Download
How fast data reaches you over one connection. It decides streaming quality and how long downloads take. Netflix recommends 15 Mbps for 4K and 5 Mbps for 1080p.
Internet speed test
Internet Speed Test
Download, upload and the number most tests skip: how much your latency rises when the line is busy.
Download--Mbps
Upload--Mbps
Idle latency--ms
Under load--ms
ServerWaitingM-Lab route
0.00Mbps
Ready when you are.
What this connection can handle
Compared with requirements published by Netflix, YouTube, Zoom, Microsoft Teams, NVIDIA GeForce NOW and Xbox Cloud Gaming. One run is a snapshot: test again at a different time, or on a cable, before drawing conclusions.
Your history
Past results on this device
Results are saved only in this browser (localStorage). They are never sent to our servers, and you can clear them at any time. Note that M-Lab itself publishes each test with its IP address, as stated above.
No saved results yet. Finish a test and it will appear here.
| Time | Down Mbps | Up Mbps | Idle ms | Bufferbloat |
|---|
Reading your result
What each number means
Upload
How fast you can send. It matters for video calls, cloud backup and live streaming. Zoom recommends 3.8 Mbps up for 1080p group video, so a 10 Mbps upload covers one or two calls but gets tight in a busy household.
Idle latency
The quickest round trip to the test server, in milliseconds. Lower is better for games and calls, and it mostly reflects distance and your access technology.
Latency under load
Round-trip time while the download is running. If it climbs far above idle latency, your router or modem is queuing too much data, often called bufferbloat. That is the usual reason calls break up when someone else starts a download.
Before you test
How to get a result you can trust
- Test where the problem happens. If video calls freeze in the back bedroom, run the test there, then once next to the router. The gap tells you whether Wi-Fi is the issue.
- Pause the big jobs. Cloud backups, game updates and other streams share the same line and will lower both numbers.
- Run it three times. Note the spread. One slow run is noise; three slow runs at the same hour is a pattern.
- Try a cable if you can. A wired result shows what your provider delivers; a Wi-Fi result shows what your device actually gets.
Different number elsewhere?
Why this test may not match Ookla or Fast.com
This test uses one connection to an M-Lab research server, which is closer to how a single file download behaves. Ookla uses several connections, often to a server inside your provider's network, so it tends to report higher figures on fast plans. Neither is wrong; they measure different paths. See exactly how we measure.
Calculators
Turn the result into an answer
CalculatorDownload Time Calculator
How long a 50 GB game or a 4 GB video takes at your measured speed, with an optional overhead allowance.
ConverterMbps, MB/s and Kbps ConverterWhy Steam shows 12 MB/s on a 100 Mbps plan, and how to convert between bits and bytes.
CalculatorHousehold Streaming CalculatorAdd up the streams and calls in your home and compare the total with your result.
Guides
When the number doesn't explain the problem
GuideWhy speed test results change between runs and sites
Servers, single vs multiple connections, browsers and evening congestion.
GuideWhy is my upload speed so slow?Cable's narrow upstream, the modem trap and what actually fixes it.
GuideVideo call speed requirementsZoom, Teams and Google Meet compared from their own documentation.
Common questions
Is this a real speed test?
Yes. It transfers real data to and from a Measurement Lab server for up to ten seconds in each direction using the open ndt7 protocol.
Does this site store my result?
No. Speed Check Hub does not store results. M-Lab, which runs the test servers, publishes every measurement with the public IP address and time, which is why you must accept its data policy first.
How much data does a test use?
It depends on your speed, because each direction runs for a fixed time rather than a fixed size. On a fast connection a single run can use hundreds of megabytes, so be careful on a capped mobile plan.
Why is there no jitter number?
The ndt7 protocol reports round-trip times from TCP, not the packet-by-packet timing needed for a meaningful jitter figure. We would rather show the latency-under-load number we can measure honestly.