The fetch-decode-execute cycle
A CPU operates in a tight loop: Fetch instruction from memory, Decode what it means, Execute the operation, Write results back. Modern CPUs do billions of cycles per second.
Instruction sets
x86-64 (Intel/AMD) dominates desktops and servers. ARM (Apple Silicon, Qualcomm) dominates mobile. RISC-V is an open-source alternative growing in embedded and AI accelerators.
Multiple cores
A core is one complete execution unit. Modern CPUs have 8 to 128 cores. Multithreaded software splits work across cores to increase throughput dramatically.
Caches
DRAM is slow relative to CPU speed. L1, L2, and L3 caches are fast SRAM between CPU and main memory. Cache-friendly software accesses memory in predictable sequential patterns.
Beyond Moore's Law
Transistor density scaling has slowed. The industry pivoted to specialised chips: GPUs for ML, TPUs for inference, NPUs in phones, and chiplet packaging for larger effective dies.
