CS305 Computer ArchitectureAutumn 2010

Lecture 02

Bhaskaran RamanDepartment of CSE, IIT Bombay

http://www.cse.iitb.ac.in/~br/

http://www.cse.iitb.ac.in/synerg/doku.php?id=public:courses:cs305­autumn10:start

Types of Computers

Q: What are the different computers you have seen or used or encountered or heard/read about?

Types of Computers: Desktops● Desktops or PCs: most well known class

Types of Computers: Servers● Servers: most powerful: individual PCs to supercomputers

Types of Computers:Embedded Computers

● Embedded computers: most specialized, vary in compute power, most prevalent but least noticed/known● Cell­phones, robots, sensors, TV, fridge, washing 

machine, etc.

Some Terms

● Processor or microprocessor: the main part of a computer

● Terms to be explained in subsequent lectures:● Instruction set architecture● n­bit architecture

32/64­bit Processors Sold

1998 1999 2000 2001 20020

200

400

600

800

1000

1200

The number of distinct processors sold between 1998 and 2002

Embedded Computer Desktops Servers

Mill

ions

of

com

pute

rs

Sale of Different MicroprocessorsSource: course text, Fig. 1.2

Source: course text, Fig 1.17

Growth of Processing Power

What is this Course About?

● What are the principles underlying the design of these various microprocessors?

● What support does the microprocessor provide for running applications?

● What are the components alongside a microprocessor?

● How to measure performance, and optimize for speed?

A Hierarchy of Languages

Algorithm

High­Level Language (HLL)

Assembly Language

Machine Language

Digital Logic

Natural language, with some symbols

Concise; Structured, yet human readable

Not concise; Human readable representation of machine language

Binary numbers (0's and 1's)

Electrical signals

Mac

hin e

 spe

cifi

cM

achi

n e 

inde

pen d

ent

Programmer

Compiler

Assembler

Computer designer

An ExampleC code:a = b + c;d = e + f;

Machine code:...0...1......0...1......0...1......0...1......0...1......0...1......0...1......0...1...

Assembly code:lw $s1, 4($s0)lw $s2, 8($s0)add $s3, $s1, $s2sw ($s0), $s3lw $s3, 16($s0)lw $s4, 20($s0)add $s5, $s3, $s4sw 12($s0), $s5

Von Neumann Architecture

Memory=

Program+

Data

Controlpath

Datapath

Proc

esso

r

Input

Output

● The stored program concept: program (instructions) as well as data are stored in memory

● Processor fetches instructions from memory, and executes them on data (also fetched from/to memory)● Example from previous slide: LW and SW instructions

The Five Components● All computers have these five components: input, 

output, memory, [data path + control path = processor]

● Underlined aspects: topics in this course

● Input: keyboard, mouse; also disk, network● Output: monitor; also disk, network● Memory: different kinds of memory● Data path + control path = processor

Inside a Computer...

Inside a Computer (continued...)

Magnetic Tape

Parts of a Computer● Integrated circuits, or chips:

● Flat and black● Processor (CPU), main memory, cache memory, etc.

● Motherboard:● Houses the various chips● Also has many I/O interfaces (PCI, USB, Serial, etc.)

● Secondary memory: non­volatile● Magnetic disks, optical (CD/DVD), tape, flash­based 

(e.g. USB pen­drives, CF cards), floppies (obsolete)

Inside a Processor

Source: course text, Fig. 1.9

The IC Revolution behindProcessors and Memory

● IC's or Integrated Circuits● Precursors: transistors, vaccum tubes● Today: VLSI, ULSI

The IC Revolution in Numbers

Year Technology used in computers

Relative performance / unit cost

1951 Vacuum tube 11965 Transistor 351975 Integrated circuit 9001995 Very large scale integrated 

circuit2400000

2005 Ultra large scale integrated circuit

6200000000

Source: course text, Chapter-1

Growth of Memory Capacity

1976 1981 1986 1991 1996 20011

10

100

1000

10000

100000

1000000

Growth of capacity per DRAM chip over time

Year of introduction

k bi

t c a

paci

ty (

k=2^

10=

1024

)

The IC Manufacturing Process

Source: course text, Fig. 1.14

IC CostCost per die=

Cost per waferDies per wafer×yield

Dies per wafer≈Area of waferArea of die

Yield=1

[1Defects per area×Area of die /2]2

Straightforward algebra

Approximation

From experience

● Unit cost of chip decreases with volume● Fixed costs amortised: design, masks in chip manufacture● Tuning to improve yield

Limits to IC density

● Fundamental physical dimension limits● Power consumption

● ~ 1 cm2 Pentium: 82 Watts!● Fan needed to sink the heat

● Observe: frequency scaling (Linux PC)

Upcoming topics

● Instruction set architecture● Assembly language programming