CPU

A CPU runs a simple cycle billions of times per second: fetch the next instruction from memory, decode it, execute it, for example by adding two numbers or jumping to another instruction, and store the result. Its clock speed, measured in gigahertz, is how many cycles it completes per second, but modern CPUs also run several instructions at once and predict which way a program will branch.

Today's processors contain several cores, each a complete CPU able to run its own thread, and many cores can run two threads at once with simultaneous multithreading. Small, very fast caches (L1, L2 and L3) sit between the cores and the much slower main memory, because waiting for RAM would otherwise waste most of the CPU's time.

The instructions a CPU understands are defined by its architecture. x86-64, from Intel and AMD, dominates PCs and servers; ARM dominates phones and is now common in laptops, such as Apple's M-series chips, and in cloud servers; RISC-V is an open architecture growing in embedded devices. Compiled programs must be built for the architecture they run on.

A common misconception is that a higher clock speed always means a faster computer. Performance also depends on how much work each cycle does, the number of cores, cache sizes and memory speed, and many programs can only use one core at a time, so a chip with fewer but faster cores can beat one with many slower cores.