Gigabit to Mebibit

1 Gbps (gigabit) broadband delivers up to 125 MB/s, which is more than sufficient for most households. It supports dozens of simultaneous 4K streams, fast game downloads, and video conferencing with headroom to spare. The limiting factor is usually the Wi-Fi router (Wi-Fi 5 maxes out around 400–600 Mbps in practice) or the speed of the remote server you're downloading from.

10 Gbps networking is standard in data centers, server interconnects, and high-performance workstations doing large file transfers (video editing, database backups). It is increasingly available in prosumer home networking equipment. At 10 Gbps, a 1 TB file transfer takes about 13 minutes under ideal conditions.

One terabit equals 1,000 gigabits (SI). Terabit-per-second (Tbps) speeds are used in long-haul fiber optic cables and internet backbone infrastructure. A single transatlantic fiber cable typically carries hundreds of terabits per second across many multiplexed channels.

Wi-Fi 5 (802.11ac) delivers up to 3.5 Gbps theoretical, but typically 400–600 Mbps real-world on a single device. Wi-Fi 6 (802.11ax) reaches 9.6 Gbps theoretical and 600–900 Mbps practical per device, with better multi-device handling via OFDMA. Wi-Fi 6E extends the same technology into the uncongested 6 GHz band, improving real-world speeds to 1–2 Gbps. Wi-Fi 7 (802.11be) pushes the theoretical maximum to 46 Gbps using 320 MHz channels and 4096-QAM, with real-world single-device speeds expected around 2–5 Gbps — the first Wi-Fi standard to reliably exceed gigabit in practice.

Modern data centers handle enormous simultaneous traffic between thousands of servers — cloud computing, video streaming, and AI training all require massive internal bandwidth. 100 Gbps links between switches are now standard; 400 Gbps is increasingly deployed for spine connections. At these speeds, a single link can move 50 GB of data per second, keeping pace with NVMe storage arrays and GPU memory transfer rates.

A megabit (Mb) = 1,000,000 bits (SI decimal). A mebibit (Mibit) = 1,048,576 bits (IEC binary = 2²⁰ bits). The mebibit is 4.857% larger. Network speeds use megabits (Mb); embedded memory and flash storage specifications use mebibits when binary precision is required.

Mebibit appears primarily in microcontroller and microprocessor datasheets (e.g. "2 Mibit flash memory"), FPGA configuration file sizes, and some wireless protocol standards (802.11 frame size limits, Bluetooth payload specifications). It is rarely seen in consumer-facing applications but is common in embedded systems engineering documentation.

Yes. In 2007, a class-action settlement required Western Digital to pay $2.1 million because their hard drives advertised capacity in decimal megabits/gigabits while operating systems reported binary values — making drives appear ~7% smaller than labeled. Similar suits hit Seagate and Samsung. These lawsuits accelerated industry adoption of IEC prefixes and pushed Apple (2009) and later Windows (2021) to clarify their capacity labeling.

SPI flash chips are addressed at the bit level during serial communication — the programr shifts data in one bit at a time over the SPI bus. Datasheets specify capacity in mebibits (e.g. W25Q16 = 16 Mibit = 2 MiB) because the serial interface operates on bits, not bytes. Calculating transfer time requires bit-level math: reading a full 16 Mibit chip at 80 MHz SPI clock takes about 0.2 seconds.

Flash memory chips organise storage in binary-aligned blocks (sectors, pages) whose sizes are powers of 2. Specifying capacity in mebibits (1,048,576 bits per Mibit) maps precisely to the physical organisation of the memory array. Using decimal megabits would result in non-integer block counts, making datasheet specifications harder to verify against hardware design.