What this tool helps you work out
A 1 Gbit/s link does not normally move files at 125 MB/s from beginning to end. Ethernet framing, TCP/IP, SMB, encryption, storage speed and competing traffic all reduce the amount available to the application. This tool converts the link rate into a practical throughput estimate using the efficiency you choose.
Enter your details
Before you begin
Use a measured rate when planning an important transfer. Link speed only describes the connection negotiated between two devices; it does not prove that the source disk, destination disk, server or route can sustain it.
Use measured values where possible and leave headroom for busy periods, overhead and future growth.
How the result is worked out
The calculator first converts bits per second into bytes per second, then applies the selected efficiency. The basic relationship is: usable MB/s = link Mbit/s ÷ 8 × efficiency. For example, 1,000 Mbit/s at 90% efficiency gives about 112.5 MB/s.
Worked example
A 2.5 Gbit/s Ethernet connection running at 88% efficiency provides roughly 275 MB/s of application throughput. A 500 GB transfer at that rate would take a little over 30 minutes before file-system and small-file overhead.
Practical tips
- Use wired Ethernet for repeatable large transfers where practical.
- Check that every switch port, cable and network adapter supports the intended rate.
- A single slow disk can become the bottleneck even when the network is fast.
- Thousands of small files usually transfer more slowly than one large file of the same total size.
Link speed is a ceiling, not a rate
A gigabit link is not a promise of a gigabit. The figure describes the signalling rate of the physical connection, and everything above that layer takes a share before your data moves at all.
On wired Ethernet the overhead is modest and predictable. Frame headers, the inter-frame gap and the preamble consume a few per cent, and TCP and IP headers a few more. A well-configured gigabit link moving one large file typically achieves around 940 Mb/s of actual data, which is about 94% — and that number is stable enough to plan around.
Wireless is a different matter. The advertised rate on a Wi-Fi product is a theoretical maximum assuming ideal conditions, maximum channel width, no interference and a single device. Real throughput of 40 to 60% of that headline is normal, and it falls with distance, obstruction and the number of devices sharing the airtime. A router advertised at 3,000 Mb/s is not going to move 3 gigabits of anything.
Throughput across a path is set by its narrowest point, and that point is often not the one being upgraded. A gigabit switch does nothing for a device connected over Wi-Fi, and a fast network does nothing for a file coming off a mechanical hard disk that can only read at 120 MB/s. Before upgrading anything, work out which link is actually the constraint.
What the numbers mean in practice
| Link | Headline | Realistic sustained | A 10 GB file takes about |
|---|---|---|---|
| 100 Mb/s Ethernet | 100 Mb/s | ~94 Mb/s | 15 minutes |
| Gigabit Ethernet | 1 Gb/s | ~940 Mb/s | 90 seconds |
| 2.5 Gb/s Ethernet | 2.5 Gb/s | ~2.35 Gb/s | 35 seconds |
| 10 Gb/s Ethernet | 10 Gb/s | ~9.4 Gb/s | 9 seconds |
| Wi-Fi 5, good conditions | 867 Mb/s | ~400 Mb/s | 3.5 minutes |
| Wi-Fi 6, good conditions | 1,200 Mb/s | ~600 Mb/s | 2.5 minutes |
The wireless figures assume a single device close to the access point with a clear path. Add a wall, a few metres, or three other devices and they fall substantially.
Diagnosing a link that underperforms
- Check what the interface negotiated. A gigabit port that has settled at 100 Mb/s is the most common cause of a mysteriously slow wired connection, and it is almost always the cable or the socket rather than the equipment.
- Check the cable. Gigabit needs all four pairs. A damaged or badly terminated cable can still carry 100 Mb/s perfectly, which is why the fault presents as slowness rather than failure. Cat 5e or better is required; older Cat 5 may or may not manage it.
- Check the storage at both ends. A mechanical drive reading at 120 MB/s cannot saturate a gigabit link, which needs about 118 MB/s sustained, so the drive and the network are closely matched — and a drive that is fragmented, busy or full will fall short.
- Test with something that removes storage from the equation. A memory-to-memory throughput test tells you what the network can do; a file copy tells you what the whole system can do. Comparing the two isolates where the limit is.
- Look for a duplex mismatch. Rare on modern equipment but devastating where it occurs: the link works, and throughput collapses under load with rising error counts.
- Count the hops. Every switch adds a little latency, and a cheap unmanaged switch under load can become the bottleneck for everything behind it.
Frequently asked questions
Why is a 1 Gbit/s network not 1 GB/s?
Network rates are quoted in bits. Eight bits make one byte, so 1 Gbit/s has a theoretical ceiling of 125 MB/s before overhead.
What efficiency should I use?
Around 85–95% can be reasonable for a healthy wired TCP transfer, but real results depend on the protocol, equipment and workload. Measure when accuracy matters.
Can Wi-Fi reach its advertised link rate?
Rarely as usable application throughput. Wi-Fi is shared, half-duplex radio and is affected by signal quality, interference, channel width and other clients.
Why does SMB copy speed fluctuate?
Caching, storage behaviour, antivirus scanning, file size and competing traffic can all change during the copy.
Does latency affect throughput?
Yes, especially over long-distance links or when TCP windowing, packet loss or application behaviour limits how much data can remain in flight.
Why do I only get about 940 Mb/s on a gigabit connection?
Because that is what a gigabit connection actually delivers. The remaining 6% is consumed by Ethernet framing, the inter-frame gap, and TCP and IP headers — overhead that is inherent to how the data is packaged rather than a fault or a limitation of your equipment. Anything in the region of 930 to 945 Mb/s on a sustained transfer means the link is working exactly as it should.
Will upgrading my switch to 2.5 or 10 gigabit help?
Only if the switch is genuinely the constraint, which it usually is not on a home network. If the devices connect over Wi-Fi, or the files live on mechanical drives, or the traffic goes out to the internet over a slower broadband connection, a faster switch changes nothing. Identify the narrowest point on the path first; it is frequently storage rather than network.
Why is my Wi-Fi so much slower than the number on the box?
The advertised figure is a theoretical maximum that adds together every radio and every stream the device supports, under ideal conditions with nothing else present. Real throughput of 40 to 60% of the headline is normal for a single nearby device, and it falls with distance, walls, interference and the number of devices sharing airtime. It is not a fault, and no product achieves its headline in a house.
Does a faster network make my internet faster?
Only up to the speed of your broadband connection. If your line delivers 100 Mb/s, upgrading an internal network from gigabit to 10 gigabit does nothing whatsoever for internet traffic. It only helps for traffic that stays inside the house — file transfers to a NAS, backups, video editing from network storage.
Related networking tools
References
- Microsoft: Network adapter performance tuning
- Microsoft: Network-related performance counters
- Microsoft: SMB file-server performance tuning
Official guidance and standards can change. Check the linked source when applying the result to an important network.