Two different questions
When people say an internet connection is “fast,” they may be talking about two different things. One is how much data the connection can move in a given time. The other is how long it takes for a signal to get there and back. The first is bandwidth. The second is latency. A connection can be strong in one and poor in the other, and different activities depend on different ones.
Bandwidth: capacity
Bandwidth describes the maximum amount of data a connection can carry per second. It is usually measured in bits per second: megabits per second (Mbps) or gigabits per second (Gbps). A plan advertised as 500 Mbps can, at best, carry 500 million bits each second.
A common point of confusion is bits versus bytes. File sizes are usually shown in bytes (megabytes, gigabytes), and a byte is 8 bits. So a 500 Mbps connection moves at most about 62.5 megabytes per second (500 divided by 8). In practice, real-world speed is lower because of protocol overhead, congestion and other limits.
A hypothetical example: a 100-megabyte file is 800 megabits. On an ideal 100 Mbps link, it would take about 8 seconds (800 divided by 100). Real transfers usually take somewhat longer.
The word “speed” is often used loosely for bandwidth, and strictly speaking the measured rate of useful data actually achieved is called throughput. Bandwidth is the ceiling, and throughput is what you experience.
Latency: delay
Latency is the time it takes for data to travel from one point to another, often measured as a round trip, meaning the time for a message to go out and a reply to return. It is measured in milliseconds (ms), where a millisecond is a thousandth of a second. The common “ping” test reports exactly this round-trip time.
A pipe analogy helps. Bandwidth is the width of the pipe, which determines how much water can flow at once. Latency is the length of the pipe, which determines how long it takes the first drop to arrive. A wide pipe does not make a long pipe shorter.
What causes latency
- Distance: signals in fiber travel at a good fraction of the speed of light but not instantly, so a server on another continent has a built-in minimum delay that no equipment can remove.
- Number of hops: each router the data passes through adds a small delay.
- Congestion: when a link is busy, packets wait in queues. This is a major source of delay variation.
- Connection type: wireless links, satellite connections and some older technologies tend to add more delay than a direct fiber or cable link. Satellites in very high orbits add a lot; newer low-orbit systems add less.
- Processing time: servers and network devices take time to handle each request.
- Local network: a weak Wi-Fi signal can add delay through retransmissions, a point covered when comparing home networks.
Jitter and packet loss
Two related measures matter for real-time use. Jitter is variation in latency: if delays swing from 20 ms to 150 ms and back, voice and video suffer even if the average is fine. Packet loss is the share of packets that never arrive. Some can be resent, but for live calls it often causes glitches or dropouts.
Which one matters for what
| Activity | Mostly depends on |
|---|---|
| Downloading a large file | Bandwidth |
| Streaming video | Bandwidth (with some buffering to hide delay) |
| Loading a web page with many small items | Both, often more latency than people expect |
| Video calls | Reasonable bandwidth, low jitter and loss |
| Online multiplayer games | Low, stable latency (data volumes are modest) |
| Remote desktop and cloud gaming | Low latency, plus moderate bandwidth |
Streaming works well even with moderate delay because the player downloads ahead and stores it in a buffer. Interactive uses cannot buffer, because the data depends on what you just did.
This is why web pages can feel slow on a high-bandwidth plan. A page involves many small requests, each paying a round-trip cost, and until responses arrive, the extra capacity goes unused. Content delivery networks and edge computing aim to reduce latency by placing servers closer to users. Claims about 5G latency follow the same logic: the headline benefit is often lower delay rather than only higher peak speeds.
Why the numbers on speed tests differ
Speed tests typically report download speed, upload speed and ping. They measure the path to a test server chosen by the tool, so results can differ from the path to the service you care about. Time of day, other devices on your network, Wi-Fi quality and the test server’s own load all change the numbers. Many residential plans have unequal download and upload capacity, which matters if you send large files or video.
Bufferbloat
Sometimes latency rises sharply when the connection is fully used, for instance during a large upload. Oversized queues in network equipment hold packets too long, and the delay builds up. This effect, known as bufferbloat, can make calls and games stutter while someone else is transferring files. Some routers include features that manage queues to reduce it.
Practical steps
- Judge a connection on both numbers. A very high-bandwidth plan with poor latency may disappoint for gaming or calls.
- For latency-sensitive use, a wired Ethernet connection is often steadier than Wi-Fi, though results vary with the home.
- Restarting a router, moving closer to it, or reducing background transfers can help.
- If problems persist, a provider can check the line, since the cause may be outside your home.
Neither number tells the whole story alone. Bandwidth tells you how much can be moved, latency how soon it begins to arrive, and a good experience usually needs enough of both.