"The internet is slow" is one of the most common complaints any admin or ISP hears, and a number is the only good way to start that conversation. A speed test measures how fast data actually moves between your device and a server right now, in both directions, plus how quickly and how consistently the connection responds.
The NetTools.tech Internet Speed Test runs entirely from your browser against the NetTools server. It reports download and upload speed over parallel streams, latency, jitter and request loss, shows live charts, and keeps a history of recent results locally so you can compare tests over time.
What is an internet speed test?
In simple terms
A speed test downloads and uploads a known amount of test data for a fixed time and divides the amount by the time. That gives a speed in megabits per second (Mbps). Before that, it sends a series of tiny requests to see how long a round trip takes (ping) and how much that time varies (jitter).
Technical background
Throughput is limited by the slowest link on the path and by TCP behavior. A new TCP connection starts slowly and increases its sending rate as acknowledgements come back (slow start and congestion control), so the first moments of any transfer are slower than the steady state. A single TCP stream may also not fill a fast or long-distance link, which is why speed tests use several streams in parallel.
Speeds are given in decimal units: 1 Mbps = 1,000,000 bits per second and 1 byte = 8 bits. That matches how ISPs advertise plans. A file manager that reports MiB/s uses binary units and bytes, so a 100 Mbps line shows roughly 11.9 MiB/s at best.
Latency in a browser test is measured with HTTP requests rather than ICMP, because web pages cannot send raw ICMP packets. Browsers expose detailed timing for each request through the Resource Timing API; the interval from sending the request to receiving the first byte of the response excludes most of the browser's own queuing.
Jitter describes how much latency changes from one measurement to the next. Real-time applications like voice, video calls and online games are sensitive to it because they have to buffer to smooth out variation.
Why use a speed test?
- Verify what you are getting. Compare actual throughput with the plan, wired and wireless.
- Find the bottleneck. Testing on Wi-Fi and then on a cable often shows that the access line is fine and the wireless link is the limit.
- Check upload capacity. Backups, video calls and remote work depend on upload, which is often much lower than download.
- Judge suitability for real-time use. Latency, jitter and request loss matter more for calls and games than raw download speed.
- Track changes. The local history shows whether evening slowdowns or a new router change the picture.
Who should use it?
- Home and small-office users checking their connection or preparing a complaint to their ISP.
- IT support and help desks who need a quick, comparable measurement from a user's device.
- System administrators validating a new uplink, VPN or office connection.
- Remote workers checking whether a connection is good enough for video calls.
How the speed test works
The measurement runs in your browser; the NetTools server only serves test data and counts uploaded bytes. Traffic goes from your browser through Cloudflare to the NetTools server in Latvia, so the result describes that whole path.
- Preparing. The page requests a short-lived test session. The session is signed, bound to your IP address and valid for 60 seconds; at most 10 sessions per hour are issued per visitor. The response also tells the page your public IP address and whether you are connected over IPv4 or IPv6. The Cloudflare edge that served the request is read from the
CF-Rayheader. - Latency. Two warm-up requests open the connection, then 20 small HTTP requests are sent one after another to an endpoint that returns an empty response. Each has a 2-second timeout. The round-trip time comes from Resource Timing (request start to first response byte) where the browser provides it, otherwise from the elapsed request time. From the 20 samples the page calculates minimum, median, average, maximum, jitter and request loss.
- Download. Four parallel streams fetch incompressible random data for about 10 seconds. If the speed is above 150 Mbps between 2.5 and 3 seconds, two more streams are added (6 in total). The cumulative byte count is sampled every 100 ms.
- Upload. Three parallel uploads send 4 MiB blocks of random data for about 10 seconds; above 100 Mbps at the same point, three more uploads join (6 in total). The server counts the bytes and discards them immediately.
- Calculation. The average speed is computed from bytes actually transferred, excluding the first 2 seconds while connections ramp up (unless the test was too short for that). The peak is the best 1-second window. Idle time at the end, for example when a session byte budget was reached, is trimmed so it does not dilute the average.
- After the test. The page asks the IP Lookup service for your network's ASN and provider name, after the measurement so it does not disturb it.
Server-side safeguards: each download request is capped at 64 MB, each upload body at 32 MB, and a session can move at most 3 GiB down and 1 GiB up. A visitor may run at most 8 parallel streams and the server 64 in total. The endpoints only produce generated data; they cannot fetch or proxy other sites.
How to use the NetTools.tech speed test
- Open the Internet Speed Test.
- For a fair result, use a wired connection if you can, pause downloads, cloud sync, streaming and updates on your network, and close other heavy tabs.
- Select START TEST. The phase indicator moves through preparation, latency, download and upload while the gauge and the live chart show current values.
- Select Stop at any time; the results so far are shown as a partial test.
- Read the Results section, use Copy summary for a short text, or the export bar (CSV, JSON, Copy all, Print) for the full data including all samples.
- Run it again with TEST AGAIN. Completed tests are added to Recent tests, where you can chart download, upload or ping over time and Clear history.
The Change server button shows how many test servers exist; currently there is one.
Understanding the results
| Result | Meaning | What to look for |
|---|---|---|
| Download | Average receive speed after warm-up, in Mbps or Gbps; peak and data volume below | Compare wired results with the plan; a peak far above the average suggests fluctuating throughput |
| Upload | Average send speed after warm-up; peak and data volume | Often much lower than download on cable and DSL; matters for calls and backups |
| Ping | Median of the 20 HTTP round trips, with min and max | Stable low values; a max far above the median means occasional spikes |
| Jitter | Mean absolute change between consecutive latency samples | Low single digits is very stable; tens of milliseconds can make calls stutter |
| Request loss | Latency requests that failed or took over 2 s, out of 20 | Anything above 0% deserves a repeat test; it is HTTP-level, not ICMP loss |
| Connection Quality | NetTools.tech score out of 100 with the five component points | Use it to compare your own tests, not as a standard rating |
| Use hints | Checks for video calls, online gaming, 4K streaming and large uploads | Each shows the thresholds it uses, e.g. gaming needs latency under 60 ms and jitter under 15 ms |
| Network details | Public IP, IP version, ASN, provider, server, Cloudflare route, HTTP protocol, streams, browser | Unexpected provider or IP can mean a VPN or proxy is in the path |
| Charts | Throughput over time for both directions and every latency sample | Dips in throughput, or a latency sample that spikes or fails |
The quality score combines download (30 points, logarithmic, full at 1000 Mbps), upload (20 points, full at 500 Mbps), latency (25 points, full at 10 ms or less, zero at 200 ms), jitter (15 points, full at 2 ms or less, zero at 50 ms) and loss (10 points, zero at 5%). 85 and above is labelled Excellent, 70 Good, 50 Fair, below that Poor.
Example: Wi-Fi versus cable
A user on a 500/50 Mbps cable plan says video calls freeze. The figures below are illustrative.
| Test | Download | Upload | Ping | Jitter | Request loss |
|---|---|---|---|---|---|
| Laptop on Wi-Fi, next room | 142 Mbps | 18.4 Mbps | 31 ms | 22.6 ms | 1/20 (5%) |
| Same laptop on cable | 468 Mbps | 47.9 Mbps | 14 ms | 1.3 ms | 0/20 (0%) |
On cable the line delivers close to the plan with stable latency, so the ISP connection is not the problem. On Wi-Fi throughput is much lower and jitter is high, with one lost request. The video-call hint requires jitter under 30 ms and loss under 2%, which the Wi-Fi test fails on loss. The fix is on the wireless side: placement, channel, band or a wired connection for the work laptop.
Common use cases
- Checking a new broadband or fiber connection against the advertised speed.
- Comparing Wi-Fi with a wired connection to locate a bottleneck.
- Testing a connection before an important video call or a large upload.
- Measuring the effect of a VPN by testing with it on and off.
- Collecting a history of results at different times of day before contacting an ISP.
Troubleshooting
Download much lower than the plan
Test on a cable first. Pause other traffic on the network, including devices you do not see such as consoles and TVs updating. If the wired result is still low, try at a different time of day; consistent low values on a quiet wired connection are worth raising with your ISP, together with your history.
Upload fails with "HTTP 413"
A proxy between you and the server rejects large request bodies. This happens with some corporate proxies and security gateways. Test from a network without the proxy.
"At most 10 speed tests per hour"
Each visitor can start 10 test sessions per hour. Wait and try again; the limit protects the server's bandwidth for everyone.
"Too many parallel test streams"
Another speed test from the same IP address is running, for example in a second tab or on another device behind the same router. Let it finish first.
Very high ping on an otherwise fast connection
A VPN, proxy, mobile or satellite link adds latency. The network details show the provider and the Cloudflare edge; an unexpected location suggests traffic is leaving through a VPN or a distant gateway.
Results vary a lot between runs
Variation usually comes from other traffic or the wireless link. Compare several runs in Recent tests; for a longer look at stability, use the Network Jitter Test.
Common mistakes
- Testing on Wi-Fi and blaming the ISP. Always compare with a wired test.
- Confusing Mbps and MB/s. 100 Mbps is about 12.5 MB/s in decimal bytes.
- Comparing with other speed tests as if they were the same measurement. Different servers, protocols and calculation methods give different numbers.
- Treating the quality score as a standard. It is a NetTools.tech model for comparing your own tests.
- Reading HTTP ping as ICMP ping. Use Ping or MTR for ICMP measurements.
Limitations
- Results describe the path from your device through Cloudflare to one NetTools server; they depend on that server's connection and its distance from you.
- A single server is available, so you cannot choose a test server close to you.
- Latency, jitter and loss are HTTP-based and include a little browser and server overhead.
- Very fast connections may be limited by the browser, the device's CPU, the 6-stream maximum or the per-session data budget rather than by the line.
- Each phase lasts about 10 seconds; short-term congestion before or after is not captured.
- The ASN and provider shown come from routing and registry data and can be missing or describe a VPN or proxy provider.
Security & privacy
- What is sent: randomly generated test data, plus ordinary HTTP requests that reveal your IP address to the server, as any website visit does. No personal data needs to be entered.
- What is stored: uploaded test data is counted and discarded immediately. The recent-tests history lives only in your browser's local storage.
- Third parties: traffic passes through Cloudflare, which the NetTools server sits behind. After the test, your IP address is looked up in routing and registry data to show your network's ASN and provider.
- Abuse protection: sessions bound to your IP, a 60-second session lifetime, byte budgets and stream limits. The endpoints cannot be used to fetch other sites.
- What not to enter: nothing; the test has no input fields.
Related tools
- Network Jitter Test — longer latency stability test from your browser or our server. Guide: jitter test guide.
- What Is My IP — shows the public IP address and connection details the server sees. Guide: what is my IP guide.
- Ping — ICMP latency from our server to a host. Guide: Ping guide.
- MTR — per-hop loss and latency when you suspect the path. Guide: MTR guide.
- Bandwidth Calculator — converts a measured speed into transfer times. Guide: bandwidth calculator guide.
- Category overview: Network diagnostics guide.
Summary
The NetTools.tech speed test measures real HTTP transfers between your browser and the NetTools server through Cloudflare: 20 latency requests, then about 10 seconds each of parallel download and upload, with the first 2 seconds excluded from the averages. Ping is the median HTTP round trip and jitter the mean change between consecutive samples. Test wired, repeat a few times and compare the local history before drawing conclusions.
Frequently asked questions
Why is my speed test result lower than my internet plan?
The test measures the whole path from your device to the NetTools server, including Wi-Fi, other devices using the line, your browser, Cloudflare and the server's own connection. Plans usually describe the maximum on the access line, so a wired connection with other traffic paused gives the closest comparison.
Is the ping in the speed test the same as an ICMP ping?
No. It is the time of small HTTP requests from your browser to the NetTools server through Cloudflare, measured with the browser's Resource Timing data where available. It includes some HTTP and server overhead, so it is usually a little higher than an ICMP ping to the same place.
How is jitter calculated in the speed test?
Jitter is the mean absolute difference between consecutive successful latency samples, taken from the 20 latency requests. Other tools sometimes use the standard deviation or the RFC 3550 interarrival jitter, so values are not always directly comparable.
What does request loss mean?
Request loss is the share of the 20 latency requests that did not complete within 2 seconds or failed. It is an application-level measure, not ICMP packet loss, but repeated loss is still a sign of an unstable connection.
What is the NetTools.tech connection quality score?
It is a NetTools.tech model, not an industry standard. Download contributes up to 30 points, upload 20, latency 25, jitter 15 and request loss 10, each from documented thresholds, and the total is labelled Excellent, Good, Fair or Poor.
How much data does a speed test use?
It depends on your speed, because download and upload each run for about 10 seconds. On a fast connection a single test can transfer hundreds of megabytes; the result cards show how much data each direction actually moved.
Where are my speed test results stored?
The recent-tests history is stored only in your browser's local storage, up to 50 entries. Nothing from the history is uploaded, and you can clear it with the Clear history button.
Why do I get different results each time?
Throughput and latency depend on other traffic on your network, Wi-Fi conditions, the Cloudflare edge you are routed to and load along the path. Run several tests at different times and compare the history rather than relying on one result.