How this instrument works
Not every computer once agreed on what a byte was. Early machines used all sorts of native word and character widths — 6-bit codes on some 1960s mainframes, 7-bit ASCII on early teletype-descended systems, 12-bit words on the PDP-8, 36-bit words on the PDP-10 — before the 8-bit byte, formally called an octet, became close to universal through the 1970s and 80s. ISO/IEC 80000-13 later fixed that convention in writing: a byte is an octet, exactly 8 bits, full stop.
The distinction between the two units survives in how they are written, not just what they mean. IEEE 1541 and the SI-derived unit-symbol convention use a lowercase b for a bit and an uppercase B for a byte, so 8 Mb and 8 MB describe quantities eight times apart — a distinction that plenty of marketing copy quietly ignores. Networking speeds are conventionally quoted in bits per second because data moves down a wire or through the air one bit at a time; storage capacity is quoted in bytes because that is the addressable unit memory and file systems actually work in.
That single difference in capitalisation is behind one of computing's most persistent points of confusion: an internet connection advertised at 8 Mbps sounds like it should download an 8 MB file in one second, but 8 megabits is only 1 megabyte, so the same link needs a full eight seconds — before accounting for protocol overhead, which trims the real figure further still. Reading a speed test result correctly means knowing which of the two units, bit or byte, is actually printed on the screen.
- Type your figure into the Bytes (B) field — it opens at 1, a single byte.
- Read Bits (bit) beneath it; the figure recalculates on every keystroke, always eight times the byte count.
- Working from an advertised network speed instead? Divide megabits per second by 8 to estimate megabytes per second.
- Only zero or positive values are accepted, since a byte count cannot run negative.
Worked example — one byte over a serial link
A microcontroller transmits a single ASCII character, one byte, down a UART serial line to a display module. Type 1 into Bytes (B) and Bits (bit) returns 8.0 — the raw data payload of that character, before the link adds anything of its own.
The wire itself carries more than those 8 bits, though. A typical UART frame wraps each byte with a start bit and one or more stop bits, so sending that single byte at a baud rate of, say, 9600 actually takes about 10 bit-times end to end, not 8 — the reason serial link throughput in bytes per second is always somewhat lower than baud rate divided by 8 would suggest.
Questions
Is 8 bits per byte a fixed rule, or does it vary?
Fixed today, though it was not always. Modern computing has settled on the 8-bit byte, formally an octet, and ISO/IEC 80000-13 defines it that way explicitly. Older architectures used other native widths — 6-bit, 7-bit, even 36-bit words — but essentially everything built since the 1980s, and every file format, network protocol and programming language in current use, assumes 8 bits to a byte without exception.
Why is an 8 Mbps connection not the same as 8 MB per second?
Because Mbps counts megabits and MB/s counts megabytes, eight times apart. An 8 Mbps link moves 8 million bits every second, which is exactly 1 megabyte per second before any protocol overhead — so downloading an 8 MB file over that connection takes roughly 8 seconds, not 1. Confusing lowercase b for uppercase B is the single most common reason an advertised internet speed feels slower than expected.
How do I tell whether a spec sheet means bits or bytes?
Check the capitalisation and the context. IEEE 1541's convention uses a lowercase b for bits and an uppercase B for bytes, so Mbps and Gbps almost always mean megabits and gigabits per second — network and radio speeds — while MB, GB and TB almost always mean bytes — storage and file sizes. When a figure is ambiguous, dividing or multiplying by 8 usually reveals which one a plausible real-world number implies.
Why did early computers use byte sizes other than 8 bits?
Because the byte started out as whatever chunk of bits a given machine's character set needed, not as a fixed standard. Six-bit codes suited some 1960s mainframes' character sets; the PDP-8 used 12-bit words; the PDP-10 used 36-bit words built for its own numeric and text conventions. The 8-bit byte won out largely because it comfortably holds one extended ASCII or Latin character while dividing evenly for binary arithmetic, and manufacturers converged on it through the 1970s.
How many bits does a single UTF-8 character take, compared with a byte?
It depends on the character. Plain ASCII text — English letters, digits, basic punctuation — takes exactly 1 byte, 8 bits, per character in UTF-8. Characters outside that range take more: accented Latin letters and many other alphabets take 2 bytes, most Chinese, Japanese and Korean characters take 3, and some symbols and emoji take 4 bytes, 32 bits, each. The byte-to-bit relationship itself never changes; only the number of bytes per character does.
Why does a UART or serial connection take longer than 8 bit-times to send one byte?
Because the raw data bits are only part of the frame. A typical asynchronous serial format wraps each 8-bit byte with a start bit and one or two stop bits, and sometimes a parity bit, so sending one byte often costs 10 bit-times on the wire rather than 8. Divide a serial link's baud rate by 10, not 8, for a more realistic estimate of its byte-per-second throughput.
Can this converter handle large data quantities, like whole files?
Yes, though for anything past a few thousand bytes it is more useful to work in kilobytes, megabytes or gigabytes and let a decimal- or binary-prefix converter handle the scaling, then apply this fixed times-8 factor to whatever final byte figure you land on. The 8-bits-per-byte relationship itself never changes regardless of how large the underlying number is.