Internet Speed Test
Tests your connection against Cloudflare's edge network. Same engine as speed.cloudflare.com, no signup, no rate limits, no fake "boost" pop-ups. About 50 MB of test traffic, takes ~30 seconds.
About this test
What's actually being measured
This test uses Cloudflare's @cloudflare/speedtest JavaScript library — the same one that powers speed.cloudflare.com. It performs HTTP downloads and uploads of various sizes against Cloudflare's edge network, then uses the browser's PerformanceResourceTiming API to extract precise timing data.
Why these specific numbers
- Download / Upload (Mbps): peak sustained throughput. The test runs multiple parallel streams at different file sizes to find the real ceiling, not just a single-stream measurement.
- Latency (ms): idle round-trip time to the nearest Cloudflare edge — what your ping would show. Anything under 50 ms is excellent, 50–100 ms is fine for everything except competitive gaming, 100–200 ms is noticeable, 200+ is uncomfortable.
- Jitter (ms): variance in latency. High jitter wrecks VoIP and video calls even if your bandwidth and average latency look fine. Sub-5 ms is great, 10–20 ms is the threshold where you start hearing it on a phone call.
What can throw off the result
- WiFi vs Ethernet: WiFi adds latency, jitter, and a hard ceiling well below your ISP's plan. To know your real connection, plug in.
- Background traffic: a Steam download, cloud backup, or someone streaming Netflix in the next room steals bandwidth.
- VPN / WARP: adds 5–30 ms latency and usually caps throughput at 50–500 Mbps depending on the service.
- Browser overhead: JavaScript-based tests are bounded by the browser's network stack. On gigabit+ connections the test can underreport by 5–15%. For real wire-speed measurements use
iperf3against a private endpoint. - Local network bottleneck: a 100 Mbps cable, an old switch, or weak WiFi caps the result regardless of your ISP plan.
What the verdict means
The "Tank's verdict" line at the end is just a vibe-check based on the numbers — it doesn't compare you to other users or report anything anywhere. Tank is offering an opinion. He has many opinions.
This speed test uses HTTP, which doesn't tell you much about UDP behavior — the protocol your phone calls actually use. For real VoIP-style testing with a live audio echo, real-time jitter and packet loss measurement, and an estimated MOS score, try the VoIP Echo Test instead.
The test runs entirely between your browser and Cloudflare's edge. Cloudflare collects aggregate measurement data per their privacy policy. We don't log, store, or transmit your results — they only exist in your browser tab. Close it and they're gone.
How this speed test actually works
The tool uses the same underlying SDK that powers speed.cloudflare.com — Cloudflare's official speed test library. Under the hood, it opens multiple parallel HTTPS connections to Cloudflare's edge network and measures throughput on both directions along with latency and jitter. The results are directly comparable to what you'd get from Cloudflare's own speed test page.
The choice of Cloudflare as the measurement target matters. Cloudflare has one of the largest edge networks in the world, with points of presence in most major cities. When you run this test, you're measuring your speed to whichever Cloudflare location is closest to you — usually a few dozen milliseconds away, sometimes single-digit milliseconds if you're near a major metro. This is close to real-world browsing performance for most modern sites, since a growing majority of the internet is served through CDNs whose edges look similar to Cloudflare's.
Why the numbers might differ from other speed tests
The most common question about any speed test result is why it differs from another speed test. The answer is that different tests measure different things, and none of them is universally "right."
Speedtest.net (Ookla)
Ookla hosts their measurement servers at ISP locations — often literally inside the ISP's data center. This produces higher numbers because the path from your device to the test server is shorter and less contended than a typical internet destination. It measures the best-case scenario of your connection to your own ISP's infrastructure, which isn't the same as measuring the connection to actual internet destinations.
Ookla's number is a good measure of "what did I pay my ISP for?" and a poor measure of "what should I expect when I stream Netflix?"
Fast.com (Netflix)
Netflix's speed test measures throughput to Netflix's CDN. Since watching Netflix is presumably why some percentage of users care about their speed, this measurement is more relevant to actual Netflix experience than a generic internet measurement. Fast.com is intentionally minimalistic — it shows one big number, download only, and hides most of the technical detail.
Cloudflare (this tool)
Cloudflare's edge network is where a huge fraction of internet traffic terminates. When you visit a site that's on Cloudflare, Fastly, Akamai, or any other CDN, your connection is to a similar edge location as this test. The number this test gives is close to what you'd experience on typical modern sites.
Why the numbers usually order Ookla > Netflix > Cloudflare
Rough rule of thumb: Ookla usually shows the highest number because it's measuring to a close server. Netflix and Cloudflare usually show similar-ish numbers because they're both measuring to CDN edges. Any individual test might buck this pattern depending on network conditions and geographic factors.
What the specific numbers mean
Download speed
How fast data comes to you. For streaming, downloads, browsing, most of what matters happens at download speed. Numbers are given in Mbps (megabits per second). If you want megabytes per second (MB/s), divide by 8.
Upload speed
How fast data leaves you. Matters for video calls, uploading files to cloud storage, running a home server, screen sharing, backing up to remote storage. Consumer cable and DSL connections usually have much slower upload than download. Fiber connections are typically symmetric or close to it.
Latency (ping)
The round-trip time for a small packet to reach the test server and come back. Measured in milliseconds. For gaming, video calls, and interactive applications, latency matters more than raw bandwidth. Under 30ms is excellent, 30-60ms is normal, 60-100ms is acceptable but noticeable in latency-sensitive applications, over 100ms is a problem for real-time use cases.
Jitter
The variation in latency over time. Consistent latency of 40ms is much better than latency that swings between 20 and 100ms even if the average is 40. High jitter causes stutter in video calls, unpredictability in gaming, and generally makes real-time applications feel unreliable. Under 10ms of jitter is good. 10-30ms is manageable. Over 30ms is noticeable.
Loaded latency
Some speed tests measure latency while the connection is under load — bandwidth is being saturated by the download/upload tests. Loaded latency is often much higher than idle latency, and the gap between them (called "bufferbloat") indicates whether your connection has proper queue management. High bufferbloat means real-time traffic gets delayed when there's other activity, which manifests as video call quality dropping when someone else on the network starts a download.
What affects the results
Time of day
Residential ISPs often have congestion in the evenings when everyone is streaming. Same connection can be significantly slower at 8pm than at 2am. This isn't a fault; it's shared infrastructure being shared.
WiFi vs Ethernet
WiFi has its own bottlenecks. Even the fastest WiFi 6E setup can be slower than a direct Ethernet connection to the same router. If you're troubleshooting a slow connection, running the test over Ethernet eliminates WiFi as a variable.
Your computer
The device running the test has to be able to keep up. A slow CPU, a busy machine, an old browser version, or too many other programs running can all cap the throughput below what the network could actually provide. Speed tests measure end-to-end performance, and the client device is part of the "end."
Number of parallel streams
A single TCP connection is often bandwidth-limited by TCP's window sizing and congestion control, especially over long distances. Multiple parallel streams can saturate a link that no single stream can. This tool uses parallel streams to get closer to the actual link capacity.
Distance to the test server
Farther servers produce higher latency and often lower throughput. This test uses whichever Cloudflare location is closest to you, which minimizes distance effects. Speed tests that let you pick a distant server can produce arbitrarily bad numbers unrelated to your actual connection.
Middleboxes and proxies
Corporate networks with proxies, deep packet inspection, or content filtering can slow down speed tests. The tools weren't designed for measurement, but they can add latency and reduce throughput. If your work laptop shows much lower numbers than your home connection, the corporate network gear might be the reason.
Common speed test problems
"I'm paying for 500 Mbps but only get 300"
Some loss between the advertised speed and the measured speed is normal — usually 5-15%. If you're consistently getting less than 80% of your rated speed on multiple different tests at multiple different times, that's worth calling your ISP about. Also worth checking WiFi vs Ethernet, other devices on the network, and your router's stats page for any error indicators.
"Speed is great but everything feels slow"
Bandwidth is the size of the pipe; latency is how quickly small packets get through. High bandwidth with high latency feels fast for downloads but slow for browsing, which involves many small round-trips. If ping is over 100ms, individual page loads feel sluggish even though a speed test shows lots of bandwidth.
"Speed test says 200 Mbps but downloads run at 50 Mbps"
Speed test uses parallel streams; a single file download uses one stream. Single-stream TCP performance is often much lower than the aggregate the link can support. This is normal. If parallel downloads (or a download manager with multiple connections) achieves higher speed, that's why.
"Upload is way slower than download"
Normal for cable and DSL. Cable connections typically have 10-20% of their download speed for upload. DSL varies but often has significant asymmetry. Fiber is usually symmetric. Wireless (5G, LTE) is usually somewhat asymmetric but less than cable.
Related tools
The Bandwidth Calculator converts speed test numbers into estimated transfer times for specific file sizes. Useful for planning "how long will it take to move X gigabytes to the cloud" or "how many devices can I run video calls on simultaneously."
The VoIP Quality Test specifically measures the packet loss, jitter, and one-way delay characteristics that matter for voice quality, which is different from raw bandwidth measurement. If VoIP quality is your concern, that test is more directly useful.
For understanding why your specific connection might be slower than expected, running MTU / MSS checks can reveal fragmentation-related throughput issues that don't show up as obvious errors.
Privacy
The speed test connects to Cloudflare's edge measurement servers directly from your browser. No data about the test transits any myipcat.com server. Cloudflare, as the measurement endpoint, sees the traffic — but as an anonymous test connection with no unique identifiers. The test results are shown only to you.