The Pipe Analogy
Bandwidth is the width of a pipe — how much water can flow through per second when the pipe is full. Latency is the length of the pipe — how long it takes a drop of water to travel from one end to the other. Increasing the width does not make water arrive faster; it just allows more of it to move simultaneously. These are independent properties, and both matter for internet performance, but they matter in different ways for different tasks.
Bandwidth is measured in megabits per second (Mbps) or gigabits per second (Gbps). Latency is measured in milliseconds (ms), usually as round-trip time. A speed test reports both, but only measuring bandwidth is like judging a road by its lane count without knowing how far away the destination is.
Where Latency Comes From
Latency has several components that stack up along the path between your device and a server:
- Propagation delay: The time for a signal to travel the physical distance. Light travels through fibre at roughly 200,000 km/s. London to New York is about 5,500 km, giving a one-way propagation delay of roughly 27 ms — meaning a round trip cannot be faster than 54 ms regardless of bandwidth. Propagation delay dominates on long-distance connections and is completely irreducible.
- Transmission delay: Time to push all the bits of a packet onto the link. At 1 Gbps, a 1,500-byte packet takes about 12 microseconds — negligible. On a 1 Mbps link it takes 12 ms — significant.
- Processing delay: Time for routers and switches to read headers, look up routes, and make forwarding decisions. Usually sub-millisecond on modern hardware.
- Queuing delay: Time a packet spends waiting in a buffer because the outgoing link is busy. This is the component most affected by congestion and the one that SQM (smart queue management) addresses.
Bandwidth-Delay Product
The bandwidth-delay product (BDP) is the amount of data "in flight" in a network path at any moment: bandwidth multiplied by round-trip time. A 1 Gbps link with 100 ms RTT has a BDP of 100 Mbits (12.5 MB). TCP must keep that much data unacknowledged in flight to fully utilise the link. Small TCP buffers on old equipment can prevent a fast, long-distance link from reaching its theoretical throughput even when neither end is congested. This is why a 1 Gbps fibre connection to a well-peered nearby server can transfer at near line rate while a connection to a server on the other side of the world may transfer at a fraction of the available bandwidth with poor TCP tuning.
High Latency + High Bandwidth: The Satellite Example
Geostationary satellite broadband illustrates the disconnect between bandwidth and latency. A traditional geostationary satellite sits 35,786 km above the equator. The one-way propagation delay alone is about 240 ms, giving a round-trip time of roughly 480–600 ms. Plans may offer 25–100 Mbps download, which is respectable bandwidth. But every click, DNS lookup, TLS handshake, and interactive request waits through that 500 ms round trip. File downloads can eventually reach rated speed because they bulk-transfer data with few round trips. Web browsing, gaming, and video calls feel sluggish because every interactive step pays the full latency penalty. Low Earth orbit (LEO) satellite services like Starlink dramatically reduce this by operating at 550–600 km altitude, yielding latency of 20–40 ms.
Low Latency + Low Bandwidth: The Legacy Modem Example
A dial-up 56k modem connected to a local ISP produced round-trip latency of 100–200 ms — slow by modern standards, but physically nearby. The connection was narrow (56 kbps) but not far away in network terms. Interactive tasks like text chat worked acceptably. Downloads were painfully slow because bandwidth was the bottleneck. This illustrates the opposite corner: proximity without capacity.
How Each Affects Different Use Cases
| Use case | Bandwidth need | Latency sensitivity | Primary bottleneck |
|---|---|---|---|
| Online gaming | Low (1–5 Mbps typical) | Very high — every frame update requires a round trip | Latency and jitter; bandwidth rarely limits games |
| VoIP / voice calls | Very low (100 kbps) | Very high — over 150 ms one-way becomes noticeable | Latency and packet loss; jitter buffer concealment helps briefly |
| Video streaming | Medium to high (5–25 Mbps per 4K stream) | Low — large buffers absorb latency variation | Bandwidth; latency only matters at start of stream |
| Web browsing | Low to medium | High — each page load involves many serial round trips | Latency for page feel; bandwidth for large media |
| Large file transfer | High — determines total transfer time | Low — initial connection setup cost is small relative to transfer | Bandwidth; BDP tuning for long-distance transfers |
Why Fixing Latency Often Matters More Than Adding Bandwidth
For interactive use, reducing latency from 80 ms to 10 ms is more noticeable than doubling bandwidth from 100 Mbps to 200 Mbps. A modern web page may involve 50–100 individual HTTP requests. Each request-response cycle is a round trip. At 80 ms RTT, 50 serial round trips take 4 seconds even if bandwidth is unlimited. At 10 ms RTT, the same 50 round trips take 500 ms. HTTP/2 and HTTP/3 reduce serial dependency through multiplexing and prioritisation, but the round-trip cost never disappears entirely.
Bandwidth still matters: if a household simultaneously streams 4K video, syncs cloud backups, and joins video calls, available bandwidth determines whether all activities can run without competing. Upload bandwidth is often the hidden constraint — many cable and fixed wireless plans provide asymmetric speeds, and saturating upload causes inbound latency to spike because the upstream queue fills. Smart queue management (SQM/CAKE) can recover much of this latency under load by actively managing the queues on both directions of the connection.
Frequently Asked Questions
What is the difference between latency and bandwidth?
Bandwidth is how much data a connection can carry per second. Latency is how long it takes a packet to travel to a destination and back. Bandwidth is capacity; latency is delay. They are independent properties that affect different aspects of the internet experience.
Which matters more for gaming?
Latency, jitter, and packet loss matter more than raw bandwidth for gaming. Most games use 1–5 Mbps but need fast, consistent packet delivery. A 10 Mbps connection with 15 ms latency will feel better than a 500 Mbps connection with 80 ms latency for real-time gameplay.
Can high bandwidth still feel slow?
Yes. A connection can have high Mbps but feel slow if latency is high, packet loss occurs, DNS is slow, Wi-Fi is unstable, or a website's server responds slowly. For interactive tasks, latency is usually the limiting factor once bandwidth exceeds a modest threshold.