How Bandwidth and Download Time Actually Work

What bandwidth actually measures

Bandwidth is the rate at which data moves across a network link, almost always quoted in bits per second. Despite the marketing language ("speed", "fast broadband"), bandwidth is not speed in the physics sense; it is throughput, the volume of data the pipe can deliver per unit of time. The two concepts get conflated because a higher-bandwidth link finishes a given download sooner, but for any individual packet the travel time across the planet is dominated by latency, not bandwidth. The bandwidth calculator ignores latency entirely and answers the throughput question: given a file of this size and a link of this speed, how long does the bulk transfer take?

Every byte that crosses a network link is eight bits. That single fact does most of the work in any bandwidth calculation, and yet it is the most common source of confusion. A "100 megabit" connection is a 100,000,000-bits-per-second connection, which translates to 12.5 megabytes per second in storage terms. The factor of eight between bits and bytes explains why a 1 GB game downloads in roughly eighty seconds on a 100 Mbps link rather than the ten seconds the labels seem to suggest. Internet providers price the bigger-looking number; storage devices use the smaller one; both are correct but neither is interchangeable with the other.

How the download-time formula works

The arithmetic behind the download time calculator is the simplest division in computing: convert the file size to bits, convert the link speed to bits per second, and divide. The complication is unit conversion, not maths. A 1 GB file is 8 × 10⁹ bits. A 100 Mbps link is 10⁸ bits per second. The transfer time is therefore 8 × 10⁹ ÷ 10⁸ = 80 seconds. Every other case is a variation on the same theme: pick consistent units, divide, and translate the answer back into something a human can read.

time (seconds) = file size (bits) ÷ link rate (bits per second) file size (bits) = file size (bytes) × 8 100 Mbps = 100 × 10⁶ bits/s = 12.5 MB/s

The base relation is exact and has been since IEC 80000-13 standardised the units in 2008. What is not exact, in practice, is the link rate itself. A 100 Mbps connection rarely delivers 100 Mbps end-to-end; the figure is a peak negotiated rate between your modem and the next hop. Real throughput depends on TCP overhead, the receiving server's upload cap, contention with other devices and the slowest link in the chain. Treating the calculator's output as a lower bound rather than a forecast keeps expectations honest.

Worked example

Consider a household downloading a 50 GB game over a 200 Mbps fibre connection. The file size in bits is 50 × 10⁹ × 8 = 4 × 10¹¹ bits. The link rate is 200 × 10⁶ = 2 × 10⁸ bits per second. Dividing gives 2000 seconds, which is thirty-three minutes and twenty seconds. The bandwidth calculator returns exactly that figure when you enter 50 GB at 200 Mbps. Run the same file over a 100 Mbps link and the time doubles to about sixty-seven minutes; over a 1 Gbps link it drops to under seven minutes.

For a more granular case, take a 4.7 GB DVD ISO sent over a USB 3.0 drive at 10 MB/s. The units here are bytes per second, not bits per second, so no factor of eight applies: 4.7 × 10⁹ ÷ 10⁷ = 470 seconds, or seven minutes and fifty seconds. The same image transferred over a saturated 100 Mbps internet connection would take 376 seconds — slightly faster, but only because internet plans are quoted in bits and the megabits per second figure happens to be larger than the megabytes per second figure on a slow drive. Get the units right and the cross-comparisons stop being mysterious.

Factors that affect real transfer time

TCP overhead and protocol losses

TCP, the protocol behind almost every download, is not a transparent pipe. Every packet carries headers, every received packet triggers an acknowledgement, and the sender ramps up its window size cautiously rather than blasting at line rate from the first byte. On a long transfer the overhead is small, perhaps three to five percent. On many short transfers, or on a lossy link where retransmits are common, the overhead can swallow more than ten percent of nominal capacity. The protocol design favours reliability over raw speed, which is the right trade-off for general use but means the calculator's theoretical figure is rarely matched.

Server and CDN limits

Your connection is only as fast as the slowest hop. A 1 Gbps home link counts for nothing if the server is throttling each connection to 10 MB/s, which many free download mirrors do to keep their own bandwidth costs predictable. Content delivery networks like Cloudflare, Akamai and Fastly shorten the distance to the user and remove this bottleneck for major sites, which is why Netflix and Steam routinely saturate residential gigabit links while a small game-mod download from a single server might struggle to crack 5 MB/s.

Wi-Fi and local network bottlenecks

Wireless adds two layers of friction. The radio link rate the laptop reports is a peak negotiated rate, not a sustained throughput; under typical conditions Wi-Fi 5 delivers roughly 60 to 70 percent of its quoted rate, and Wi-Fi 6 closer to 75 to 85 percent. The link is shared among all connected devices, and a phone backing up to iCloud in the next room can steal a meaningful slice of the available airtime. For a fixed download, a wired Ethernet link of any modern generation will outperform Wi-Fi most of the time, even if the Wi-Fi link rate looks higher in the menu bar.

Disk write speed

On very fast connections the bottleneck shifts from the network to local storage. A spinning hard drive sustains 100 to 150 MB/s on a good day; a budget SATA SSD does 400 to 500 MB/s; an NVMe SSD is comfortably above 2 GB/s. A 10 Gbps internet link can in theory deliver 1.25 GB/s, which exceeds what a spinning drive can absorb. For large downloads on gigabit-plus connections the drive's write speed often becomes the limiting factor, and the bandwidth calculator's network figure overstates how quickly the file can actually be written.

Time of day and ISP congestion

Residential broadband contracts are sold on a peak-rate basis, but the underlying infrastructure is shared. Evenings, especially the 19:00 to 23:00 window in most countries, see sustained throughput drop noticeably on contended cable and DSL plans. Full-fibre links are less affected because the local loop is not shared in the same way, but the upstream peering still is. The download time tool cannot model this; if a large transfer matters, scheduling it for the early morning often beats paying for a faster plan.

How to get closer to the calculator's figure

• Use a wired connection where possible. Even Wi-Fi 6 cannot match a gigabit Ethernet cable for sustained throughput, and the gap widens as the file gets larger. For a 100 GB game, the difference is often twenty to thirty percent.
• Test with multiple parallel streams. Tools like aria2, axel and many download managers split a single file into chunks fetched in parallel. On servers that allow it, this routinely doubles or triples the effective download rate by working around single-connection throttles.
• Pick a fast mirror. Linux distributions, large open-source projects and game launchers expose mirror lists. The geographically nearest mirror is usually the fastest; the one on the project's home page is usually the slowest.
• Check the disk first. On gigabit-plus connections, run a quick local write test before blaming the network. A full HDD or a USB 2.0 stick will throttle any link.
• Update the router firmware. Old routers cap throughput well below their paper rating, especially when NAT and stateful firewall rules pile up. A reboot every few months is often enough to recover lost speed.
• Mind the upload direction. If you are uploading rather than downloading, the link speed to use is the upstream figure, which on cable and ADSL is often a tenth of the download number.

Common mistakes

The biggest single mistake is conflating Mbps with MB/s. A "100 Mbps" plan does not deliver 100 MB/s — it delivers 12.5 MB/s at most, and usually less. The error is so common that it is worth saying out loud whenever an estimate looks wrong: check the units first.

The second is confusing decimal and binary prefixes. A "1 TB" hard drive holds 10¹² bytes, which Windows reports as 931 GiB but calls "931 GB" — the prefix is decimal, the interpretation is binary, and the storage is identical. The data storage converter shows the gap explicitly. For network speed it almost never matters because IEEE 802 uses decimal prefixes throughout, but for file size it absolutely does, and the calculator lets you pick the prefix that matches the source you are reading.

The third is assuming download time scales perfectly linearly. It does, in the limit of very large files. For small files, TCP slow-start dominates: a 1 MB file on a gigabit link does not finish in 8 milliseconds because the connection never gets to gigabit before the last byte is sent. For very large files, disk write speed dominates. The calculator is most accurate in the comfortable middle, roughly 10 MB to 10 GB.

A fourth mistake is ignoring the upload direction on asymmetric links. Cable and ADSL plans often advertise the download number ("up to 500 Mbps") while burying the upload figure (often 20 to 50 Mbps). Cloud backup, video calls and large file uploads all bottleneck on the upload link, and the time on a 1 GB upload at 20 Mbps is forty seconds, which is the same as a 1 GB download on a 200 Mbps link — five times longer than the headline download number would suggest.

When the calculation is not enough

For genuinely large transfers — hundreds of gigabytes or more — even an accurate bandwidth estimate is not the whole story. AWS Snowball, Google Transfer Appliance and Azure Data Box exist because at petabyte scale the cheapest, fastest option is to ship a physical drive across the country. The break-even point is roughly: if the network transfer time exceeds a few days at your sustained rate, a physical shipment is worth pricing. The bandwidth tool is the right place to find the network estimate that goes into that decision.

For business applications where throughput matters — backup windows, replication, large data pipelines — sustained rate over a working day is more useful than peak rate. ISPs rarely publish sustained-rate guarantees on consumer plans; business and enterprise contracts do, and the difference between "up to 1 Gbps" and "committed 200 Mbps" is the difference between a calculator estimate and a planning figure.

Frequently asked questions

The FAQ panel below answers the questions most readers arrive with: why real speeds lag theoretical ones, how Mbps and MB/s relate, where the MB and MiB confusion comes from, what speed is reasonable at home, and how the formula applies to streaming and uploads. The underlying maths in every case is the same line of arithmetic the bandwidth calculator runs: file size in bits over rate in bits per second. Everything else is unit hygiene and an honest estimate of the overhead.