# CPU Architectures, x86 and ARM

> Why a program built for an Intel chip will not run on an ARM one. What x86-64 and ARM are each good at, and what 32-bit and 64-bit mean.

**Course:** [Operating Systems and Files](https://learn.benibix.com/learn/development-platforms)  
**Time:** 12 min

**After this lesson you can:**

- Explain what a CPU architecture is
- Tell x86-64 from ARM and say where each is used
- Explain the 4 GB limit of a 32-bit program

## Why code built for x86 will not run on ARM

A CPU architecture is the set of basic orders a chip knows, like add two numbers or fetch a value. A program has to be turned into these orders before it runs. Two families run nearly every device today. x86-64 is in most Windows PCs and servers. ARM is in almost every phone.

Chips in the same family run the same machine code. Chips in different families do not. Hand an ARM chip a program built for x86-64 and it cannot read a single order. That is why a download page asks you to pick x64 or ARM64, and why the wrong pick gives you a file that will not open.

x86-64 grew out of Intel's older 32-bit x86 chips. AMD made the 64-bit version first, in chips sold from 2003, and Intel took it up too. You will see it written as x64, AMD64 or x86_64, and all three mean the same thing. It still runs most old 32-bit programs, which is a big part of why it lasted.

ARM was built to use little power, which is why phones and tablets run on it. The firm Arm designs the plans and sells the right to use them, and others build the real chips. Since 2020, Apple's Macs have used ARM chips of Apple's own design. Some Windows laptops and many cloud servers use ARM too.

The 32 or the 64 is the size of the numbers a chip works with, counted in bits. A memory address is one of those numbers. With 32 bits there are only about 4.3 billion addresses, so a 32-bit program can use at most 4 GB of memory. A 64-bit program can reach far more than any computer holds.

Moving to a new family does not always mean waiting for new builds. Apple ships a translator that lets x86-64 apps run on ARM Macs, and Windows on ARM has one too. The app works, but it is slower and uses more battery than a build made for the chip. A translator is a bridge, not a home.

People often think ARM chips are slow phone chips. That stopped being true years ago. Apple's ARM Macs match or beat x86-64 laptops at many jobs while using far less power. The real rule is simpler. Neither family is best at everything, and code has to be built for the one it runs on.

## Which one fits

1. **x86-64.** Fits when you need old Windows programs or games. It also fits a server that runs old software as it is.
2. **ARM.** Fits when battery life matters most, as on phones and thin laptops. It also fits cheaper cloud servers.

![Two halves. On the left, Your code splits into an x86-64 build for Intel and AMD chips and an ARM build for phones and Apple chips, and each build runs only on its own kind of chip. On the right, two bars show the memory a program can reach: a tiny 32-bit bar marked 4 GB at most, and a long 64-bit bar marked about 17 billion GB.](https://learn.benibix.com/diagrams/dev-cpu-arch.svg)

### Example: Jack rents a cheaper server

Jack's web app costs $40 a month on an x86-64 cloud server. An ARM server of the same size costs $32.

His app is written in Python, so most of the code needs no change. He rebuilds one part that was made for x86-64, tests it for a day, and saves $96 a year.

**Words in this lesson:** [CPU Architecture](https://learn.benibix.com/term/development/cpu-architecture), [x86-64](https://learn.benibix.com/term/development/x86-64), [ARM Architecture](https://learn.benibix.com/term/development/arm-architecture), [32-bit and 64-bit](https://learn.benibix.com/term/development/32-bit-and-64-bit)

> **The one thing to remember.** x86-64 and ARM speak different machine code. A program must be built for one or the other, or run through a slower translator.

**Question:** Why can a 32-bit program use only 4 GB of memory?

**Answer:** Its addresses are 32 bits long, so there are only about 4.3 billion of them. Every byte of memory needs its own address. A 32-bit number can only count to about 4.3 billion, which is 4 GB of bytes, so the program cannot point past that.

---

Source: https://learn.benibix.com/learn/development-platforms/cpu-architectures-x86-and-arm
Author: Behnia Zarei (a developer in Vienna, Austria, who builds and teaches data, AI and business intelligence at BENIBIX)
Publisher: BENIBIX Learn (https://learn.benibix.com/)
Last updated: 2026-11-19
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Quote freely with attribution.
