Download time calculator
A 100 Mbps connection is 100 million bits per second, which is 12.5 million bytes — 12.5 MB/s. A 5 GB file therefore takes at least 400 seconds, not 50. Connections are sold in bits and files are measured in bytes, and that single factor of eight is behind almost every "my internet is slower than advertised" complaint.
Five related sums on one panel: how long a download or an upload takes, what connection speed a fixed time window needs, and conversions between Mbps and MB/s. The factor of eight runs through all of them: a 100 Mbps line moves 12.5 MB a second at best, so a 5 GB file needs 400 seconds, not 50.
How to use the download time calculator
The arithmetic is size in bytes times eight, divided by speed in bits per second. The factor of eight is the whole difficulty: broadband is sold in megabits per second and files are measured in megabytes, so a 100 Mbps line moves 12.5 MB every second at absolute best. A 5 GB game takes about seven minutes on that line, not fifty seconds.
The theoretical figure is never reached. TCP/IP headers, acknowledgements, retransmissions and TLS all consume capacity, and the shared segment between you and your exchange is contended with your neighbours. Fifteen per cent is a fair default for a wired connection; over wifi, twenty-five to forty is more honest, because the airtime is shared and half-duplex.
Upload is the same sum with a different number in it, and that number is usually much smaller. Cable and ADSL are asymmetric by design: a "500 Mbps" package commonly uploads at 20 or 50, and the advertised figure is always the download one. Full fibre is often symmetric, which is the practical reason it matters to anyone who sends large files. The asymmetry is also why a video call degrades the moment someone else starts a backup. The upload path saturates first, and that is the direction your own camera needs.
The speed-needed mode runs the sum backwards: a volume of data and a window give the floor, and the floor is not what you buy. The second figure adds twenty per cent, and that is the one to provision against, because queuing delay rises sharply as utilisation approaches capacity, so keep sustained use below about 70% if latency matters at all. For concurrent users the number that counts is the simultaneous peak, not the headcount; thirty people at 3 Mbps is 90 Mbps in theory and less in practice, because they are not all talking at once.
The two conversion modes exist because the units are distinguished only by letter case. Lowercase b is a bit, uppercase B is a byte, and it is the only unit in everyday use where capitalisation changes the value by 800%. Network units are also strictly decimal: a megabit is 1,000,000 bits, never 1,048,576, so converting Mbps to MB/s is a clean divide by eight with no binary correction anywhere in it. The binary prefixes belong to storage, where a "1 GB" file in a Windows listing is really 1 GiB: 1,073,741,824 bytes, seven per cent more than a decimal gigabyte, and enough to notice on a large transfer.
One thing worth checking before blaming the connection: the source. A transfer runs at the slowest link in the chain, and a server limiting you to 5 MB/s will do so whether your line is 100 Mbps or gigabit. If several simultaneous downloads each run at the same speed, the limit is at the far end.
What people use it for
- Estimating how long a large download will take
- Estimating how long an upload or an offsite backup will take
- Sizing a connection so a nightly transfer finishes inside its window
- Turning an advertised Mbps figure into the MB/s a download manager shows
- Turning an observed MB/s back into Mbps to compare against a contract
- Checking a monthly data cap against a sustained transfer rate
Questions
Almost always the factor of eight: connections are sold in megabits and files are measured in megabytes. A 100 Mbps line downloads at about 12.5 MB/s at best.
12.5 MB/s in theory. About 11 MB/s is a healthy real result, because protocol overhead costs 10–15% before anything else.
100 Mbps. Multiply megabytes by eight to get the line rate your contract is written in.
About 15% on a wired connection. Over wifi, 25–40% is more honest because the airtime is shared and half-duplex.
Usually the source. In order of likelihood: a server capping per-connection speed, then wifi, then an old router, then the disk.
Most consumer connections are asymmetric by design. The advertised speed is the download one; upload is often a tenth of it.
A speed test measures it. It is rarely the number on your bill.
A GB is 1,000,000,000 bytes and a GiB is 1,073,741,824: about 7% more. Windows says "GB" but means GiB.
No. Network units are strictly decimal. A megabit is exactly 1,000,000 bits.
Networks count bits because that is what goes on the wire; storage counts bytes because that is the addressable unit.
The second one, with 20% added. A link at its theoretical capacity has no headroom for retransmissions or bursts.
Keep sustained utilisation below about 70% if latency matters. Queuing delay rises sharply above that.
By simultaneous peak, not headcount. Contention makes the real peak lower than the sum of everyone at once.
The upload path saturates first, and that is the direction your camera needs.
Cloud storage clients often throttle uploads by default to keep the connection usable. Check the client’s bandwidth setting.
About 32 TB at 100 Mbps, which puts most data caps in perspective.
They measure different things. The test measures the line; the download measures the whole chain including the far end.
No. For a large file it is negligible; for many small files the per-request overhead dominates and this will be optimistic.