Starlink internet speed is not one number. It is download, upload and latency working together, and those three do not move in lockstep. Expect a usable broadband link whose results vary by hour, place, hardware and test method. This guide explains that variation and how to judge whether it fits your use case.
What “speed” actually includes
People say “speed” when they mean three different things:
- Download is how quickly data arrives: video, maps, updates, most browsing.
- Upload is how quickly data leaves: calls, camera, cloud backup, sending files.
- Latency is delay: how long a round trip takes before much data has moved.
A connection can download a movie quickly and still feel sluggish on a call if upload is weak or latency is jumpy. It can also post a high download figure and still stall if packet loss is elevated. For the vocabulary of a result screen, use how to read speed test results. For upload as its own problem, use the upload guide. For delay, use latency explained.
Starlink’s own plan pages list expected ranges that change over time. Treat the current official page as the commercial claim for a plan. Treat a local measurement as a snapshot. Treat coverage research as regional context, not a promise for your roof.
Public user reports and third-party aggregators show wide ranges that shift by region, hour and hardware. Treat those figures as context rather than a guaranteed speed for an address.
Download, in plain language
Download is the number most speed tests highlight. Streaming, app stores, and large incoming files care about it. Many households never notice upload until a call or a backup fails.
On Starlink, download is usually the stronger direction. That is normal for this service class; it is not a defect by itself. What matters is whether download stays high enough for the people and apps in the house at the same time.
Download moves around because the path is shared and wireless at several layers: the dish to the satellite, the satellite network, the ground station, and then the public internet. Add your Wi-Fi, and the test is no longer “Starlink.” It is Starlink plus the room.
A single high download number does not mean evenings will look the same. A single low number does not mean the kit is broken. Look at the pattern.
Upload, in plain language
Upload is the quieter constraint. Video calls, cloud photo backup, sending a large file, and some VPNs are upload-limited. If download looks fine and people still freeze on camera, you are probably looking at upload, latency, or Wi-Fi—not “Starlink is slow” as a blob.
Starlink upload is commonly reported as a smaller share of download in public tests, with plenty of scatter. Do not treat a round number you saw on a forum as your plan speed. Measure when you care, and read the upload article for improvement steps that belong there, not here.
Latency, in plain language
Latency is wait time. Throughput is volume. You can have plenty of volume and still wait.
Starlink is a low-Earth-orbit service, so delay is typically much lower than older geostationary satellite internet. Public reports often describe latency that is usable for video calls and many games, while still higher and more variable than a good fiber circuit. Exact results vary by location, routing, Wi-Fi and load.
If the question is “why does ping jump?”, go to latency explained. If the question is Starlink versus fiber as a buying decision, that comparison lives on Starlink vs fiber latency, not here.
Why the same dish posts different numbers
Scatter is the rule. The usual drivers:
Place and network neighborhood. Capacity and routing are not uniform. Two towns a hundred miles apart can feel different at the same clock time. Use Starlink Coverage for regional context, not as a street-level guarantee.
Time of day. Shared capacity shows up as evening softness more often than as a dead link. A Tuesday 10 a.m. test is not an 8 p.m. test.
Sky and weather. Obstructions, heavy rain, and snow on the dish can cut throughput or raise delay. Clear air with a blocked view still fails. See bad weather, rain, and the obstruction map.
Hardware and install. Mini versus a standard dish, a trip mount versus a fixed pole, a loose cable, and a roasting router all change the result. So does a phone on the far side of the house.
How you test. Browser versus app, Wi-Fi versus Ethernet, one stream versus a busy household, a VPN on or off. The test method is part of the number.
What else is running. A backup, a game download, and four cameras will eat the same pipe the test is trying to measure.
None of this requires a proprietary forecast weight. It is ordinary shared-network behavior plus a phased-array dish on your roof.
Why one test proves almost nothing
A speed test is a short sample of a moving system. It answers “what happened in these few seconds on this device on this path.” It does not answer “what is Starlink” or “what will Friday’s meeting be like.”
One test is useful as a clue. It becomes evidence when you repeat it:
- Same device and same room, wired if you can.
- More than one time of day, including the hour you actually care about.
- After you change one thing (VPN off, closer to the router, Ethernet).
- Written down with time, weather notes, and whether the house was busy.
Three calm tests that agree tell you more than one spectacular screenshot. The results guide is how to interpret the line items. The slow-connection guide is the diagnosis order if the pattern is bad.
Marketing screenshots, a neighbor’s story, and a single airport test are not your baseline.
Fit for common uses
Judge Starlink by the job, not by a round Mbps target.
Streaming and general home use. Download usually carries this if the install is clear and Wi-Fi reaches the TVs. Evening scatter is the main annoyance, not a zero.
Video calls and remote work. Upload, latency, and stability matter more than a peak download. Wired or a strong nearby Wi-Fi hop helps. Test at meeting hours, not only at noon.
Cloud backup and sending large files. This is an upload-duration problem. Start the job when the household is quiet. See the upload guide.
Gaming. Many titles are playable when delay is steady; some are miserable when jitter and loss spike. That tradeoff belongs in Is Starlink good for gaming?, not in a generic speed pitch.
Many devices at once. The kit can associate a lot of clients and still feel slow if they all want the airtime. Technical connection count is not comfort. Spread heavy jobs, and do not diagnose the dish until Wi-Fi and load are honest.
If you need a dated sense of weather-related risk for a trip, use the Connection Outlook. Speed fit is still a local measurement plus regional context.
Plan, kit, and mobility without a speed table
Residential, roam, and lite-style plans are commercial products whose advertised ranges change. Mini, standard, and performance kits are different radios and different installs. Public user reports often treat them as different clusters, not one number. TNET will not invent a Mbps table for those SKUs.
What you can do without a table:
- Read the current official plan page for the commercial claim.
- Measure the kit you actually own, at the hour you actually use it.
- Do not copy a Mini-on-a-tripod airport test onto a fixed residential pole, or the reverse.
- Hardware comparison as a buying decision belongs in Mini vs Standard vs Performance when you need that, not here.
Standby or paused service, if your account has it, is also not your active-plan speed. Test the mode you will live in.
How to measure without making this page the test
When you are ready to measure, use TNET’s Starlink Speed Test. That tool’s job is the run. This article’s job is expectation: what the numbers are, why they move, and which use cases care.
For a fair sample:
- Pause large downloads and backups.
- Stand near the router, or use Ethernet.
- Turn off a VPN for one run, then compare.
- Repeat at the hour you actually use the link.
- Read how to interpret the results instead of chasing a single headline number.
If the pattern is poor, do not keep hammering the test. Follow why Starlink is slow.
Regional context, not a promise
“What speed is Starlink?” is partly “where.” Coverage and performance research exist because national slogans hide local scatter. Open Starlink Coverage for country and regional context. Combine that with your own repeated tests. Neither layer is a contract for a street address.
Starlink remains a satellite broadband service with a clear-sky install, shared capacity, and weather sensitivity. It is often the right fit where fiber does not reach. It is a weaker fit if you need fiber-like delay and a flat speed all evening. Speed, in that decision, is the whole picture—download, upload, latency, and how much they move—not the biggest number on one afternoon test.