Ping, Jitter and 0 ms: Read Your HTTP Latency Test
Read browser latency and jitter results, understand 0 ms and reused connections, and see why HTTP/3 uses QUIC instead of TCP. Free tests with clear limits.
Read median, range and jitter together
The latency tool records eight sequential HTTP response times after one warm-up request. The median is the middle of the sorted measurements; the range shows the smallest and largest. Jitter here is the mean absolute change between neighboring samples, not a packet-level measurement.
For an illustrative sequence of 20, 30, 10 and 40 ms, the median is 25 ms and the adjacent-sample variation is 20 ms. A single number cannot tell you whether Wi-Fi contention, the network path, browser scheduling or the endpoint caused a slow response. Repeat under similar conditions before interpreting a difference.
What a 0 ms or missing timing means
The response-time samples and the setup breakdown are separate measurements. A reused connection can carry a new request without another DNS lookup or handshake.
| Display | Meaning | Next step |
|---|---|---|
| Cached or connection reused | No new DNS phase was measured for this request | Do not read it as the DNS resolver’s lookup speed |
| No new setup (connection reused) | The resource timing entry contains no new connection setup interval | Compare request response times rather than treating setup as round-trip latency |
| TCP not used (HTTP/3 · QUIC) | This request used HTTP/3, whose transport is QUIC | Read QUIC RTT if the edge reported one |
| Not reported | The required timing was absent or unusable | Keep it unknown; it is not a measured zero |
| Reported 0 ms (precision unavailable) | The edge supplied zero without enough precision to interpret it further | Do not infer zero physical delay or a reused connection from this field |
Compare Wi-Fi, mobile data or a VPN fairly
Use the same device and browser for a small comparison. The result is a troubleshooting observation, not a calibrated benchmark of your whole Internet connection.
- Pause downloads, streams and other heavy traffic, then run the latency tool and note the median, range and variation.
- Change only the network or VPN configuration you want to compare, wait for it to settle and repeat.
- Record several runs and the test time. If the observed edge or route changes, treat the result as a comparison of the complete paths.
- Use TLS Check for the HTTP protocol and edge-reported RTT. That metadata belongs to its own inspection request, so it need not equal a latency sample.
Keep a comparison log that can explain a change
For each run, note the time, network label, VPN setting, median, range and variation. If you want to compare a VPN, keep Wi-Fi and the device unchanged. If you want to compare Wi-Fi with mobile data, keep the VPN setting unchanged. A larger median after changing two things at once does not reveal which change caused it.
Repeat an unexpected result after the connection has settled. A one-off high sample can come from scheduling or other traffic, while several consistently slower runs provide better context for a support request. Keep setup labels in the log as well: cached DNS or a reused connection explains why a fresh request can have no new handshake interval. Do not add edge RTT to the browser’s total request time; the two values can overlap and come from different requests.
Why browser ping differs from game ping
An HTTP request includes browser and server work as well as network time. A game communicates with another endpoint and may use another protocol. A nearby Cloudflare result cannot predict the latency to a specific game server, and this tool does not measure ICMP ping, packet loss or download speed.
If the goal is hosting a game, use the Gaming IP page to prepare an address and inspect router WAN clues. For lag during play, also inspect the game’s own network statistics and repeat tests when the problem occurs. CGNAT clues and high latency are separate observations; one does not establish the other.
Frequently asked questions
Why does TCP RTT show “not used” with HTTP/3?
HTTP/3 uses QUIC rather than TCP. The TCP field does not apply to that request. IPNixa shows QUIC RTT separately when the edge reports it.
Does 0 ms mean my connection has no delay?
No. A setup phase may be skipped on a reused connection, or a server may report a zero without enough precision to interpret it. Those cases are labeled separately from missing values.
Is browser latency the same as ICMP ping?
No. These samples time small HTTP requests to IPNixa, including browser and endpoint processing. ICMP, games and other destinations use different measurements.
Can this test identify packet loss or bufferbloat?
No. Eight small HTTP requests do not measure packet loss or latency under a controlled upload or download load. Use a tool designed for those measurements.