Download—Mbps
Cloudflare edge test
Performance, not guesswork
Why do speed tests give different results?
Ookla, Fast.com, Cloudflare and M-Lab can report different speeds on the same broadband connection. Here is what each test measures—and where RaceMyNet fits.
Jump to the full guide ↓DownloadUploadPingJitterLoss
Upload—Mbps
Ping—ms
Jitter—ms
Packet loss—%
This is a methodology comparison based on provider documentation and RaceMyNet’s implementation, not a hands-on ranking or a claim that one service is always more accurate. RaceMyNet publishes this guide and is included in the comparison.
Ookla, Fast.com, Cloudflare, M-Lab and RaceMyNet compared
On smaller screens, swipe across the table to see all columns.
| Service | What it measures | Useful for | Keep in mind |
|---|---|---|---|
| Ookla Speedtest | Download, upload and latency between your device and a test server. | Checking throughput to a selected test destination. | Server choice and connection mode affect the result; it is not a guarantee of speed to every website. |
| Fast.com | Downloads and uploads to Netflix servers; expanded results include loaded and unloaded latency. | A quick check of the route to Netflix infrastructure. | A Netflix route is not the same as a route to a game server. |
| Cloudflare Speed Test | Transfers and latency measurements against Cloudflare’s network, with connection-quality diagnostics. | Exploring throughput alongside responsiveness. | Its measurement method and destination can differ from other tests. |
| M-Lab NDT | Single-stream download and upload performance, plus latency metrics. | Comparing how one connection performs. | Single-stream performance is not interchangeable with multi-connection capacity. |
| RaceMyNet | Cloudflare-based download, upload, ping and jitter checks, plus a browser loss check. | A visual connection check with links to troubleshooting guides. | Uses Cloudflare infrastructure with custom settings; it is not an independent test network. |
How do internet speed tests work?
A download test receives data from a server and measures how quickly it arrives. An upload test sends data in the other direction. Results usually use megabits per second (Mbps); a download app showing megabytes per second (MB/s) uses a different unit. Divide Mbps by eight for a rough MB/s conversion before allowing for overhead.
Latency measures the delay to a particular endpoint. Unloaded latency checks a relatively quiet connection; loaded latency checks it while traffic is flowing. Jitter describes variation in latency. Those distinctions matter when a connection downloads quickly but games or calls still stutter.
Ookla Speedtest
Speedtest tests the path between your device and a test server. Its server selection and connection settings are relevant when comparing runs. Record the server and use the same mode each time: changing the destination changes the path being measured. A fast result to one test server does not promise the same throughput to a distant download service.
Fast.com by Netflix
Fast.com transfers data to and from Netflix servers. Download speed is the main result; “Show more info” reveals upload speed and loaded/unloaded latency. Its settings also let you change test duration and parallel connections. It can give a useful second perspective when a different speed test reports another number.
Cloudflare Speed Test
Cloudflare tests against its own network and explains connection quality beyond a single bandwidth number. Its test method includes transfers of different sizes and latency measurements. A difference from Ookla is not automatically an error: the destination and method are different.
M-Lab NDT
M-Lab’s Network Diagnostic Tool measures single-stream performance and reports download, upload and latency metrics. A single stream can behave differently from tests using several connections. M-Lab publishes measurement data for research; review its privacy information before testing.
How RaceMyNet measures your connection
RaceMyNet uses the Cloudflare speed-test library and endpoints for bandwidth, ping and jitter, with its own transfer sizes, timing and result settings. Sharing infrastructure does not mean its results will exactly match speed.cloudflare.com. RaceMyNet currently measures unloaded latency, not a full loaded-latency or bufferbloat test.
The loss stage attempts a WebRTC measurement. If unavailable, it falls back to counting failed or timed-out HTTPS requests. That fallback is a request-failure estimate, not a direct measurement of IP packet loss; successful requests can hide retransmissions. Neither method proves what happens on your route to a specific game server.
Why do different speed tests show different speeds?
- Different destinations: each server can use a different route through your provider and the wider internet.
- Different workloads: connection count, transfer size, duration and result calculation affect the reported number.
- Different conditions: Wi-Fi signal, other household traffic, VPNs and time of day can change between runs.
- Different devices: browser overhead, adapters, cable links and device performance can limit a test.
How to compare speed tests fairly
- Use the same device and connection. Start with Ethernet where possible; check that the port can support your broadband speed.
- Pause downloads, cloud backups and streaming. Keep VPN settings consistent and note whether a VPN is in use.
- Run one service at a time. Simultaneous tests compete for bandwidth.
- Repeat each service three times, rotating the order. Record the time, server, settings, download, upload and latency.
- Compare the middle result and the spread. Do not select only the highest score or mix loaded and unloaded latency.
- Repeat during your usual busy period and compare with your actual app or game experience.
Record your own results; this guide does not invent benchmark figures. Speed tests transfer real data, so keep data allowances in mind.
Common speed-test questions
Why is Fast.com faster than Ookla Speedtest?
It can be faster—or slower—because Netflix and the selected Speedtest server use different routes and test settings. Repeat under the same conditions. One difference alone does not demonstrate ISP throttling or show which service is “wrong”.
Which speed test is most accurate?
Accuracy depends on the question. Checking capacity to one server, Netflix-route performance and single-stream throughput are different tasks. Look for repeatable results and an explained method rather than treating the largest number as the correct one.
What are useful Speedtest alternatives?
Fast.com, Cloudflare Speed Test, M-Lab NDT and RaceMyNet offer other ways to inspect a connection. Choose a different network destination when investigating a route-specific problem; RaceMyNet and Cloudflare share underlying infrastructure.
Which speed test should I use for gaming?
Use bandwidth as a baseline, then investigate latency variation and loss. Check your game’s network statistics because a browser test does not measure its exact server route. If ping rises when someone downloads, use a loaded-latency test and read our bufferbloat guide.
Does a good speed test mean my Wi-Fi is good everywhere?
No. Test in the room where you play or work, then compare beside the router and over Ethernet. Our Wi-Fi vs Ethernet guide helps interpret the difference.
Check your connection, then investigate the symptoms
Gaming connection checks · Slow downloads on PC · High ping explained
Sources and transparency
Provider documentation checked on 28 September 2026. Features can change. Service names belong to their respective owners; inclusion does not imply endorsement or a partnership.
COMPARE INTERNET SPEED TESTS
Choose the test that answers your question.
Use a repeatable throughput test for broadband checks, loaded latency for congestion, and in-game telemetry for your actual game connection. No single score describes the entire internet.