computer architectures cs401 sabanci university erkays@sabanciuniv

04/20/23 Erkay Savas 1

Computer Architectures

CS401

Sabanci Universityerkays@sabanciuniv.edu

04/20/23 Erkay Savas 2

Motivation• The IEEE/ACM Computer Curricula 2001, prepared by the Joint

Task Force on Computing Curricula of IEEE Computer Society and ACM lists computer architecture as one of the core subjects that should be in the curriculum of all students in computer science and engineering.

• The computer lies at the heart of the computing. Without it most of the computing disciplines today would be a branch of theoretical mathematics. To be a professional in any field of computing today, one should NOT regard the computer just a black box that executes program by magic. All students of computing should acquire some understanding and appreciation of a computer system’s functional components, their characteristics, their performance and their interactions. There are practical implications as well. Students need to understand computer architecture in order to structure a program so that it runs more efficiently on a real machine. In selecting a system to use, they should to able to understand the tradeoff among various components, such as CPU clock speed vs. memory size.

04/20/23 Erkay Savas 3

Instructor• Dr. Erkay Savaş• Office: FENS 1098• e-mail: erkays@sabanciuniv.edu• url: http://people.sabanciuniv.edu/~erkays• class webpage:

http://people.sabanciuniv.edu/~erkays/cs401/cs401.html

• Phone: x9606 (not a preferred way of communication)– Use sucourse

• Office hour: Tuesday (09:40-11:30)– or by appointment

04/20/23 Erkay Savas 4

Instruction

• Midterm: 30%– After the term break– In lab/recitation session

• Final: 40%• No official lab sessions

– homework assignments will do

• Homework assignments: 4-5 per term– 15%

• Project: 10%• Participation & attendance: 5%

– Quiz & in-class questions

04/20/23 Erkay Savas 5

ENIAC

world’s first general-purpose computer

04/20/23 Erkay Savas 6

ENIAC by Eckert & Mauchly

• Electronic Numerical Integrator And Calculator

• Technology: Vacuum tubes (19000+)• Cost like 300 million USD.• The main components

– 20 registers (each 10-digit wide, 60 cm long), addition/subtraction and temporary storage

– Initiator: powering up/shutting down the computer, starting computation

– Master programmer: controls the execution of programs

04/20/23 Erkay Savas 7

ENIAC– Multiplier

• multiplication of a 10-digit number by a d-digit number (for d up to 10) took d+4 cycles, so a 10 by 10-digit multiplication took 14 cycles, or 2800 microseconds—a rate of 357 per second

– Function tables: used for programming

• Programming: by wiring the cables and setting 3000 switches manually. Done for every program.

• Energy Consumption: 200 Kilo Watt

• 30 tons

• Purpose: Used for calculation of missile trajectories.

04/20/23 Erkay Savas 8

Programming(!) ENIAC

04/20/23 Erkay Savas 9

Dispute• The English also claimed that the world’s first

general-purpose computer was built in England– COLOSSUS in 1943

– The problem was about the definition of general-purpose computer.

• Also by British, EDSAC (Electronic Delay Storage Automatic Calculator)– by Maurice Wilkes of Cambridge University in 1949

– Stored-program computer

• Non-electronic computers– Harvard Mark I – IBM ASCC (1944), Zuse Z3 (1941)

04/20/23 Erkay Savas 10

Difference Engine

• By Charles Babbage in 1850s

What about now?

What are the capabilities of contemporary computers?

04/20/23 Erkay Savas 12

Intel CentrinoTM Architecture

• CPU• Chipset• Wireless Network

Interface

Chipset

04/20/23 Erkay Savas 13

Intel Core 2• Core 2 brand refers to a range of Intel's

consumer 64-bit dual-core and MCM quad-core CPUs with the x86-64 instruction set,– Produced: From 2006 – Max CPU clock: 1.06 GHz to 3.20 GHz – FSB speeds: 0533 MT/s to 1600 MT/s – Process: 65 nm to 45 nm – Instruction set: x86, MMX, SSE, SSE2, SSE3,

SSSE3, x86-64, SSE4 (SSE4 is for only Penryn-based processors)

– Microarchitecture: Intel Core microarchitecture– Cores: 1, 2, or 4 (2x2)

04/20/23 Erkay Savas 14

Yorkfield XE Processor• Core 2 Extreme QX9650:

– On November 11, 2007 – The first Intel desktop processor to use 45 nm

technology and high-k metal gates.– Features a dual-die quad core design– Two unified level-two (L2) caches, with a total of

12 MiB (2 × 6144 KiB). – Features a 1333 MHz FSB – clock speed of 3 GHz. – The processor incorporates SSE4.1 instructions – total of 820 million transistors on 2x107 mm²

dies.04/20/23 Erkay Savas 15

Mobile Processors: Merom• Core 2 Duo ULV U7700 • Clock speed: 1333 MHz • Second level Cache Size: 2048 KiB • Front Side Bus Speed: 533 MT/s • Voltage: 0.80 - 0.975 V • Power: 10 W • Release Date: December 30, 2007 • Price: $289• Technology: 65 nm• Die Size: 111 mm2.• Virtualization and Trusted Execution Technology

04/20/23 Erkay Savas 16

04/20/23 Erkay Savas 17

How is this progress possible?

What is the key technology behind it and what is the law that is governing the

progress in this technology?

04/20/23 Erkay Savas 18

Semiconductor Technology

• Transistors: tiny devices that can be realized easily(!) in silicon which is abundant on earth.

• A transistor is basically a switch that can be used to implement some logical operations.

• A collection of transistor which implement a logical operation is called as gate.

• A gate implements a logic primitive (AND, OR)

• Using logical gates we can realize millions of useful operations we can think of.

04/20/23 Erkay Savas 19

Moore’s Law• The observation made in 1965 by Gordon

Moore, co-founder of Intel, – the number of transistors/in2 on IC had doubled

every year since the IC was invented. – Moore predicted that this trend would continue

for the foreseeable future.

• The pace slowed down a bit, – transistor density doubles approximately every

18 months, – and this is the current definition of Moore’s Law.

• Most experts expect Moore’s Law to hold for at least another one or two decades

04/20/23 Erkay Savas 20

Architecture & Organization • More transistors and better process

technology faster processors• It is important to know what to do with

these resources– more memory– larger cache– another cache level– a powerful multiplier unit– unit for networking operations– new instructions

04/20/23 Erkay Savas 21

Extremes: Fastest Computers

• Earth Simulator Computer (ESC)– Built by NEC – Fastest from 2002 to 2004

• Claimed Applications– high resolution global models

• predictions of global warming.

– high resolution regional models• predictions of el Niño, monsoon.

– Simulation of earthquake generation process

• Suspected Application– Simulation of nuclear weapon explosions

04/20/23 Erkay Savas 22

Fastest Computer: ESC• Processor technology: NEC SX

– vector processor

• Multiprocessor system– 5120 processors in total

• 10 TB of memory• 700 TB of disk

– 450 TB system– 250 TB users

• 1.6 PB of mass storage in tape drivers• Area of the computer: 4 tennis courts, three

floors• 35.86 trillion calculations per second

(TFLOPS)

04/20/23 Erkay Savas 23

Birds-Eye View of ESC

04/20/23 Erkay Savas 24

Cross-Sectional View of ESC

04/20/23 Erkay Savas 25

New ESC Facilities

04/20/23 Erkay Savas 26

Wiring of ESC

Blue Gene/L• The first computer in the Blue Gene series

– 2004– Linpack benchmark: 36.01 TFLOPS– 8-cabinet system, with each cabinet holding 1,024

compute nodes– On October 27, 2005, reaching 280.6 TFLOPS on

Linpack, • 65,536 "compute nodes" (i.e., 216 nodes) • an additional 1024 "I/O nodes" in 64 air-cooled cabinets. • a 900TB filesystem.

– During an upgrade in 2007 (LLNL BlueGene/L the performance increased to 478 TFLOPS sustained.

04/20/23 Erkay Savas 27

04/20/23 Erkay Savas 28

Other Extreme: Smallest• Sensor Nodes

– Berkeley Mote:

– 8-bit RISC processor– 4 MHz clock– 8 Kbytes Flash Memory

for programs • OS code space: 3500 bytes• Available code space: 4500 bytes

– 512 bytes RAM Data Memory– 10 Kbps radio

04/20/23 Erkay Savas 29

Other Extreme: Smallest

commercially available

vs.

in development

vs.

what it needs to be (smart dust)

Kris Pister (University of California at Berkeley)

• Cost: – today ~ $100 – need to be < $1

• Size:

04/20/23 Erkay Savas 30

New Ideas, Concepts

• Ubiquitous computing– Everywhere but not too visible,– Easy to interact– Real, the opposite of VR (virtual reality).

• Pervasive computing• Numerous, casually accessible, often invisible

computing devices • Frequently mobile or embedded in the

environment • Connected to an increasingly ubiquitous network

structure

• Wearable computer• Embedded processors

04/20/23 Erkay Savas 31

Embedded Processors

• Today, large majority of computation devices are not in desktop computers or workstations but embedded in video games, laser printers, automobiles, etc.

• Tomorrow, computation devices will be definitely and literally everywhere;

• watches, roads, our clothes, our desk, and any small household item, walls of your house will have some sort of computational capacity if not certain type of intelligence.

04/20/23 Erkay Savas 32

Embedded Processors

04/20/23 Erkay Savas 33

Different Processors