Cloudflare speed test
A Cloudflare speed test measures how your current connection performs to a nearby Cloudflare edge: download, upload, latency, jitter, and packet loss. It is a snapshot of that path, not a grade of your ISP’s name.
How it works
The public tool is speed.cloudflare.com. It runs in the browser against Cloudflare’s anycast network, which aims to use a nearby data center by routing, not by guessing your city from an IP map. Download and upload are measured with a series of payloads. Latency is measured idle and also while the connection is busy. Jitter is how much consecutive latency samples disagree. Packet loss is checked with UDP via a WebRTC TURN path. The page also shows quality-style scores for uses such as streaming, gaming, and video chat after the run finishes.
A speed test Cloudflare publishes is therefore a quality check as much as a megabit check. A high download with ugly loaded latency can still feel bad on a call. A modest download with stable ping can still feel fine for pages and mail. Read the set, not the hero number.
Because the far end is Cloudflare’s edge, you are mostly measuring the path through your ISP to that edge. You are not measuring a random website in another country. A slow government form or a slow game server can still sit behind extra networks this test never touches.
How to read the meters
Download: how fast data arrives from the edge, usually shown in megabits per second. Upload: how fast data leaves toward the edge. Unloaded latency: round-trip delay while the line is idle. Loaded latency: round-trip delay while a transfer is already running.
Jitter: average jump between consecutive latency samples. Packet loss: share of probe packets that did not complete the round trip. Cloudflare has described taking a high percentile of throughput samples rather than advertising the single fastest spike, so the headline speed is meant to look like a repeatable rate, not a lottery peak. Idle latency is collected early. Loaded latency is the number that explains why a video call falls apart while a file copy still looks “fast.”
A nearby edge is not the whole internet. It is the first honest hop; the slow site you care about may still live farther away.
How to run a fair check
Use the device and the access you actually complain about. A laptop on Ethernet, a phone on Wi-Fi, and a phone on cellular are three tests. Close other transfers. Repeat once. If you are comparing ISPs or rooms, keep the same device and the same tool. Do not bounce between this page and a different vendor’s meter and call the gap “the network.”
VPN, corporate proxies, and private DNS can all move the path. If you live on a VPN, test once with it on (your real working shape) and once with it off (the raw access). They will not match, and they should not. The Cloudflare run follows whatever exit you gave it.
If packet loss or jitter is the ugly line, look at Wi-Fi first: distance, interference, a crowded band. Then look at the access gear. A clean wired run that still shows loss is a different ticket from a messy radio run. The tool can tell you the symptom. It cannot see the wall.
Questions and answers
- What does the Cloudflare speed test measure?
- Download, upload, idle and loaded latency, jitter, and packet loss to a nearby Cloudflare data center, plus simple quality scores for common uses.
- Why might this result differ from my ISP’s test?
- Different servers, different methods, and different paths. Cloudflare’s test aims at its own edge. An ISP test often aims at a server on or near that ISP.
- Does a good Cloudflare result mean every website will be fast?
- No. It means the path to Cloudflare looked healthy. A site hosted far away, or a busy origin, can still be slow.
Related guides
Time Zones Explained: UTC Offsets, GMT, and Local Time
A clear guide to time zones, UTC offsets, half-hour and quarter-hour zones, GMT, daylight saving time, and accurate conversion between places.
Daylight Saving Time: How Clock Changes Work
Understand why daylight saving time exists, when clocks change in the United States and Europe, where it is not used, and how reliable time tools handle changing rules.
2 3 years dagestan
Two to three years is a span of 24 to 36 months, about 104 to 156 weeks, plus leftover days. The phrase is a duration, not a weekend, and not a single season.