Time and Frequency for Computer Architecture

Types of computers

Digital vs. analog computer systems: we will only discuss digital systems (systems where signals take on only a discrete number of states, rather than a continuous range of values) Binary nature of essentially all hardware components: switches which are open or closed, bistable memory elements (which hold a single bit of information)
Electronic systems (as contrasted to mechanical or fluidic computers): information represented by electical signals

• define correspondence between 0/1 and particular values of voltage or current
• basic element is a switch: device whereby one electrical signal can be used to control another
• switches can be used to perform logical functions (and, or, not)

Time and frequency

Time is measured in seconds and in fractions of a second:

millisecond (ms): 0.001 second

microsecond (us): 0.000 001 second

nanosecond (ns): 0.000 000 001 second

picoseconds (ps): 0.000 000 000 001 second

For a repetitive phenomenon, the rate at which it repeats is the frequency, measured in cycles per second or Hertz. For higher frequencies we have

1 kilohertz (kHz) = 1 000 cycles / second

1 megahertz (MHz) = 1 000 000 cycles / second

1 gigahertz (GHz) = 1 000 000 000 cycles / second

There is a reciprocal relationship between frequency and the time for a cycle (the 'period'):

frequency = 1 / clock period

clock period = 1 / frequency

For example, if a CPU operates at 100 Hz, its "clock cycle" is 0.01 second = 10 ms; if it operates at 100 MHz, its clock cycle is 0.000 000 01 second = 10 ns.

Switching elements and computer generations

(text, sec. 1.10 (on CD)) Relays were used in the earliest equipment: electronic accounting machines, and early computers made in the 1930's and early 1940's at Bell Labs and by Zuse in Germany. Relays are mechanical and hence relatively slow --- their response time is measured in milliseconds.
Vacuum tubes were dominant about 1945 to 1960 ('first generation of computers'). The first general purpose computer was the ENIAC (1946): 18,000 tubes, 20 10-digit accumulators, 100 kHz clock, 200 microsecond add time, 200 KW power. Tubes were much faster than relays (because they had no mechanical moving parts), but they were bulky, required high power, and had a relatively short life (crucial because of the large number of components).
Discrete transistors ('second generation computers') were used in computers starting around 1960. They took less power , were smaller, and had a longer lifetime than vacuum tubes.
Integrated circuits ('third generation computers') were created by fabricating several transistors on a single chip, allowing computers to be made smaller. IC's were introduced in the late 1960's, and gradually increased their level of integration (number of transistors on a chip).
Very large scale integration ('fourth generation computers') represented the ability to put more and more transistors on a chip, until an entire processor (a microprocessor) could be fabricated on a single chip. Initially VLSI was used to make personal computers (Apple II - 1977; IBM PC - 1981); now all computers are made from VLSI.

Technology Trends

We continue to learn, at a steady pace, how to fabricate smaller and smaller transistors. This allows for

• a higher level of integration (currently, hundreds of millions of gates on a chip)
• (until recently) faster circuits (compare the initial PC of 5 Mhz with current PCs of 2-3 GHz)

The cost of a chip has remained (very roughly) constant, so price/performance has been rapidly decreasing. How should this extra circuitry and extra performance be used?