| | {{infobox |
| | | title = Z3 |
| | | image = zuse-z3-archive.jpg |
| | | caption = The Z3 as drawn in Zuse's archive: "first working program-controlled computing device — relay technology / binary / floating point (completed 1941)" |
| | | designer = [[Konrad Zuse]] |
| | | completed = May 1941 — Berlin-Kreuzberg |
| | | destroyed = December 1943, air raid — replica at Deutsches Museum, Munich |
| | | technology = ~2,600 telephone relays |
| | | memory = 64 words × 22 bits — binary floating point |
| | | program = punched 35 mm film, read sequentially |
| | | speed = ~5–10 Hz · add ~0.8 s · multiply ~3 s |
| | }} |
| | |
| | The **Z3** — the third of Zuse's machines, after the mechanical **[[Z1]]** and the relay-hybrid **[[Z2]]** — was the machine [[Konrad Zuse]] demonstrated in Berlin on **12 May 1941** — and the machine that quietly answers the question "when was the computer invented?" Built from roughly 2,600 discarded telephone relays in a Kreuzberg living room, it was **binary**, **programmable**, **floating-point**, and **fully automatic**: punch the program onto film, press start, and the relays clicked through the work alone. It predates [[Colossus]] by two years and ENIAC by four. Most histories give the Z3 the title it earned: **the first working, programmable, fully automatic digital computer.** |
| | |
| | ```fact The day computing switched on |
| | On 12 May 1941, engineers from the German Aeronautics Research Institute watched the Z3 run calculations in Zuse's parents' apartment. It was the first time a machine had executed a stored instruction sequence, entirely on its own, from beginning to end — the actual birthdate of the working computer, and almost nobody noticed. |
| | ``` |
| | |
| | ## What was inside |
| | |
| | [[File:z3-drawing.svg|right|340px|The Z3 laid out: memory cabinet (2,000 relays), arithmetic unit (600 relays), console, and the film reader — CC0]] |
| | |
| | The Z3 was not a calculator wired to repeat; it was a real architecture, divided the way computers are still divided: |
| | |
| | - **The memory** — about 2,000 relays holding **64 words of 22 bits** each, addressable and random-access. Numbers were **binary floating-point**: a sign bit, a 7-bit exponent, and a 14-bit mantissa (15 effective bits, the leading 1 implied). Special values for zero and infinity were built in — the machine could report *"this overflowed"* instead of lying. |
| | - **The arithmetic unit** — about 600 relays, a two-stage floating-point engine doing **add, subtract, multiply, divide, and square root**, plus automatic conversion between the operator's decimal and the machine's binary. |
| | - **The program reader** — instructions punched as holes in recycled **35 mm film stock**, eight operations encoded per line, read sequentially. To loop a program, you literally glued the film into a loop. |
| | - **The console** — a small keyboard for decimal input and a lamp panel for decimal output: the human spoke decimal, the machine thought binary, and the hardware translated both ways. |
| | |
| | ```chart Data through the Z3 |
| | input film "film program" |
| | process ctrl "control" |
| | process alu "arithmetic\n600 relays" |
| | process mem "memory\n2,000 relays" |
| | output out "decimal\nlamps" |
| | |
| | film -> ctrl -> alu |
| | alu -store-> mem |
| | alu -show-> out |
| | ``` |
| | |
| | ## What it could do — and what it couldn't |
| | |
| | In operation the Z3 was honest arithmetic at relay speed: an addition in well under a second, a multiplication in about three seconds, the whole clock ticking at somewhere between five and ten hertz. It worked real problems — wing-flutter and structural calculations for aviation researchers, the very statics Zuse had been too lazy to keep doing by hand. |
| | |
| | Its limitation was fundamental and simple: **it could not decide.** The film program ran straight through, instruction after instruction, with no conditional branch — the machine could loop only if its tape was glued into a circle, and it could not alter its course based on a result. Every program was one fixed sequence. |
| | |
| | ```note Could the Z3 compute anything? |
| | In 1998, computer scientist **Raúl Rojas** proved a surprising result: despite having no conditional jump, the Z3's design is **Turing-complete** in principle — looping the tape and exploiting its arithmetic lets it simulate any computation, given unlimited tape and time. What Zuse built in a living room was not a toy version of the idea. It was the idea. |
| | ``` |
| | |
| | ## Three years, then fire |
| | |
| | The Z3 did its wartime work in obscurity. The Reich's institutions never grasped what they had — Zuse's requests for serious funding were dismissed, in his account, by officials who believed the war could be won without computers. The machine clicked away at aerodynamic matrices while the country that owned it decided not to want it. |
| | |
| | In December 1943, Allied bombs destroyed the Methfesselstraße workshop, and the Z3 with it — two and a half years after switching on, the first working computer became scrap. Its successor, the **[[Z4]]**, survived only because Zuse evacuated it south, eventually to a barn in the Allgäu, where it waited out the end of the war. |
| | |
| | ## Afterlife |
| | |
| | - **1960** — Zuse KG built a working replica of the Z3 for the **Deutsches Museum** in Munich, where it still stands — and has been demonstrated, clicking through real programs. |
| | - **1987–89** — a nearly eighty-year-old Zuse personally supervised the reconstruction of his earlier Z1 in Berlin. |
| | - **1998** — Rojas's Turing-completeness proof settled the theoretical question the skeptics had left open. |
| | - The original film programs and fragments of documentation survive; the Z3 is reconstructed in scholarship, in replicas, and in the architecture of every machine since. |
| | |
| | [[File:z3-relay-finder.jpg|left|300px|A relay of the type the Z3 was built from — the entire machine was ~2,600 of these clicking switches — CC0]] |
| | |
| | ## Where it sits in the line |
| | |
| | ```fact The 1940s race, by the clock |
| | **Z3** 1941 — binary, programmable, automatic · **Atanasoff–Berry** 1942 — electronic, not programmable · **[[Colossus]]** 1943 — electronic, programmable, special-purpose, secret · **Mark I** 1944 — automatic, decimal · **[[ENIAC six|ENIAC]]** 1945 — electronic, general, rewired to program · **[[Manchester Baby]]** 1948 — first stored program in memory. |
| | ``` |
| | |
| | The Z3 was electromechanical where ENIAC would be electronic, sequential where the [[Manchester Baby|Baby]] would be stored-program — but it ran first, and it ran alone. Between Babbage's unfinished dream and the electronic machines of 1945, the bridge was a film tape and 2,600 relays in an apartment that no longer exists. |
| | |
| | ## See also |
| | |
| | - [[Konrad Zuse]] — the man in the living room |
| | - [[Plankalkül]] — the language he designed while the machine burned |
| | - [[Analytical Engine]] · [[Charles Babbage]] — the design Zuse unknowingly reinvented |
| | - [[Colossus]] · [[ENIAC six]] · [[Manchester Baby]] · [[EDSAC]] · [[EDVAC report]] |
| | - [[Alan Turing]] · [[History of Programming]] |
| | - [[History of Electricity]] — the relay's story |
| | |
| + | ## References |
| + | |
| + | 1. Zuse, Konrad. *The Computer — My Life* (Springer, 1993, tr. of *Der Computer – Mein Lebenswerk*, 1970/1984). |
| + | 2. Rojas, Raúl. "Konrad Zuse's Legacy: The Architecture of the Z1 and Z3." *IEEE Annals of the History of Computing* 19(2), 1997. |
| + | 3. Rojas, Raúl, and Ulf Hashagen, eds. *The First Computers: History and Architectures* (MIT Press, 2000). |
| + | |
| | [[Category:Hardware]] |
| | [[Category:History]] |
| | |