Why your ping is high even on fibre
Every week someone tells us they pay for a gigabit line and still sit at 120 ms in a game. Those two numbers have almost nothing to do with each other, and once you see why, the rest of this stops being mysterious.
Bandwidth is not latency
Bandwidth is how wide the pipe is. Latency is how long the pipe is. Upgrading from 100 Mbps to 1 Gbps makes the pipe wider — a download finishes sooner — but a packet still has to travel the same physical distance, through the same routers, at roughly two thirds the speed of light in fibre.
An old LAN game sends tiny packets, often under 200 kbps per player. It has never once been limited by your bandwidth. It is limited entirely by how long each of those tiny packets takes to arrive, and how consistently.
Where the milliseconds go
A rough budget for a player in Bucharest connecting to a host in Berlin, on a good day:
| Stage | Typical | Why |
|---|---|---|
| Your machine to your router | 0.3–1 ms | Cable. On Wi-Fi it becomes 2–15 ms and jumps around. |
| Router to your ISP | 2–6 ms | The local loop. Fibre is genuinely better than cable or DSL here. |
| Across your ISP's network | 3–10 ms | Depends on how well they peer. |
| Bucharest to Berlin, physics | ~9 ms | 1,300 km at two thirds light speed. Nothing can beat this. |
| Routing detours | 0–60 ms | The interesting part. Traffic often goes through Vienna, sometimes London. |
| Queuing on congested links | 0–40 ms | Evening peak. This is also where jitter comes from. |
| The host's own line | 1–30 ms | Their Wi-Fi and their upload are your problem too. |
Add it up and 120 ms is easy to reach without anything being broken. The physics part is 9 ms. Everything above that is routing and queuing, which is exactly the part that can be changed.
Rule of thumb: if your ping to a host is more than three times the straight-line physics estimate, you are not looking at a distance problem. You are looking at a path problem.
What you can actually fix
Get off Wi-Fi
This is the single biggest win in most households, and it is free. Wi-Fi adds a few milliseconds on average but a lot of variance, and variance is what makes a shooter feel bad. Twenty milliseconds of jitter is worse than forty milliseconds of steady ping.
Kill the queue at your own door
If someone in the house is uploading, your packets wait behind theirs. Most modern routers have SQM or fq_codel under QoS. Turning it on costs you a few percent of throughput and removes most bufferbloat.
Change the path
You cannot tell your ISP to peer better, but you can send traffic through a different set of hops. That is the whole idea behind LanOrbit's routing: measure several paths continuously, use whichever answers fastest right now, and switch when it degrades. On the Bucharest to Berlin example above, that usually takes 120 ms down to somewhere near 30.
What you cannot fix
Distance. If the host is in Sydney, 250 ms is the floor and no product on earth will change it — anyone claiming otherwise is selling something. Also the host's own upload: if they have 1 Mbps up and eight players in the room, everyone suffers and the fix is to host somewhere else.
Measuring it properly
A browser speed test tells you about TCP to a nearby CDN. It says nothing about UDP to the machine you are actually playing against. Ping the host directly if you can, watch the variance rather than the average, and check loss over a few minutes rather than a few seconds. Our ping test page shows the node side of that, and the app measures the rest once it is installed.
