1.2 Historical Development & the Laws of Moore and Rock

Theory: Historical Development & the Laws of Moore and Rock

To fully appreciate today’s computers, it helps to understand how things got the way they are. Computing machinery evolved over several centuries, and in modern times its history is usually divided into generations according to the dominant (salient) technology of the era (Null & Lobur, §1.5). The dates are approximate and the generations overlap.

The key idea to remember: each generation is defined by its switching technology, and each new technology made computers smaller, faster, cheaper and more reliable than the last.

Diagram loads as you scroll.

Generation Zero: Mechanical Calculating Machines (1642–1945)

Machines built from gears, wheels and levers, with no electronics.

Person	Machine	Significance
Wilhelm Schickard (1592–1635)	Calculating Clock	Early mechanical calculator (add, subtract)
Blaise Pascal (1623–1662)	Pascaline	The first widely known adding machine (add and subtract with carries)
Charles Babbage (1791–1871)	Difference Engine; designed (never built) the Analytical Engine	Difference Engine computed polynomial tables. The Analytical Engine had a “mill” (processor), a “store” (memory) and punched-card input: the blueprint of a general-purpose computer. Ada Lovelace wrote programs for it
Herman Hollerith (1860–1929)	Punched-card tabulating machines	Used to process the 1890 US census; his company later became IBM

Mini-example

“A machine with a mill and a store, programmed by punched cards, never actually built” → Babbage’s Analytical Engine, generation 0.

First Generation: Vacuum Tube Computers (1945–1953)

Technology: the vacuum tube (an electronic switch/amplifier). Machines were huge, used enormous power, produced lots of heat and tubes burned out often.

ENIAC (Electronic Numerical Integrator and Computer), John Mauchly and J. Presper Eckert, University of Pennsylvania, 1946. Known as the first all-electronic, general-purpose computer: about 17,468 vacuum tubes, about 30 tons, roughly 1,800 square feet of floor space, about 174 kW of power. It was programmed by moving plugs and wires (see section 1.3).
UNIVAC I (1951): the first commercially sold computer in the US, built by Mauchly and Eckert’s company.
IBM 650 (1955): the first mass-produced computer (hundreds sold). It is a vacuum-tube machine, so it belongs to this generation even though it appeared slightly after 1953.
Context: during World War II, Alan Turing and colleagues at Bletchley Park built the electromechanical Bombe to break the German Enigma cipher, and the British built Colossus to attack the Lorenz cipher.

Second Generation: Transistorized Computers (1954–1965)

Technology: the transistor, invented at Bell Laboratories (AT&T) in 1947 (Bardeen, Brattain and Shockley), first made from germanium.

Switches between 0 and 1 in microseconds.
Compared with vacuum tubes: about 10× faster, 1/10 the power, 1/10 the space, and far more reliable.
Magnetic core memory used for the first time to build large random access memories (RAM).
High-level languages appear: FORTRAN, COBOL, ALGOL, SNOBOL.
Machines: IBM 7000 series (e.g. IBM 7094), DEC PDP-1, CDC 6600 (the first supercomputer, designed by Seymour Cray).
Computer science begins as an academic discipline.

Third Generation: Integrated Circuit Computers (1965–1980)

Technology: the integrated circuit (IC): many transistors made together on one small piece of silicon (invented independently by Jack Kilby at Texas Instruments and Robert Noyce at Fairchild, 1958–59).

About 10× better than the second generation; CPUs reach about 1 million instructions per second (1 MIPS).
Online (interactive) systems become possible, e.g. dynamic production control systems and airline reservation systems. Time-sharing and multiprogramming operating systems appear.
Industry led by IBM and the “BUNCH”: Burroughs, Univac, NCR, Control Data and Honeywell.
Machines: IBM System/360 (the first family of computers sharing one architecture, from small to large models: the organization/architecture idea of 1.1 in action), DEC PDP-8 and PDP-11 (minicomputers), and the Cray-1 vector supercomputer (1976).
Dennis Ritchie (creator of C) and Ken Thompson rewrote UNIX in C for the DEC PDP-11 (1973), making the operating system portable to other machines.

Fourth Generation: VLSI Computers (1980–present)

Technology: Very Large Scale Integration (VLSI): more than 10,000 components per chip, making it possible to put an entire CPU on one chip, the microprocessor.

Integration scale	Components per chip
SSI (small)	10 – 100
MSI (medium)	100 – 1,000
LSI (large)	1,000 – 10,000
VLSI (very large)	more than 10,000
ULSI (ultra large)	more than 1,000,000

The first microprocessor was the 4-bit Intel 4004 (1971, about 2,300 transistors).
Later chips, the 8080 and the 808x family (8086/8088, used in the 1981 IBM PC), created personal computing (PC).
A variety of parallel architectures developed, building on vector supercomputers such as the Cray-1: multiprocessors, massively parallel machines, clusters.
UNIX and C spread across workstations and PCs.

Fifth Generation: Modern Computers with ULSI (2000–present)

Today’s chips use ULSI (Ultra Large Scale Integration, billions of transistors). Because this is still microelectronic (integrated-circuit) technology, some people believe we are still in the 4th generation. Those who call it a new generation point to:

Multicore processors and systems-on-chip (CPU + GPU + memory controller + AI engine on one chip, e.g. Apple M-series).
GPUs and AI accelerators (TPUs, NPUs) for artificial intelligence and machine learning (neural networks).
Mobile, cloud and massively parallel computing.

Mini-example

“Uses magnetic core memory and is programmed in FORTRAN” → 2nd generation. “Has a single-chip 4-bit CPU” → 4th generation (Intel 4004 microprocessor).

Moore’s Law

Gordon Moore (Intel co-founder), 1965:

“The density of transistors in an integrated circuit will double every year.”

The contemporary version: “The density of silicon chips doubles every 18 months.” (Moore himself revised his estimate to every two years in 1975.)

Doubling is exponential growth. If $N_0$ is today’s count and the doubling period is $d$ , then after time $t$ :

N(t) = N_0 \times 2^{t/d}

Steps for any Moore’s-law question:

Find the number of doublings $k = t / d$ (time elapsed ÷ doubling period, in the same units).
Multiply by $2^k$ .
To go backwards (how long to grow by a factor $F$ ?), solve $2^k = F$ , i.e. $k = \log_2 F$ , then $t = k \times d$ .

But this “law” cannot hold forever: transistors cannot shrink below the size of a few atoms, and heat and quantum effects become serious at very small sizes.

Mini-example

A chip has 500 million transistors. With an 18-month doubling period, after 3 years: $k = 36 / 18 = 2$ doublings → $500 \times 2^2 = 2{,}000$ million $= 2$ billion transistors.

Rock’s Law

Arthur Rock (Intel financier):

“The cost of capital equipment to build semiconductors will double every four years.”

In 1968, a new chip plant cost about USD 12,000. At the time, USD 12,000 would buy a nice home in the suburbs, and an executive earning USD 12,000 per year was “making a very comfortable living”.
In 2012, chip plants under construction cost well over USD 5 billion, more than the gross domestic product of some small countries (e.g. Barbados, Mauritania, Rwanda).

Moore vs. Rock. Moore’s law says chips keep getting denser; Rock’s law says the factories needed to make those denser chips keep getting more expensive. For Moore’s law to hold, Rock’s law must fail, or vice versa: eventually nobody can afford the plant needed for the next doubling. No one can say which will give out first.

Mini-example

Rock’s law over 8 years: $8 / 4 = 2$ doublings → cost × $2^2 = 4$ . A USD 1 billion plant would become a USD 4 billion plant.

Worked Examples

Exercise 1

Identify the generation (0–5) of each machine or description:

(a) ENIAC, with about 17,000 vacuum tubes. (b) A machine using magnetic core RAM, programmed in COBOL, built from germanium transistors. (c) The IBM System/360 family built from integrated circuits. (d) A PC built around an Intel 8088 microprocessor. (e) The Pascaline. (f) A smartphone chip with 8 CPU cores, a GPU and a neural engine, with about 15 billion transistors.

Show solution ↓

Solution

Look for the switching technology in each description.

	Technology clue	Generation
(a)	Vacuum tubes	1st (1945–1953)
(b)	Transistors, core memory, COBOL	2nd (1954–1965)
(c)	Integrated circuits	3rd (1965–1980)
(d)	Single-chip microprocessor (VLSI)	4th (1980–)
(e)	Mechanical gears (Pascal)	0 (1642–1945)
(f)	ULSI, multicore, AI hardware	5th (2000–)

Exercise 2

A chip today contains 2 billion transistors. Using the contemporary form of Moore’s law (doubling every 18 months), predict the transistor count after (a) 3 years, (b) 6 years, (c) 9 years.

Show solution ↓

Solution

Convert years to months and count doublings $k = t / 18$ .

Time	Months	$k$	Factor $2^k$	Transistors
(a) 3 years	36	$36/18 = 2$	4	$2 \times 4 = 8$ billion
(b) 6 years	72	$72/18 = 4$	16	$2 \times 16 = 32$ billion
(c) 9 years	108	$108/18 = 6$	64	$2 \times 64 = 128$ billion

Check (c) step by step: 2 → 4 → 8 → 16 → 32 → 64 → 128 billion (6 doublings) ✓.

Exercise 3

In 1965 the most complex chips held about 64 components. Using Moore’s original law (doubling every year), how many components did that predict for 1975? Compare with the 18-month version over the same period.

Show solution ↓

Solution

Original law: $1975 - 1965 = 10$ years → $k = 10$ doublings.

64 \times 2^{10} = 2^6 \times 2^{10} = 2^{16} = 65{,}536\ \text{components}

(Moore’s famous 1965 prediction was indeed “about 65,000 components per chip by 1975”.)

18-month version: 10 years $= 120$ months → $k = 120 / 18 = 6.67$ doublings.

64 \times 2^{6.67} \approx 64 \times 101.6 \approx 6{,}500\ \text{components}

The doubling period matters enormously: over only 10 years the two versions differ by a factor of about 10.

Exercise 4

How many years does it take for transistor density to increase 1,000-fold if it doubles every 18 months? What if it doubles every 2 years?

Show solution ↓

Solution

Solve $2^k = 1000$ :

k = \log_2 1000 = \frac{\ln 1000}{\ln 2} = \frac{6.908}{0.693} \approx 9.97 \approx 10\ \text{doublings}

(Quick check: $2^{10} = 1024 \approx 1000$ ✓.)

18-month doubling: $t = 10 \times 1.5 = 15$ years.
2-year doubling: $t = 10 \times 2 = 20$ years.

Handy rule: 10 doublings ≈ ×1000.

Exercise 5

A chip plant cost about USD 12,000 in 1968.

(a) What does Rock’s law (doubling every 4 years) predict for 2012? (b) The actual cost in 2012 was about USD 5 billion. Roughly what doubling period does the actual growth correspond to?

Show solution ↓

Solution

(a) $2012 - 1968 = 44$ years → $k = 44 / 4 = 11$ doublings:

12{,}000 \times 2^{11} = 12{,}000 \times 2{,}048 = 24{,}576{,}000 \approx \text{USD } 24.6\ \text{million}

(b) Growth factor:

F = \frac{5 \times 10^9}{12{,}000} \approx 416{,}667

Number of doublings: $k = \log_2 416{,}667 \approx 18.7$ (since $2^{18} = 262{,}144$ and $2^{19} = 524{,}288$ ).

Doubling period $= 44 / 18.7 \approx 2.4$ years.

So plant cost actually grew much faster than Rock’s 4-year rule, doubling roughly every 2.4 years. This is the economic pressure that threatens Moore’s law.

Exercise 6

Explain in a few sentences why “for Moore’s law to hold, Rock’s law must fall, or vice versa”.

Show solution ↓

Solution

Moore’s law says chip density keeps doubling (every 1–2 years). Each doubling needs smaller features and therefore newer, more precise and more expensive fabrication equipment.
Rock’s law says the cost of that equipment doubles every 4 years: exponential growth in the cost of building chip plants.
Both are exponential. If density keeps doubling, plant cost keeps doubling too, and eventually a plant costs more than any company (or country) can recover by selling chips. At that point the next density doubling will not be built, so Moore’s law stops.
The only way for Moore’s law to continue indefinitely is if manufacturing costs stop doubling (Rock’s law fails), e.g. through new cheaper manufacturing methods.

Hence the two “laws” cannot both continue forever; one must give out first, and no one knows which.