It could have been anything
In the previous article, we learned to read eight switches as one value. In the first, we learned why those switches exist at all. This article asks what happens when you stop looking at individual bits and start asking: how many should we bundle together?
If you have switches, you can group them however you want. Nothing in physics says a group must contain eight. So why does every computer talk about bytes?
Bits are just switches
Eight switches. Not a number yet. Just on and off.
The question isn't what they mean. It's how many belong in a bundle. Group too few and you can't represent enough. Group too many and everything gets expensive. Somewhere in the middle, engineers had to pick.
Trying different group sizes
Imagine you're designing the first character-based machine. Each group of switches must represent symbols: letters, digits, punctuation, control codes. How many switches per symbol?
The search for enough roomSixteen symbols. The alphabet alone needs twenty-six. Not even close.
Now the letters fit. But numbers, punctuation, spaces, and control characters still need room. Still too tight.
Sixty-four symbols. A real improvement. Enough for a useful subset, but still cramped for a general-purpose text system.
Ever wondered why Base64 uses exactly 64 characters? Because every Base64 digit encodes exactly six bits. The encoding inherited the room size of a 6-bit world: 2⁶ = 64. The standard outlived the hardware that inspired it.
One hundred twenty-eight characters. Letters, digits, punctuation, control codes. ASCII proved that seven bits was enough to write English. For a while, it seemed sufficient.
Double the space. Room for accented characters, symbols, extended sets, and future expansion. Not because seven failed, but because eight left headroom. Engineering prefers margin.
History had other ideas
Knowing that eight bits is a good size doesn't explain why it became the size. For years, it wasn't.
Early computers used 6-bit, 7-bit, 8-bit, and even 9-bit bytes, depending on the manufacturer. There was no universal standard. A "byte" originally meant the smallest unit a machine used to store one character, and that unit varied from architecture to architecture.
That changed when IBM standardized the System/360 around 8-bit bytes in 1964. One decision, one ecosystem, millions of machines aligned. The argument didn't end overnight, but the direction became clear.
The word "byte" comes from "bite" (as in a small bite of data), deliberately misspelled to avoid confusion with "bit." When Werner Buchholz coined it in 1956, a byte wasn't fixed at eight. It meant whatever chunk of bits a given machine used for one character. Eight won later. The name stayed.
A good trade-off
Eight wasn't mathematically special. It sat in the middle of a trade-off engineers still make today.
Not enough symbols. Not enough range. Constant workarounds to fit data that doesn't belong.
More memory. More wiring. More cost. Paying for capacity you don't use yet.
Standards are often born from good trade-offs, not perfect ideas. That pattern shows up far beyond hardware: protocols, languages, APIs. Good enough, widely adopted, hard to replace.
Meet the byte
Today, a byte almost always means eight bits. That's the convention this trilogy has used in every example.
Eight switches, one bundle, one addressable unit of memory. We've treated every value so far as unsigned: 0 through 255. Signed numbers, where the highest bit means something different, are a story for another article.
Four bits have a name too
Half a byte already has a name: a nibble (sometimes spelled "nybble"). Four bits. Sixteen possible values. Enough to represent one hexadecimal digit, though we're not going there yet.
Split any byte down the middle and you get two nibbles. That split will matter soon. For now, just notice that the byte has structure inside it.
If "byte" is a bite of data, "nibble" is half a bite: a nibble. It's a historical pun that stuck around in engineering slang. Not official terminology you'll find in every spec, but you'll hear it in debug sessions and datasheets often enough to remember.
Why this matters
Once bytes became the standard, everything built on top of them.
These tools don't show bits because humans struggle to read long strings of ones and zeros. They show bytes, grouped the way history grouped them. Not an arbitrary choice by the toolmaker. A consequence of the standard that won.
If you grew up with game consoles, you've heard this vocabulary before. The NES era was "8-bit." The SNES and Mega Drive were "16-bit." The Nintendo 64 put the number right in the name. It sounds like the same idea as this article. It isn't quite.
When the industry said "8-bit," it usually meant the processor: 8-bit CPUs like the 6502 inside the NES. Registers, data paths, the width of a single operation. That aligned with a world already thinking in bytes, but "8-bit console" was shorthand for the CPU generation, not a claim about how memory was labeled on paper.
The jump to "16-bit" followed the same logic: wider processors, more capable hardware, bigger games. Still bytes underneath. Still eight bits per addressable unit in memory.
Then came the Nintendo 64. The "64" marketed a 64-bit CPU: wider registers, a broader address space, a bigger number on the box. Memory in that console was still organized in 8-bit bytes. The PlayStation, marketed as "32-bit" in the same era, also used 8-bit bytes. Same byte. Different processor generation. Different marketing.
The lesson: when someone says "N-bit," ask bits of what? Processor width, bus width, byte size, and brand name are not the same conversation. Marketing is often born from the biggest number that fits on the box.
Mental model
A byte has eight bits because that's the definition.
A byte has eight bits because eight turned out to be a practical engineering standard.
Three sentences to keep
Nothing in physics says a byte must contain eight bits.
Eight bits didn't become a standard because they were mathematically special. They became a standard because they were a good engineering compromise.
Standards are often born from good trade-offs, not perfect ideas.
Further reading
- Why Binary Won? Essay 1 in this series.
- Think in Binary Essay 2 in this series.
- Charles Petzold, Code. Chapter 12 on bytes and hexadecimal notation.
What's next
We now know why computers group bits into bytes. The Binary Dump trilogy ends here: physics, cognition, convention.
Hexadecimal is next. After that: what happens when the highest bit stops being weight and becomes sign. But first, the notation every tool uses to hide the ones and zeros.