Measure your Starlink connection without an invented baseline.

The browser requests a large download from Cloudflare but stops reading after about 8 seconds and calculates throughput from the bytes received. It then repeatedly POSTs 2 MB upload blocks until the upload time limit. Five small HTTP requests provide a median round-trip latency; this is not ICMP ping. Results are stored only in this browser.

A browser test is useful for tracking your connection, but individual satellite runs can vary. Repeat the test at different hours and compare like-for-like conditions. Avoid testing during firmware updates or heavy weather, both of which can temporarily affect throughput.

The US median Starlink download was 129.61 Mbps in Q3 2025, according to Ookla 2025 Global Satellite Broadband Performance Report. This is an independent sample median, not a guarantee for your address. Your actual speed depends on cell congestion, dish placement, obstructions, local networking, and time of day.

The most common causes are obstructions (trees, buildings blocking the dish's field of view), peak-hour congestion between 7-11 PM local time, and suboptimal dish placement. Check the Starlink app's obstruction map first. If your dish has a clear sky view and speeds are still low, congestion in your cell is the likely culprit.

Testing once a week at a consistent time gives you a useful trend line. If you suspect a problem, test 3-4 times across different hours in one day — morning, midday, evening, and late night — to isolate whether the issue is congestion-related or persistent.

Yes. The test runs entirely in-browser and works on any device with a modern browser — phone, tablet, laptop, or desktop. Make sure your device is connected to your Starlink router's Wi-Fi, not cellular data, or you will be testing your cell carrier instead.

Fiber and cable are generally faster and lower-latency where modern wired service is available. Fixed 5G depends heavily on local signal and tower load, while legacy geostationary satellite generally has much higher latency. Check address-level availability, current plan terms, and your own repeated measurements instead of relying on an unsourced category average.