Online Ping Test

Measure round-trip latency to any hostname or IP address from our servers. Ping sends ICMP echo requests and times the replies, which answers two questions at once: is the host reachable, and how long does the round trip take?

Because the test runs from our infrastructure rather than your browser, it measures the path between our server and the target — useful for checking whether a host is up at all, and for comparing against a test from your own machine to work out which side of the internet a problem sits on.

Reading the result

• Minimum — the best case, effectively the speed-of-light floor for that path plus switching time. This number barely moves; it is set by distance.
• Average — the typical round trip, and the figure most people mean by "ping".
• Maximum — the worst sample. A maximum far above the average means something on the path is queueing or congested rather than simply distant.
• Packet loss — replies that never arrived. Sustained loss is a fault; a single missing reply to a busy router usually is not.
• Standard deviation or spread — how much the individual samples vary. A tight spread is a stable path; a wide one is jitter, which matters more than the average for calls and games.

What counts as good latency

Under 20 ms is excellent and typical of a nearby server on a wired connection. 20–50 ms is good and perfectly usable for calls and gaming. 50–100 ms is fair. Above 150 ms interactive applications start to feel sluggish, and above 250 ms voice conversations become difficult, because the delay exceeds what people naturally tolerate before talking over one another.

Distance matters more than most people expect. A round trip from London to Sydney cannot beat roughly 230 ms no matter what you pay your ISP. Light in fibre covers about 200 km per millisecond and real routes are never straight lines, so a large figure to a distant host is geography rather than a fault.

The number to watch is not the average on its own but how much it varies. A steady 60 ms is far more pleasant to use than an average of 40 ms that swings between 10 and 90.

Why a bad ping result is often not a real problem

Many routers deprioritise or rate-limit ICMP, because generating replies costs control-plane CPU that is better spent forwarding traffic. A router that ignores some pings can be passing your actual data perfectly.

This is the single most common misreading in network diagnostics. If ping looks poor but everything works, run a hop-by-hop MTR instead — it distinguishes a router that declines to answer from one that is genuinely dropping traffic, because real loss continues through every subsequent hop while ICMP rate limiting does not.

When to use something else

• Latency changes only under load — that is bufferbloat, and it needs a test that measures latency while the link is busy rather than idle.
• Calls break up but ping is fine — measure jitter and loss over UDP with the full connection test; a reliable transport hides the loss that ruins real-time media.
• You need to know where the problem is — ping tells you that something is wrong, not where. Traceroute and MTR locate it.