Bandwidth, Latency, and Throughput: What's the Difference?

These three terms appear in nearly every discussion about the internet, and are often used as if they meant the same thing. In fact they measure different things, and different complaints usually trace back to one of them specifically.

The motorway analogy

Picture data as vehicles travelling from one city to another.

• Bandwidth is the number of lanes. More lanes, more vehicles passing at once.
• Latency is the travel time of a single vehicle from origin to destination. Adding lanes doesn't make one car arrive sooner.
• Throughput is how many vehicles actually arrive per hour. That figure is shaped by lanes, congestion, and road conditions.

The analogy has limits, but it's enough for one important point: adding bandwidth doesn't fix latency.

Bandwidth, the number they sell

This is what's printed on the subscription: 30 Mbps, 100 Mbps, 300 Mbps. The unit is megabits per second, not megabytes. One byte equals eight bits, so 100 Mbps means about 12.5 MB per second under ideal conditions.

That difference is what confuses people when downloading files: a 100 Mbps plan shows a download speed of around 12 MB/s in the browser, and that's correct.

Bandwidth determines how much can run simultaneously:

A household of four watching, working, and gaming at once rarely exceeds 60 Mbps simultaneously. This is why jumping from 100 to 300 Mbps often makes no perceptible difference.

Latency, the number that determines how it feels

Latency is the time a single data packet takes to travel to its destination and back. It's measured in milliseconds and usually called ping.

For interactive activities, latency matters far more than bandwidth:

• Video calls. High latency makes people talk over each other because the delay is noticeable.
• Online gaming. A 50 ms difference decides who appears to shoot first.
• Typing on a remote server. Letters appear behind your keystrokes.
• Opening a web page. One page needs dozens of round trips before it renders fully.

What determines latency is mainly physical distance and the number of devices in the path. Light in optical fibre travels about 200,000 km per second, so there's a floor no technology can break through. Jakarta to the United States and back is at minimum around 160 ms, purely because of the distance.

Jitter, unstable latency

Jitter is the variation in latency between packets. An average ping of 30 ms sounds fine, but if the figure jumps between 10 and 200 ms, video calls will sound choppy.

Calling apps work around this by holding data briefly before playing it. The greater the jitter, the larger the buffer required, and the more noticeable the delay.

On Wi-Fi, jitter often comes from competing for airtime with neighbouring networks. A LAN cable almost always gives lower jitter. How to measure it is covered in the speed test guide.

Throughput, what you actually get

Throughput is the real-world result after every factor plays out: available bandwidth, network congestion, Wi-Fi signal quality, the destination server's capacity, and packets lost along the way.

Throughput is always lower than bandwidth. A 10–20 per cent gap is ordinary, because some capacity goes to the protocol's own overhead.

A much larger gap points to an obstruction. The order of checks is in the slow Wi-Fi guide.

Bufferbloat: why video breaks up when someone's downloading

This is the most commonly misdiagnosed symptom. The story is typical: a video call is running smoothly, then someone in the house starts uploading a large file, and the call falls apart, even though bandwidth remains.

The cause is queuing. When the connection fills up, network equipment stores not-yet-sent packets in a buffer. An oversized buffer makes packets wait a long time for their turn, so latency leaps from 20 ms to hundreds of milliseconds.

How to check: run a speed test that measures latency under load, such as the Waveform Bufferbloat Test. Compare ping while idle with ping while a download runs. A gap above 100 ms indicates a problem.

The fix lives in the router. Look for a feature called Smart Queue Management, SQM, Adaptive QoS, or fq_codel. It caps queues to keep latency low, trading away a little peak speed. That's a trade almost always worth making in a shared household.

Why upload is often far smaller

What this means for video meetings and live streaming, and how to find what is uploading quietly in your home, is in the guide to why upload is slower than download.

Many home plans are asymmetric: 100 Mbps down but only 20 Mbps up. This is a legacy of the assumption that home users receive more than they send.

That assumption changed once online meetings became routine. A video call sends your video stream upward, and uploading work files needs the same capacity.

If your work involves a lot of video calls or sending large files, the upload figure deserves more attention than the download figure. Unfortunately it's rarely highlighted in plan advertising.

Matching the complaint to the cause

This table explains why upgrading the plan often doesn't resolve the complaint. If the problem is latency or bufferbloat, adding lanes to a road with a chaotic queue doesn't help much.

Frequently asked questions