On a wired, symmetric 1 Gbps connection, the theoretical line rate is 125 MB/s. A real file transfer will be lower after protocol and system costs. The current calculator models a decimal 1 TB one-way transfer at 2 h 29 m using its 12% planning-overhead assumption.
That is a planning estimate, not an observed benchmark or guarantee. A healthy path might differ, and a “1 Gbps” residential plan may provide that speed only for downloads.
Translate gigabit service into file-copy units
Network rates use bits; file-copy tools usually use bytes. NIST distinguishes bits, bytes, and decimal and binary prefixes.
1,000,000,000 bits/s ÷ 8 = 125,000,000 bytes/s
So 1 Gbps is 125 decimal MB/s before Ethernet, IP, transport, encryption, request, storage, and application work. It is 119.2 MiB/s in binary units. A copy dialog showing “110 MB/s” is therefore much closer to gigabit than someone expecting “1 GB/s” might think.
The current calculator adds 12% to payload size. That allowance provides a consistent planning model; it is not a statement that every network is exactly 88% efficient.
Modeled times for common package sizes
These are one-way, decimal-size estimates at 1 Gbps:
| Dataset | Current calculator model | |---|---:| | 10 GB | 1.5 min | | 100 GB | 14.9 min | | 250 GB | 37.3 min | | 500 GB | 1 h 15 m | | 1 TB | 2 h 29 m | | 5 TB | 12 h 27 m |
If a recipient must download after the upload completes, add that leg. With symmetric gigabit at both ends, one 1 TB upload plus one download is 4 h 59 m before operational margin.
“Gigabit” must be symmetric to upload at gigabit
Fiber plans are often symmetric, but not always. Cable and fixed-wireless plans commonly advertise download first and provide a smaller upload. The FCC's U.S. Thirteenth Measuring Broadband America report, based on selected September–October 2022 measurements, documented upload asymmetry outside fiber. That is historical panel evidence, not a description of every current plan.
For contrast, 100 GB at 35 Mbps upload takes 7 h 07 m in the current model. The same package at 1 Gbps takes 14.9 min. Both users might see “gigabit” in advertising if the former refers only to download.
Verify the provisioned upstream rate, connect by Ethernet, and run an upload test to a relevant region during the planned window.
What prevents 125 MB/s?
Protocol and application work
Headers, acknowledgments, encryption records, request metadata, and retransmissions consume capacity. Exact efficiency varies by frame size, protocol, and workload. A theoretical protocol ceiling is still not an internet-path promise.
Wi-Fi
Wi-Fi advertises radio link rates, not sustained file payload. It is shared and half-duplex, and channel width, signal quality, contention, and client capability change constantly. A nominal link rate above 1 Gbps can still produce a transfer below wired gigabit.
Latency, loss, and TCP windows
A distant path needs enough in-flight data to fill the connection. ESnet's host-tuning guidance uses bandwidth × round-trip time to size this requirement: 1 Gbps over 50 ms represents about 6.25 MB in flight. Small windows or packet loss can leave capacity unused.
Storage and file count
Source media must sustain roughly the payload rate and the destination must write it. A single modern hard drive can exceed 125 MB/s for favorable sequential work, but fragmented sources, USB bridges, simultaneous I/O, and many small files can be slower. File creation and metadata operations make a 500 GB directory of tiny files different from one 500 GB file.
Destination and host
VPN gateways, CPU, endpoint scanning, encryption, per-connection throttles, cloud request limits, and verbose logs can become the ceiling. Azure publishes standard storage-account ingress targets far above 1 Gbps in listed configurations, but those regional account targets do not guarantee any individual transfer.
Test whether your result is healthy
- Confirm that upstream service is actually 1 Gbps.
- Use Ethernet and pause unrelated traffic.
- Benchmark representative source and destination storage locally.
- Transfer one large approved test file to separate file-count cost.
- Record observed MB/s, latency, loss, CPU, and disk activity.
- Repeat against another approved endpoint or region.
- Compare the evidence with the large-transfer bottleneck checklist.
Avoid a universal “good” percentage. A transcontinental encrypted upload and a same-switch LAN copy have different paths. Baseline your own production route, retain measurements, and plan from the lower repeatable rate rather than the best short burst.
When gigabit is enough
Gigabit is substantial capacity for occasional 100–500 GB deliveries and can support one decimal TB in a few hours when both endpoints keep up. Recurring multi-terabyte daily workflows need scheduling, recipient capacity, and aggregate egress planning. A 5 TB upload may fit overnight in the model, but the recipient leg, interruptions, and other traffic can push the complete handoff past a shift.
If the concern is unit conversion, start with 1 Gbps versus GB per second. If the target is a daily volume, use the 1 TB-per-day bandwidth guide.
Where TeraAirlift fits
TeraAirlift is a Windows-first desktop operations console for large-file delivery. Queue, progress and ETA, cancellation, retry-friendly operations, transfer history, diagnostics, recipient-controlled delivery, and SHA-256 verification support unattended gigabit workflows. They do not guarantee that a path will sustain 1 Gbps.
