Turning Learners Into Developers
Codekilla
CODEKILLA
// computer.fundamentals · Chapter 04 / 20

Five Generations of Computers

Vacuum tube → Transistor → IC → Microprocessor → AI: how size, speed and cost changed

1st genVacuum tube · 19462nd genTransistor · 19563rd genIC chip · 19644th genMicroprocessor · 19715th genAI · Quantum · nowsize of one switch (not to scale) ↓ speed ↑ 10³ → 10¹⁸ ops/s
The switching element shrank from a bulb-sized vacuum tube to a transistor a few atoms wide — and speed went up a billion times.
GenerationYearsMain technologySize & powerSpeedLanguages / SoftwareExamples
1st1940–1956Vacuum tubes; magnetic drum memory; punched cardsRoom-sized; 10–150 kW; huge heat; tonnesmilliseconds (10⁻³ s), ~1,000–5,000 ops/sMachine language (binary)ENIAC, EDVAC, UNIVAC I, IBM 701, IBM 650
2nd1956–1963Transistors; magnetic core memory; tapeCupboard-sized; a few kW; less heatmicroseconds (10⁻⁶ s), ~10⁵ ops/sAssembly, FORTRAN, COBOL, ALGOL; batch OSIBM 1401, IBM 7090, CDC 1604, UNIVAC 1108, PDP-1
3rd1964–1971Integrated Circuits (SSI/MSI); keyboards & monitors; disksDesk-sized minicomputers; hundreds of Wnanoseconds (10⁻⁹ s), ~10⁶–10⁷ ops/sHigh-level languages (BASIC, Pascal, C); multiprogramming, time-sharing OSIBM System/360, PDP-8, PDP-11, Honeywell 6000
4th1971–presentMicroprocessor (LSI → VLSI → ULSI); semiconductor RAM; GUI; networks & internetDesktop → laptop → phone → watch; 5–500 Wpicoseconds (10⁻¹² s), 10⁹–10¹² ops/sC++, Java, Python; Windows, macOS, Linux, Android, iOSIntel 4004, Altair 8800, Apple II, IBM PC, Macintosh, Pentium, Core i9, Ryzen, Apple M4
5thPresent → futureAI, parallel processing, GPUs/NPUs/TPUs, quantum & neuromorphic chips, nanotechnologyCloud data centres (MW) + tiny edge devices (mW)exaFLOPS (10¹⁸) in supercomputers; TOPS in phonesNatural language (prompting), ML frameworks, LLMs (GPT, Gemini, Claude)Frontier, El Capitan, Nvidia Blackwell, Google Willow, ChatGPT, Copilot+ PCs, robots
  • ISwitching device shrank: vacuum tube (size of a bulb, ~2 W each) → transistor (pea-sized) → IC (thousands per chip) → VLSI (millions) → today 3 nm transistors (about 15 silicon atoms wide).
  • IIMemory: magnetic drum → magnetic core → semiconductor RAM → DDR5 (up to 8,400 MT/s) → HBM3e for AI.
  • IIIInput/Output: punched cards & plugboards → teletype → keyboard + CRT → mouse + GUI → touch → voice → gesture/brain interface.
  • IVCost per calculation has fallen by more than a trillion times since 1946.
// Keep exploring
All chapters →