1 Gbps equals 125 MB per second at theoretical line rate, not 1 GB per second. There are eight bits in a byte, so divide network speeds in gigabits per second by eight to get decimal gigabytes per second. Protocol overhead and real conditions reduce file payload throughput further.
On a wired symmetric gigabit path, the current calculator uses a planning rate equivalent to about 111.6 MB/s of payload over its modeled duration. It does that by adding 12% to the payload, not by asserting a universal network efficiency. The 12% allowance is a planning assumption, not a law or measurement.
Bits and bytes differ by a factor of eight
Capitalization matters:
bmeans bit.Bmeans byte.Mbpsmeans megabits per second.MB/smeans megabytes per second.Gbpsmeans gigabits per second.GB/smeans gigabytes per second.
Therefore:
1 Gbps = 1,000 Mbps = 125 MB/s = 0.125 GB/s
A copy dialog showing around 110 MB/s on a nominal 1 Gbps link is not reporting one-tenth of the speed. It is showing bytes while the network plan is sold in bits, with additional overhead and system constraints.
Decimal and binary units create another difference
NIST's binary-prefix guidance distinguishes SI decimal units from IEC binary units:
| Unit | Bytes | |---|---:| | 1 GB | 1,000,000,000 | | 1 GiB | 1,073,741,824 | | 1 TB | 1,000,000,000,000 | | 1 TiB | 1,099,511,627,776 |
A decimal 1 TB drive contains about 931 GiB. Nothing disappeared; the label and operating-system display use different unit systems. Conversely, a 1 TiB package contains about 10% more bytes than a 1 TB package.
Use “GB” for decimal and “GiB” for binary when precision matters. If software labels a binary value as GB, note that convention before comparing totals.
Separate four speed numbers
The same transfer can have four valid numbers:
- Advertised line rate: 1,000 Mbps.
- Theoretical payload conversion: 125 MB/s before protocol costs.
- Calculator-modeled rate: the rate implied by the current 12% payload allowance.
- Observed throughput: what the transfer reports on the actual path.
Do not call an observed number “overhead” until you have ruled out other bottlenecks. Wi-Fi contention, an asymmetric internet plan, source storage, destination storage, latency, packet loss, VPNs, security processing, file count, and service limits all affect observation.
For a decimal 1 TB package, the current model estimates 2 h 29 m one way at 1 Gbps. The theoretical floor is 2 hours 13 minutes 20 seconds. Both assume the full stated rate remains available; only a real test produces an observed time.
Why protocol overhead exists
File bytes travel inside transport, internet, and link-layer structures. Ethernet framing, TCP/IP headers, acknowledgments, encryption records, request metadata, and retransmissions use capacity without becoming bytes in the delivered file.
The exact fraction depends on frame size, protocol, path, and workload. Jumbo frames can reduce framing cost on a controlled LAN, but every device on the path must support the configuration. They do not solve a slow disk, packet loss, or a constrained internet uplink.
The calculator deliberately avoids pretending it knows the exact stack. Its 12% planning allowance creates consistent estimates for comparison; it is neither a measured Ethernet percentage nor a guarantee that observed time will land within 12%.
A binary-size example
Suppose an application reports 500 GiB. That is 536.87 decimal GB, not 500 GB. At nominal 1 Gbps, the current model estimates 1 h 20 m for one direction.
Someone who reads “1 Gbps” as “1 GB/s” might predict 500 seconds. That mistake combines a factor-of-eight bit/byte error with a decimal/binary mismatch and ignores all other costs. Converting units before discussing performance prevents the largest error.
Read storage specifications carefully
Network vendors usually quote bits per second. Storage vendors and copy tools usually quote bytes per second. A drive rated at 270 MB/s has a nominal data rate of 2.16 Gbps before its own workload effects. That can feed a 1 Gbps sequential transfer but cannot by itself feed a 10 Gbps link.
Published storage figures are often best-case sequential rates. Small-file metadata, fragmentation, cache behavior, thermal limits, RAID, filesystem work, and simultaneous reads or writes can lower sustained output. Compare observed storage throughput in the same units as observed network throughput.
Use the units to diagnose, not to promise
First convert the advertised rate. Then measure a wired local test, the sender's internet upload, the recipient's download, and source/destination storage. If the transfer is slow, follow the bottleneck diagnostic guide. For practical gigabit expectations, use the 1 Gbps planning guide.
Keep theoretical, modeled, and observed values labeled in reports and client estimates. “The circuit is 1 Gbps” describes one boundary. It does not certify the end-to-end file path.
Where TeraAirlift fits
TeraAirlift is a Windows-first desktop operations console that shows package size before queueing and progress and ETA during delivery. Those displays help operators compare units and observed progress, but they are not speed guarantees. The console also provides retry-friendly operations, history, diagnostics, recipient-controlled delivery, and SHA-256 integrity verification.
