Data Transfer Calculator
Convert Mbps to MB/s, decimal to binary units, and get real transfer times.
100% in your browser. Nothing you type ever leaves this page.
- Effective throughput
- 35.2 MB/s
- Same speed in bits
- 282 Mbps
- Finishes at
The same transfer on other links
Unit converter
Decimal units (KB, MB, GB) are powers of 1000, what drive makers and ISPs sell. Binary units (KiB, MiB, GiB) are powers of 1024, what your OS often reports. The gap reaches 9.95% at the TB scale.
Self-test passed: conversions and transfer math verified against reference values on load.
The factor-of-eight that explains most disappointment
Network links are sold in bits per second, files are measured in bytes, and the industry has politely declined to fix the confusion for forty years. Divide by eight: a 300 Mbps fiber line moves at most 37.5 MB per second, and after Ethernet, IP and TCP take their headers, a real-world 34 to 35. That single division explains the majority of "my internet is slower than advertised" tickets that ever reached a helpdesk. The calculator does the division, applies an honest overhead haircut, and shows both spellings side by side so nobody has to do mental math during an incident.
Estimates are floors, and here is what raises them
The time shown assumes the link is your bottleneck and the data streams continuously. Three things break that assumption daily. Small files: a million 50 KB files transfer dramatically slower than one 50 GB archive, because each file pays protocol and filesystem round trips; tar first, then move. Disks: a 10 Gbps LAN cannot outrun a source NAS reading at 110 MB/s. And latency: on high-latency paths, TCP needs window scaling to fill the pipe, which is why a transatlantic transfer can crawl on a fat link. When a transfer underperforms the calculator badly, the answer is almost always in one of those three places, not the cable.
Planning real operations with it
The comparison bars answer planning questions in one glance. A 2 TB backup over the office 300 Mbps uplink is a 15-hour affair, so it runs at night or it does not run. The same backup to a USB 3 disk is under an hour. Seeding a replica across a 10 Gbps datacenter link beats both. This is also where the ship-a-box services stop sounding silly: past a few dozen terabytes on ordinary WAN links, a courier carries data faster than the wire, with better jitter.
GB versus GiB, settled
The converter shows both ladders because both are in daily use. Storage vendors, ISPs and the SI standard mean powers of 1000; RAM, many OS dialogs and most of computing history mean powers of 1024, properly written KiB, MiB and GiB since the IEC named them in 1998. Neither is wrong, but mixing them silently costs real money in capacity planning: provision "10 TB" of decimal disk for 10 TiB of binary data and you are 950 GiB short. Our subnet calculator lives one habit away: in networking, the powers of two always win.
Frequently asked questions
Why does my 100 Mbps connection download at 12 MB/s?
Because bits and bytes differ by a factor of eight, and marketing always quotes bits. 100 megabits per second is 12.5 megabytes per second before overhead, and protocol headers shave a few more percent. Nothing is broken; two units are wearing similar names.
What is the difference between GB and GiB?
GB is decimal (10^9 bytes), GiB is binary (2^30 bytes, about 7.4% larger). Drive makers sell decimal gigabytes; Windows reports binary ones while labeling them GB, which is why a "1 TB" disk shows up as 931 GB. Both are correct numbers describing the same bytes in different dialects.
How accurate are the transfer time estimates?
They are physics plus a realistic protocol haircut, so treat them as best-case floors. The estimate assumes the link is the bottleneck; in real transfers, disk speed at either end, latency on small files, and shared bandwidth often dominate. A million small files transfer far slower than one big archive of the same size.
What does the protocol overhead option simulate?
The headers wrapped around your data: Ethernet, IP and TCP consume roughly 5 to 8 percent of raw line rate for typical packet sizes, so the toggle applies a 6% haircut. VPNs, TLS and small-packet workloads cost more; jumbo frames cost less. It is a sane default, not a promise.
Is it ever faster to ship drives than to transfer over the network?
Routinely, at scale. A 20 TB drive driven across town in an hour averages about 44 Gbps, which beats most WAN links, and cloud providers run that logic as a product (AWS Snowball). The classic rule applies: never underestimate the bandwidth of a station wagon full of tapes.