File Transfer Time Calculator

Enter the size of the file and your connection speed. You get the transfer time with protocol overhead counted, the effective speed in both Mbps and MB/s, and the point at which shipping a drive beats your connection.

Size of the transfer (the file, folder, or backup you are moving)

Size unit (GB unless you know otherwise)

Connection speed (the number on your internet plan)

Speed unit (plans are sold in Mbps; progress bars show MB/s)

Honesty tier (how much of the advertised speed actually carries your file)

Custom efficiency (%) (used only when the tier above is Custom; 94 is a clean wired network, 80 a busy or wireless one)

Was this calculator helpful?

How the file transfer time calculator works

Every wrong download estimate on the internet is the same mistake, so let us clear it first: internet speed is sold in bits, files are measured in bytes, and a byte is 8 bits. Your plan says 500 Mbps with a lowercase b: megaBITS per second. Your game says 100 GB with a capital B: gigaBYTES. Divide the one raw number by the other and your estimate is wrong by nearly an order of magnitude, every time, in the disappointing direction. Think of it this way: your connection delivers slices, but your files are weighed in whole pizzas, eight slices apiece. Put both sides in the same unit before you divide, or the delivery estimate is off by a factor of 8. That single capital letter is the hardest working character in networking.

Two honest corrections make the simple division trustworthy. First, overhead: no real transfer runs at wire speed, because every packet spends part of itself on addressing before any of it carries your file. The protocol floor is about 94 percent on a clean connection, and an ordinary day with wifi and busy servers lands nearer 80; the honesty tier above lets you pick which story you want. Second, units that mean what you think they mean: the size menu carries both the decimal units on the download page and the binary units in your file manager, because the two families quietly disagree by 7 to 9 percent and this page refuses to fudge it.

The formula

bits to move = size in bytes × 8
effective speed = advertised speed × efficiency
transfer time = bits to move ÷ effective speed
MB/s = Mbps ÷ 8

Size in bytes comes from your unit exactly: decimal units step by 1,000 (1 GB = 1,000,000,000 bytes) and binary units by 1,024 (1 GiB = 1,073,741,824 bytes). Efficiency is the share of the advertised speed that actually carries your file: 1.00 at the theoretical tier, 0.94 at the protocol floor, 0.80 for a typical real day. The last line is the one to memorize. And here is the sanity anchor every correct calculator on earth agrees on: 1 GB at 100 Mbps wire speed is exactly 80 seconds.

Worked example

A 100 GB transfer on a 500 Mbps connection, protocol overhead counted: 100 GB is 100,000,000,000 bytes, and at 8 bits per byte that is 800,000,000,000 bits to move.

The connection's 500 Mbps is 500,000,000 bits per second, which reads as 62.5 MB/s in file-manager units. At the 94 percent protocol tier the effective rate is 500 × 0.94 = 470 Mbps, or 58.75 MB/s: the number a healthy progress bar would actually show. The division is 800,000,000,000 ÷ 470,000,000 = 1,702 seconds, so the transfer takes about 28 min 22 s.

And the courier line for that connection: at an effective 470 Mbps, a full day of transfer moves about 5.08 TB. So on this rather good connection, anything over about 5.08 TB arrives sooner as a drive in a box with 24-hour shipping (over 10.15 TB, even 48-hour shipping wins). A 20 TB NAS migration takes 3 days 23 h down the wire and one day in a padded envelope.

Why you never get the advertised speed

Even on a perfect line, part of every packet is not your file. A standard Ethernet frame carries 1,500 bytes, of which 40 are TCP/IP headers (the addressing that gets each packet to the right machine and reassembles the stream in order), and the wire adds another 38 bytes of framing around each one. Do the division and about 94.9 percent of the wire carries data; enable TCP timestamps, as most modern systems do, and it is about 94.1. That is our protocol tier: the ceiling for a clean, wired, uncongested path, and the reason a 940 Mbps speed test on gigabit fiber is a pass, not a problem.

The 80 percent tier is different in kind: it is our editorial figure, not a protocol constant. It stands in for the ordinary frictions of a real day, a wifi hop that retransmits, a server that throttles per connection, a shared building uplink at 8 pm, a distant route with high latency. Measured transfers on busy networks routinely land in the 70s and 80s, so 80 is an honest round middle, and the custom field is there when you know your own number. One honesty note in the other direction: some connections briefly beat their sticker, because ISPs deliberately provision a margin above the advertised rate (and some boost the first seconds of a transfer). That is marketing generosity, not physics; the wire-speed tier is still the ceiling for the plan you actually have.

The 7 percent hiding in your units

There are two families of size units and they quietly disagree. Decimal units step by 1,000: a gigabyte is a billion bytes, and that is what download pages, drive labels, and this page's GB mean. Binary units step by 1,024: a gibibyte (GiB) is 1,073,741,824 bytes, and that is what Windows actually computes, even though it prints GB on the label. The gap is about 7 percent at gigabyte scale and widens to about 9 percent at terabyte scale, which is most of the famous mystery of the shrinking drive: a 1 TB drive holds a true trillion bytes, and a trillion bytes is 931 binary gigabytes, so nothing was stolen. (Formatting and filesystem bookkeeping take a further small slice, but the unit switch is the bulk of it.) Pick the unit family that matches wherever you read your number and the math here is exact either way.

The station wagon theorem

Computer networking's most durable joke is also a real engineering principle. Andrew Tanenbaum put it in his textbook decades ago: "Never underestimate the bandwidth of a station wagon full of tapes hurtling down the highway." A shipped drive has enormous bandwidth and terrible latency. A 20 TB drive delivered in 24 hours works out to 160,000,000,000,000 bits in 86,400 seconds, which is an effective 1.85 Gbps: faster than almost every business connection in the country, with a first-byte latency of one day.

This page computes where the crossover sits for your connection: the payload at which a day of your effective speed equals a day of shipping. Past that size, the box wins, and the calculator says so whenever your transfer runs long enough for shipping to be a live option. Industry took the joke seriously long ago: AWS Snowball, a ruggedized appliance holding 210 TB, exists so customers can mail petabytes instead of uploading them, and drive manufacturers will courier you an empty disk for the same reason. When the numbers get big enough, the fastest network protocol is FedEx.

When the progress bar and this page disagree

Trust the progress bar. It is not estimating; it is measuring the actual throughput of your actual transfer, right now, with every real-world loss already included. What this page tells you is whether that measured number is reasonable. If your 500 Mbps plan should deliver about 58 MB/s at the protocol tier and the bar shows 52, everything is healthy. If the bar shows 7, something specific is wrong, and the shape of the number is the diagnosis: almost exactly one eighth of what you expected usually means a bits-and-bytes mix-up in your expectation, not a slow network. A number far below even the one-eighth line points at the server, the wifi, or a throttle. This is the same shape of problem as charging an electric car, a fixed bucket, a rated pipe, and honesty about losses, which is why our EV charging time calculator reads like this page's sibling: there the confusion is kW versus kWh, here it is bits versus bytes, and in both cases the cure is putting the two sides in the same unit before dividing.

Sources

Where the numbers on this page come from. We go to the body that publishes the figure, not to another calculator. See how we verify.

Actual throughput on Gigabit Ethernet, Rickard Nobel. The protocol-overhead arithmetic behind our 94 percent tier: a standard 1,500 byte MTU frame carries 38 bytes of Ethernet framing plus 40 bytes of TCP/IP headers, leaving about 94 to 95 percent of the wire for your data.
TCP over IP Bandwidth Overhead, Packet Pushers. Corroborates the figure: with TCP timestamps enabled the net efficiency of a 1,500 MTU Ethernet path works out to about 94.1 percent, and roughly 94.9 percent without them.
AWS Snowball Edge device hardware information, Amazon Web Services. The capacity class cited in the sneakernet section: the Snowball Edge Storage Optimized device holds 210 TB. AWS ships customer data by courier precisely because of the arithmetic this page computes.

Frequently asked questions

How long does it take to download 100 GB?

At 100 Mbps about 2 hours 22 minutes, at 500 Mbps about 28 minutes, and on gigabit fiber about 14 minutes, all with protocol overhead counted at 94 percent. The math is size times 8 (bytes to bits) divided by speed. The times are 8 times longer than dividing the two raw numbers suggests, because plans are sold in megabits and files are measured in megabytes.

Why is my download slower than my internet speed?

Almost always the factor of 8: a 500 Mbps plan moves at most 62.5 MB per second, so the 58 MB/s your download shows is actually excellent. Beyond that, 5 to 6 percent of the wire carries packet headers rather than your file, and the far server, the wifi hop, and the busiest link between you each take their cut. A download running at about 80 percent of the plan's MB/s equivalent is a healthy one.

What is the difference between Mbps and MB/s?

One capital letter and a factor of 8. Mbps is megabits per second, the unit internet plans are sold in; MB/s is megabytes per second, the unit file managers and progress bars display. A byte is 8 bits, so divide the plan number by 8 to get the file number: 500 Mbps is 62.5 MB/s. Mixing the two up is the single most common reason a download estimate is wrong by nearly an order of magnitude.

How fast is 1 Gbps in real life?

About 117 MB/s of actual file throughput: 1,000 Mbps divided by 8 is 125 MB/s, and protocol overhead trims roughly 6 percent. That downloads a 5 GB movie in about 43 seconds and a 100 GB game in about 14 minutes. It will not make web pages feel ten times faster than a 100 Mbps line, though, because page loads are dominated by latency, not bandwidth.

Why does my 1 TB drive show less than 1 TB in Windows?

The drive really holds a trillion bytes; Windows reports capacity in binary units where each step is 1,024, then labels them GB anyway. A trillion bytes is 931 binary gigabytes, so nothing is missing. The divergence is about 7 percent at gigabyte scale and about 9 percent at terabyte scale, and this calculator's unit menu carries both families so your numbers mean what you think they mean.

Is it faster to ship a hard drive than to transfer files online?

Above a surprisingly modest size, yes. A 20 TB drive delivered in 24 hours works out to about 1.85 Gbps of effective bandwidth, which beats most business connections. On a 500 Mbps line, anything over about 5 TB arrives sooner by overnight courier. This is not a joke calculation: AWS Snowball, a data migration appliance holding 210 TB, exists because trucks still beat cables at scale.

Does upload time follow the same math?

Exactly the same math, with one trap: most home plans are asymmetric, so the upload side can be a tenth of the sticker number or less. A plan sold as 500 down may be 20 up, which turns a 28 minute download of 100 GB into an 11 hour 49 minute upload of the same files. Put your plan's upload number in the speed field and this page prices it honestly.