divided by how many people are actively using it nearby.

This is why Starlink performance has a daily rhythm. Evening peak — when everyone in your region is streaming at once — is measurably slower than three in the morning, sometimes by a large factor. It is also why a sparsely populated rural area often outperforms a densely subscribed suburb on the same hardware.

Priority tiers sit on top of this. Business and priority plans get allocated ahead of residential traffic during congestion, and mobile/roam traffic is deprioritised behind fixed residential users. During an uncongested hour these tiers behave almost identically; during peak they diverge sharply, which is precisely what the price difference buys.

If your speed is fine at 2am and poor at 8pm, you are seeing congestion, not a fault. No amount of repositioning the dish will change it.

Layer three: your own installation

This is the layer you control, and it is where most fixable problems live. The dominant issue is obstructions. Starlink needs a wide, clear view of the sky — not a single sightline, but the whole arc satellites travel through. A tree branch clipping one edge of that arc will cause a dropout every single time a satellite passes behind it, producing brief, repeated stutters that look like an unreliable service rather than a physical blockage.

The Starlink app logs obstruction events. If you are seeing periodic one-to-two second losses, check that log before assuming anything about the network.

• Obstructions — trees, roof lines, chimneys, poles; check the app's obstruction map, not just the sky by eye
• Cable runs — damaged or sharply bent cable between dish and router degrades the link
• Router placement — a strong satellite link with weak Wi-Fi at the far end of the house measures as slow internet
• Snow and ice — the dish heater handles most of it, but heavy accumulation still attenuates
• Overheating — dishes in extreme heat can throttle; ventilation and mounting position matter

Why speed tests can mislead

A speed test measures one instant on one path. Because your connection hands over between satellites every few minutes, and because the route to a test server changes with those handovers, consecutive tests can differ substantially without anything being wrong.

Test several times across different hours before concluding anything, and pay attention to the shape of the results rather than the peak. A connection that measures 180 Mbps once and 40 Mbps four times is a 40 Mbps connection with a good moment, not a 180 Mbps connection.

For latency specifically, watch jitter — the variation between consecutive pings — rather than the average. Video calls tolerate a consistent 60 ms far better than an average of 40 ms that regularly spikes to 200 ms.

What improves over time, and what does not

Congestion improves as SpaceX launches more satellites into the shells serving your latitude, and as newer hardware generations with higher per-satellite throughput replace older ones. Both are happening continuously, which is why long-term Starlink users often report performance improving without changing anything.

The physics floor does not improve. Latency will not drop below what the distance allows, and the only lever there is routing — more ground stations closer to more users, and laser paths that avoid unnecessary hops. Those help at the margin; they do not change the fundamental number.

Your installation does not improve on its own either. If a tree is in the way, it will grow.

See the geometry for yourself

Much of what determines your instantaneous performance is visible in the orbital data. The coverage view on this site shows how many satellites are above your horizon right now and how high each sits — and elevation is exactly the variable that sets your slant range and therefore your latency floor.

A location with many satellites high overhead has better geometry than one where everything is skimming the horizon. That is not the whole story of your speed, but it is the part you can actually watch happening.