Download Time Calculator
Work out how long a transfer takes at a given line rate, with realistic protocol overhead applied instead of the marketing arithmetic.
Result
Same file at other line rates
| Link speed | Realistic time | Goodput |
|---|---|---|
| 10 Mbit/s | 12h 26m 33s | 8.93 Mbit/s |
| 100 Mbit/s | 1h 14m 39s | 89.3 Mbit/s |
| 200 Mbit/s | 37m 20s | 178.6 Mbit/s |
| 500 Mbit/s | 14m 56s | 446.5 Mbit/s |
| 1000 Mbit/s | 7m 28s | 893.0 Mbit/s |
| 2500 Mbit/s | 2m 59s | 2.23 Gbit/s |
| 10000 Mbit/s | 45s | 8.93 Gbit/s |
About Download Time Calculator
Transfer time is size divided by rate - but only after you account for the difference between line rate and goodput. Headers, acknowledgements, retransmissions and the fact that no real transfer sustains 100% of the wire all eat into the number a naive calculation produces.
When the network is not the bottleneck
Before buying more bandwidth, check what actually limits the transfer. A single TCP stream on a high-latency path is capped by the window and by loss, not by capacity. Many small files are dominated by per-file overhead and metadata operations rather than throughput. Encryption, compression and checksumming can saturate a CPU core well below line rate, and a spinning disk or a saturated storage array often tops out first. Measuring a single large file transfer and a many-small-files transfer separately tells you which problem you have.
Where the missing capacity goes
A 1500-byte Ethernet frame carries 1460 bytes of TCP payload over IPv4, plus 38 bytes of framing per packet including preamble and inter-frame gap. That is about 6% lost before anything goes wrong. Add IPv6 (20 more bytes of header), a VPN tunnel, or a 1492-byte PPPoE MTU and it grows. Single-stream TCP over a long path is limited further by the bandwidth-delay product, which is a separate calculation entirely.
Common use cases
- Estimating a backup or migration window before committing to it.
- Explaining to a customer why their 100 Mbit/s service downloads at 11 MB/s.
- Comparing whether a link upgrade actually shortens a nightly job enough to matter.
Edge cases and gotchas
- Storage vendors count 1 GB as 10^9 bytes; operating systems often show GiB (2^30). The 7% difference shows up in every capacity argument.
- A single TCP stream on a high-latency path may not fill the link regardless of size - check the BDP calculator.