Internet Speed Test
Pick the kind of line you have. A test sized for a cable modem will not fill a fiber circuit, and a test sized for fiber will sit there for three minutes on DSL — so this asks first instead of guessing.
The test runs between your browser and the nearest Cloudflare edge, which is the same path your traffic already takes — so it measures your connection rather than the internet at large. Your IP address, provider and edge location are read from the connection and shown under the results; we don’t store any of it, and it only reaches us if you choose to send us your results. The privacy policy has the detail. The guide that goes with this page explains what to do with each number.
On a metered or mobile connection, read this first. A speed test has to move real data to measure anything, and this one runs for a fixed length of time rather than a fixed size — so the faster your line, the more it transfers. On a gigabit connection the fiber profile can move well over a gigabyte in each direction. It will count against your allowance.
Choose Your Connection
Why Copper And Fiber Get Different Tests
Most speed tests run one test at one size for everybody. That produces two predictable failures, and they pull in opposite directions.
Too much data, not enough line
Push a fiber-sized transfer down a 20 Mbps DSL line and the test takes minutes, times out, or gets abandoned halfway. Worse, a long single transfer on a cable modem hides the thing that actually matters on coax: the upload, which is a fraction of the download and is what collapses first when someone in the house starts a video call.
Not enough data, too much line
One connection carrying a few megabytes will never fill a gigabit circuit. It finishes during the ramp-up and reports a number well under what you’re paying for. This is the single most common reason someone calls us convinced their new fiber is broken when it is working perfectly.
| Copper profile | Fiber profile | |
|---|---|---|
| Parallel connections | 4 | 8 |
| Transfer size per request | 3 MB down · 512 KB up | 16 MB down · 4 MB up |
| Measuring window | ~9 seconds per direction | ~11 seconds per direction |
| Ramp-up discarded | First 1.5 s | First 2.5 s |
| What the result emphasises | Upload headroom and latency under load | Whether you are actually getting the circuit you bought |
In both profiles the first seconds are thrown away rather than averaged in. A connection does not start at full speed — it ramps — and including the ramp is how a test quietly under-reports a fast line. What you see is the sustained rate after it settles.
Read These Three Together
Download is the number everyone quotes and the one that explains the least. Two others usually explain the actual complaint.
| Reading | What it tells you |
|---|---|
| Download | How fast bulk data arrives. Matters for large files and for several 4K streams at once. Almost never the reason a video call is bad. |
| Upload | Your side of every video call, every cloud backup, every large email attachment and every file you put into a shared drive. On copper this is typically a small fraction of the download figure. |
| Latency, idle | Round trip time on a quiet line. Under about 30 ms is comfortable, and this is what makes a connection feel responsive rather than fast. |
| Jitter | How much that round trip varies. Steady latency is what audio and video need; jitter is heard as choppy audio even when the average looks fine. |
| Latency under load | Round trip time measured while the line is saturated. This is the honest one. |
| Bufferbloat | How far latency climbed once the line was busy. A large climb is why the call falls apart the moment somebody uploads something — and it is a fixable equipment setting, not a speed problem. |
If the download figure looks fine and something still feels wrong, the answer is almost always in the last three rows. A line that tests at 900 Mbps and adds 300 ms of delay under load will fail a Teams call that a well-behaved 50 Mbps line handles without a stutter.
Before You Read Too Much Into One Run
A few things change the result more than the connection does:
Wi-Fi. If you tested over Wi-Fi, you measured Wi-Fi. A wireless client rarely sustains much more than half its advertised link rate, and on the far side of a block wall it will be a fraction of that — which is a coverage problem, not an internet problem. Our guide on Wi-Fi dead zones covers that case. Test again on a cable if you can, and compare the two.
A gigabit port. A 1 Gb Ethernet port tops out somewhere around 940 Mbps of real throughput once protocol overhead is accounted for. If you bought a gigabit plan and see roughly that, nothing is wrong — you have found the port, not the limit of the line. Faster plans need 2.5 Gb or better hardware end to end.
Everything else on the network. A backup, a game update or a cloud sync running in the background is measured too. So is an old router that cannot route at line rate no matter what it is fed.
One reading is a snapshot. Congestion is time-of-day dependent, particularly on shared coax. If you are building a case with your provider, run the test a few times across a day, wired, and keep the results.
Numbers Not Matching What You Pay For?
Send us the results. We’ll tell you whether it’s the line, the router, the Wi-Fi or the plan — and which of those is worth spending money on — before anyone schedules anything.
More Guides
What download, upload, latency, jitter and bufferbloat each actually tell you.
Why the mesh kit didn't fix it, and what does.
The issues we see most often in Tampa Bay homes — and what to try first.
What to do — and what to stop doing — when a drive starts clicking, before the data is gone for good.
How to know which type of service gets you fixed faster.
What proactive support actually saves you compared to calling only when something breaks.
How a monthly maintenance plan pays for itself in prevented downtime.
The proactive steps that keep your business off the list of ransomware victims.
How to tell when antivirus alone isn't covering you anymore.
What managed detection & response catches that antivirus can't.
Why a stolen password is now the most common way in — and how it's stopped.
What HIPAA, PCI DSS and NIST reporting should actually look like.
A quick check on whether your current IT setup is costing more than it should.
What actually happens during a remote session, step by step.
The three layers of cyber threats — and which one actually hits home users and small businesses.
Real, upfront rates for computer repair and managed IT services — no hidden fees.
📞 727-579-4335 · ✉️ support@baygeeks.com · Privacy Policy · Terms of Service