Network Transfer Time Calculator
Work out how long a copy actually takes before you start it. Enter the data size and the link speed, pick the share of line rate you expect to see, and get the wall-clock time for a NAS backup, a media library sync or an offsite replication job.
At your expected rate
1d 1h 23m 49s
Theoretical minimum
17h 46m 40s
Effective throughput
88 MB/s
The same transfer across link speeds
| Link | Ceiling | At your rate | Typical use |
|---|---|---|---|
| 100 Mbit (Fast Ethernet) | 13 MB/s | 10d 13h 58m 6s | Legacy or long cable runs |
| 1 GbE | 125 MB/s | 1d 1h 23m 49s | Default NAS and office LAN |
| 2.5 GbE | 313 MB/s | 10h 9m 31s | Modern switch, single large file copies |
| 5 GbE | 625 MB/s | 5h 4m 46s | NVMe shares and NAS to NAS |
| 10 GbE | 1250 MB/s | 2h 32m 23s | Server room backbone |
| 25 GbE | 3125 MB/s | 1h 57s | Data centre leaf, rarely needed at home |
The ceiling column is link rate divided by eight, with no protocol overhead subtracted. Treat it as the number you could never exceed, not a target.
Frequently Asked Questions
Why does 1 GbE never reach 125 MB/s?
Because the link rate is quoted in bits per second, not bytes. 1 GbE is 1000 Mbit/s, which is 125 MB/s only before any overhead is subtracted. Ethernet framing, inter-frame gaps, TCP and SMB or NFS overhead all take a cut, so a healthy 1 GbE transfer typically lands somewhere around 110 MB/s. The calculator works in bits throughout for exactly this reason.
What is a realistic efficiency figure to pick?
It depends entirely on the two ends. Two modern machines with a direct cable on 2.5 GbE can approach 90% of line rate. A NAS over gigabit ethernet to a laptop with a spinning drive on the other end is often limited by the disk, not the network, and might sit at 50% or less. Pick the share of line rate you actually expect and treat the theoretical figure as the ceiling.
Why is my backup slower on Wi-Fi than on ethernet?
Wi-Fi advertises a link rate that is shared and negotiated per frame, and the usable throughput is a fraction of it. A client on 5 GHz with a 866 Mbit link will see far less than 866 Mbit of real transfer because of contention, retries and protocol overhead. That gap between advertised and actual throughput is also what makes bufferbloat visible, since a saturated link queues packets and latency climbs before throughput does.
Does this account for compression or deduplication?
No. The calculation is raw wire throughput for uncompressed data, which is the worst case and therefore the number to size a link against. Compression on a backup source with compressible content will shorten the transfer; heavily deduplicated data shortens it further. Neither can be predicted without inspecting the data.