Latency is how long a packet takes to travel to a destination and back, measured in milliseconds. It’s commonly called ping, after the tool used to measure it.
The thing worth understanding before anything else: latency and speed are completely independent. You can have gigabit fibre and terrible ping. Upgrading your speed tier usually does nothing for latency at all — which is why so many people pay more and notice no improvement in video calls or gaming.
Key Takeaway:
Bandwidth is how much data you can move at once. Latency is how long it takes to start arriving. They’re different problems with different fixes, and the fast-broadband upsell only addresses one of them.
Latency vs Bandwidth
Picture a motorway. Bandwidth is how many lanes it has — how much traffic can flow at once. Latency is how long one car takes to drive the route.
Adding lanes doesn’t make any individual car arrive sooner. That’s why a gigabit connection to a server on the other side of the world still feels sluggish, and why a modest connection to a nearby server feels instant.
Which matters more depends entirely on the task. Downloading a large file is a bandwidth problem. A video call, an online game, or a page that loads dozens of small resources is a latency problem. Full breakdown: Bandwidth vs Internet Speed.
What’s a Good Ping?
| Ping | Feels Like | Good Enough For |
|---|---|---|
| Under 20 ms | Instant | Competitive gaming, anything |
| 20–50 ms | Responsive | Gaming, video calls, everything normal |
| 50–100 ms | Fine | Browsing, streaming, casual gaming |
| 100–200 ms | Noticeable lag | Browsing OK, calls stutter |
| 200 ms+ | Frustrating | Real-time anything struggles |
Context matters more than the number. 90 ms to a server three countries away is excellent. 90 ms to something in your own city means something is wrong.
The Floor You Cannot Beat
This is the part that saves people hours of pointless troubleshooting.
Light travels through fibre at roughly 200,000 km per second — about two-thirds of its speed in a vacuum, because glass slows it down. That works out to about 5 ms per 1,000 km in each direction, or 10 ms round trip.
So London to New York is roughly 5,600 km. Physics alone puts the round trip at ~56 ms before you add a single router, and real-world figures of 70–80 ms are close to optimal. No router upgrade, no premium package and no “gaming” ethernet cable will beat that. The distance is the cost.
Practical consequence: if your ping to a distant server seems high, work out roughly how far away it is first. If you’re within about 20 ms of the theoretical floor, your connection is doing fine and the only real fix is choosing a closer server.
What Actually Causes High Latency
- Distance. Usually the dominant factor, and the one people ignore. See above.
- Wi-Fi. Adds a few milliseconds at best, far more with interference, weak signal or a crowded 2.4 GHz band. Ethernet is consistently better.
- Congestion. Shared infrastructure in your neighbourhood during peak hours, or too much happening on your own network.
- Bufferbloat. The big one almost nobody knows about — see below.
- ISP routing. Traffic sometimes takes a needlessly indirect path. Traceroute is how you prove it.
- Old or overloaded router. Cheap hardware handling many devices adds delay of its own.
- VPNs. Extra distance plus encryption. A nearby server costs little; one on another continent costs a lot.
- Satellite. Traditional geostationary service has a physical floor around 600 ms. Low-orbit services are far better but still above fibre.
⚠️ Bufferbloat: The Cause Nobody Mentions
If your ping is fine normally but becomes terrible whenever someone streams, uploads or downloads, you almost certainly have bufferbloat — and it’s fixable.
Routers and modems contain memory buffers to hold packets during busy moments. Manufacturers made these buffers far too large, reasoning that dropping packets is bad. The result is that under load your traffic sits in a long queue, and latency balloons from 20 ms to several hundred while the connection is technically working at full speed.
You’ll recognise it as: video calls falling apart the moment someone else starts a download, or game lag that appears only when the house is busy.
The fix is smart queue management — often listed in router settings as SQM, QoS, or under names like fq_codel and cake. Enabling it, and setting your bandwidth limits slightly below your actual line speed, keeps the queue short. Many routers don’t offer it at all, which is one of the few genuinely good reasons to replace one.
Latency, Jitter and Packet Loss
Three separate problems that get lumped together:
| Problem | What It Is | How It Feels |
|---|---|---|
| Latency | Round-trip delay | Everything responds late, but consistently |
| Jitter | Variation in that delay | Choppy calls, rubber-banding in games |
| Packet loss | Packets never arrive | Audio dropouts, freezes, disconnects |
A steady 60 ms beats an average of 30 ms that swings between 10 and 200. Real-time applications adapt to consistent delay; they cannot adapt to unpredictable delay. If calls feel unstable while your ping “looks fine,” jitter is the thing to measure.
How to Measure It Properly
A single ping tells you almost nothing. Measure over time:
- Windows:
ping -t 1.1.1.1— runs continuously, Ctrl+C to stop and see the summary - Mac / Linux:
ping -c 50 1.1.1.1— 50 probes, then statistics - Both: watch the min/avg/max line. A wide spread is jitter.
- Under load: run it while someone downloads something large. If latency multiplies, that’s bufferbloat.
- Our Speed Test reports latency and jitter alongside throughput.
Once you know there’s a problem, traceroute tells you where in the path it starts.
How to Reduce Latency
Ranked by how much difference they actually make:
- Choose a closer server. Biggest win available, every time. Game region, VPN location, CDN endpoint.
- Use Ethernet. Removes wireless variability entirely — the best fix for jitter.
- Enable SQM/QoS to kill bufferbloat, if your router supports it.
- Move to 5 GHz or reduce interference if wireless is unavoidable.
- Replace an old router — worthwhile mainly if it lacks queue management.
- Escalate to your ISP with traceroute or mtr output if the delay starts on their network.
What Won’t Help
- Paying for more bandwidth. Unless you’re genuinely saturating your line, a faster tier changes nothing about latency. This is the most commonly wasted money in home networking.
- “Gaming” cables and adapters. A working Ethernet cable is a working Ethernet cable.
- A VPN, for latency. It adds a hop by definition. It can help if your ISP is throttling or routing you badly, but it is not a latency optimiser.
- Closing every background app. Worth doing if something is genuinely saturating your upload, otherwise placebo.
Frequently Asked Questions
Is ping the same as latency?
Effectively yes in casual use. Strictly, latency is the delay and ping is the tool that measures it — but “my ping is 30 ms” is universally understood and perfectly fine.
Why is my ping high on fast internet?
Because they’re unrelated. Speed is capacity, latency is delay. Distance, Wi-Fi and bufferbloat all cause high latency on a fast line.
Does a VPN increase ping?
Almost always, since traffic takes a longer route and gets encrypted. A nearby server keeps it small. The exception is when your ISP is routing you badly and the VPN happens to take a better path.
What causes lag spikes in games?
Usually jitter or bufferbloat rather than raw latency — especially if the spikes coincide with someone else in the house using the connection. Test your ping while a large download runs.
Will Ethernet really lower my ping?
Usually by a few milliseconds, and far more importantly it removes variability. For gaming and calls, the stability matters more than the average.
What’s the lowest ping physically possible?
Roughly 10 ms per 1,000 km round trip, set by the speed of light in fibre. Within your own city you might see single digits; across an ocean, never.
Why does my ping vary so much?
That’s jitter — typically Wi-Fi interference, network congestion, or bufferbloat. Try Ethernet first, then test under load to check for bufferbloat.
Is 100 ms bad?
Depends on distance. To another continent it’s good. To a server in your own city it indicates a problem worth investigating.