Introduction to

High performance Computer Archiceture

INTRODUCTION
– Computer performance has been
increasing phenomenally over the
last five decades.
– Brought out by Moore’s Law:
●
Transistors per square inch roughly double
every eighteen months.
– Moore’s law is not exactly a law:
●
but has held good f o r nearly 50 years.

2
 Moore’s Law

Gordon Moore (co-founder of

Intel)
predicted in 1965: “Transistor density Moore’s Law: it’s worked
of minimum cost semiconductor for a long time
chips would double roughly every 18
months.”

Transistor density is correlated to

processing speed.
Trends Relat ed t o Moore’s Law

• Processor performance:
• Twice as fast after every 2
years (roughly).
• Memory capacity:
• Twice as much after every 18
months (roughly).
How Did Performance Improve?
●
Since 1980s, most o f the performance
improvements have come from:
– Architectural and organizational
innovations

●
What is the difference between:
– Computer architecture and computer
organization?
 Architecture vs. Organization
●
Archit ect ur e :
– Also known as Instruction Set Architecture (ISA)
– Programmer visible part o f a processor: instruction set,
registers, addressing modes, etc.
●
Organization:
– High-level design: how many caches? how many arithmetic
and logic units? What type o f pipelining, control design, etc.
– Sometimes known as micro-architecture
– A microarchitecture is a hardware implementation of an
ISA (instruction set architecture). An ISA is a structure of
commands and operations used by software to
communicate with hardware. A microarchitecture is the
hardware circuitry that implements one particular ISA.
 Computer Architecture

• The structure o f a computer that a

machine language programmer must
understand:
• – To be able to write a correct program
f o r that machine.
•A family o f computers o f the same
architecture should be able to run the same
program.
• – Thus, the notion of architecture leads to
“binary compatibility.”
●
Modern processors such as
Intel Pentium, AMD Athlon, etc. use:
– Many architectural and organizational
innovations
– Innovation in memory, bus, and storage
designs as well.
– Multiprocessors and clusters
Single-core computer
Single-core CPU chip
the single core
 Multi-core architectures
Replicate multiple processor cores on a
single die.
Core 1 Core 2 Core 3 Core 4

Multi-core CPU chip

Intel Core 2 Duo

• Homogeneous cores
• Bus based chip
interconnect.
• Shared on-die Cache
Memory.

Source: Intel Corp.

Intel Core 2 Duo
– Replacement f o r Pentium 4 and Pentium D
CPUs.
●
I n t e l claims:
– Core 2 duo provides 40% more performance at

40% less power compared to the Pentium D.

●
All Conroe processors are manufactured with 4
MB L2 cache:
Figure :Since the Pentium processor was first released in 1993, the
standard has been for a processor to have two arithmetic logic units so that
it can process two instructions at once
Figure: Quad-core processing with L1, L2, and L3 cache and the
memory controller within the processor housing
Courtesy: Course Technology/Cengage Learning
Intel Processors

Table Current Intel processors

 Intel Processors
(cont’d.)

Table Current Intel processors (continued)

Intel® Core™ processors
• A processor is like a computer’s brain - and Intel® Core™
processors are the most powerful. They have multiple cores
for more power and smoother multi-tasking.
• There are 4 main categories: i3, i5, i7 and i9.

• Each year the Intel® Core™ processors are updated. 2018’s updates are known
as the 8th generation.
• The number after the hyphen on its serial number should give you a clue – for
example, an Intel® Core™ i7-7820HQ is 7th gen

• 8th Generation Intel® Core™ processors

• Intel®’s 8th Generation of processors has moved with the times and delivered
some exciting new features. Including:

• Incredible VR experiences
• High-quality 4K UHD content
• Two more cores – up to 6 instead of the original 4
• Introduction of the super powered Intel Core i9 and X-series.
• With a Core™ i3 you can:(no of core 2-4)

• Browse multiple webpages smoothly

• Work in Word or Excel
• Stream movies and TV shows from Netflix in HD
• Listen to music on Spotify
• Multi-task efficiently with Intel® Hyper-Threading technology

With a Core™ i5 (no of core(2-4),you can:(

• Smoothly multitask - work on spreadsheets, stream music and browse

the web
• Work on complicated tasks – like rendering big Excel files
• Edit in Photoshop and sketch in Illustrator
• Create, share and watch 4K content
• Play intensive PC games – see more Intel®'s gaming processors
• Benefit from faster repeated tasks thanks to the large cache size
• Stream from multiple sites
• Get a temporary boost when using demanding programs with Intel®
Turbo Boost Technology 2.0
• With a Core™ i7(no-of core 4-6) you can:

• Encode video more efficiently

• Work smoothly in 3D modelling programs
• Smoothly edit in Photoshop and sketch in Illustrator
• Watch and edit 4K UHD content and 360° videos
• Work productively with demanding creative programs – each core uses 2
‘threads’ rather than 1 with Intel®'s Hyper-Threading technology
• Benefit from faster repeated tasks thanks to the large cache size

• Core™ i9 – Extreme gaming, mega-tasking and high-end content creation

• The newest addition to the Intel family, the Core™ i9 X-Series, is Intel’s
most powerful processor with 18 cores and 36 threads. And with updated
Intel® Turbo Boost Max Technology 3.0, it elevates everything you do to
new heights.
• With a Core™ i9 (no-of core-8-16)you can:

• Produce, edit and share 4K UHD content and 360° videos

• Work smoothly in 3D modelling programs
• Produce and edit high-quality music
• Smoothly edit in Photoshop and sketch out in Illustrator
• Enjoy the ultimate gaming and VR experience
• Work more productively when using demanding creative programs
• Benefit from faster repeated tasks thanks to the large cache size
 Classes of Computer

The computer systems can be

classified on the following basis:

1. On the basis of size.

2. On the basis of functionality.
3. On the basis of data handling.
 Classes of Computer
Classification on the basis of size Classification on the basis
• super computer of Functionality
• mainframe computer • Server
• mini computer • Workstation
• microcomputer • Embedded Computer

Classification on the basis of Data

Handling
• Analog
• Digital
• Hybrid
 Classification on the basis of size:super computer
• A supercomputer is a computer with great speed and memory. This kind of
computer can do jobs faster than any other computer of its generation. They are
usually thousands of times faster than ordinary personal computers made at that
time.
• A supercomputer can be built by stacking computer processors in a giant box and
interconnecting them to work on a complex task through parallel processing. Such
an arrangement is called a Cluster supercomputer. Here, each individual computer
in the cluster is called a node.
• Supercomputers speed are measured in floating point operations per second
(FLOPS) in units of :
• megaflops (MFLOPS)
• gigaflops (GFLOPS)
• teraflops (TFLOPS)
Used for solving high calculation and intensive tasks like:

Weather forecasting
Analysis of data and information
Astronomical Observation
Integrate design of engineering products
For solving large input scientific calculations and advanced scientific problems.
supercomputer manufacturer
Aspen Systems,SGI,IBM,Cray Research,Compaq, Hewlett-Packard, Thinking
Machines,Cray Computer Corporation,Control Data Corporation
super computer chhallenges
Generates large amount of heat.

The speed of data transfer will limit the super computer’s performance.

Supercomputer’s consume and produce massive amount of data in a very short

period of time .
world fasted supercomputer:Jaguar
Pratyush [Cray XC40*]
High speed(1.759 Petaflop) Owner -> Indin Institute of
Great performance Tropical Meteorology
High data transfer rate(284GB/s) Processor -> Intel Xeon Broadwell
High system memory(362TB) E5-2695 v4 18-core processors
with a clock speed of 2.1GHz
Rank -> 45th
 Classes of Computer
Personal Internet of
Clusters/
warehouse-
Feature mobile device Desktop Server things/
scale computer
(PMD) embedded

Price of $5000– $100,000– $10–

system $100–$1000 $300–$2500 $10,000,000 $200,000,000 $100,000

Price of $10–$100 $50–$500 $200–$2000 $50–$250 $0.01–$100

microproces
sor

Critical Price,
system Cost, energy, Price- Throughput, Price-performance, energy,
design
issues media performance, availability, throughput, energy application-
energy, scalability,
performance, graphics energy proportionality specific
responsivene performanc
ss performance e