Top Banner
Indira Gandhi Indira Gandhi National Open National Open University presents University presents
28

Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

Dec 27, 2015

Download

Documents

Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

Indira Gandhi Indira Gandhi National Open University National Open University presents presents

Page 2: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

A Video LectureA Video Lecture

Course:Course:Computer Platforms Computer Platforms

Page 3: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

Performance of Performance of A Computer System A Computer System

Page 4: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

ObjectivesObjectives

define performance related terminology

define the need of benchmarks

define various factors of performance of processor/memory/bus/storage

Page 5: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

What is meant by What is meant by performance of a performance of a Computer?Computer?

The ability to execute instructions in accordance to the specified configuration capabilities

Page 6: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

Performance is measured Performance is measured forfor Processor The memory The bus Storage technologies, I/O

devices & The overall system

Page 7: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

What is meant by a 32 bit Processor? 32 bit integer data may be

processed at a time The size of addresses may be 32

bits

232 = 4 Giga Byte Memory

Processor performance Processor performance

terminologyterminology

Page 8: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

Processor performance Processor performance terminologyterminology Clock speed MIPS FLOPS

Page 9: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

Clock SpeedClock Speed

An internal clock regulates the rate of execution of instructions and synchronizes various computer components.

Clock speeds are expressed in Megahertz. 1 MHz = 1 Million clock

cycles/second

Page 10: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

A faster clock speed A faster clock speed faster computer? faster computer?

YES NO

One instruction/clock cycle One instruction/clock cycle

is possible in RISCis possible in RISC

More than one instruction More than one instruction per cycle is possible in Super per cycle is possible in Super scalar Architecturesscalar Architectures

Page 11: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

MIPSMIPS

A old measure of computer speed and power

Million instructions per second

Different instruction may require different times

No standard method for MIPS calculation

Page 12: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

Does the MIPS stands Does the MIPS stands for?for?

Meaningless Indicator of Performance.

It just provides the general idea of CPU capability

Pentium based system may run at 100 MIPS.

Page 13: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

FLOPSFLOPS Floating-point

operations/second. Measures the speed of

floating point unit (FPU).

Page 14: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

Super Computer needs to have FPUs :

MFLOPS = 1 Million FLOPSGFLOPS = 1 Million MFLOPS

Page 15: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

Which is the most frequently used transfer for Instruction execution?

The Processor to Memory or vice versa

Therefore, for better performance, the CPU need better memory capabilities

Page 16: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

Performance terms for Performance terms for Main MemoryMain Memory Access Time

Cycle Time

Page 17: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

Access TimeAccess Time The time taken to perform a

Read or Write Operation

The time from the instant that an address is presented to the memory to the instant the data have been stored or made available for use

Page 18: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

Typical Access TimeTypical Access Time

55-250 nanoseconds

read-only memory (ROM)

55-250 nanoseconds

EPROM

50-70 nanosecondsdynamic RAM (DRAM)

5-15 nanosecondsstatic RAM (SRAM)

Page 19: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

Memory Cycle TimeMemory Cycle Time Measures how quickly two

back-to-back accesses can be made

Access Time + Any additional time required before a second access can commence.

Page 20: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

Performance of the BusPerformance of the Bus Bus Width: For example the

width of PCI bus is 64 bits. Clock speed: PCI have 33 or

66 or even 100 MHz for some new configurations.

PCI may be implemented as 32 bit bus.

Page 21: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

Performance of the BusPerformance of the Bus

Throughput: The amount of data transferred from one place to another.

Typically, throughputs are measured in Kbps, Mbps and Gbps

The PCI bus at 32 bits and 33 MHz, yields a Throughput rate of 133 MBps

Page 22: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

Performance factors for Performance factors for Storage technologiesStorage technologies Access time

Data Transfer rate

Page 23: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

Access Time on DisksAccess Time on DisksAccess Time on disks consists of:

Seek time: Time taken to position the head on the required track

Rotational Latency: Time taken by a sector to reach under the head while rotating.

Page 24: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

Access TimeAccess Time

80-800 milliseconds

CD-ROM

19-100 milliseconds

Erasable Optical

6-12 millisecondsHard Disk Drive

Page 25: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

Data Transfer RateData Transfer Rate The speed with which data can

be transmitted from one storage device through the interface such as UDMA to system.

Data rates are often measured in Megabits or Mega Bytes per second (Mbps or MBps).

Another term for data transfer rate is: throughput

Page 26: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

BenchmarkBenchmark A test used to compare

performance of hardware and/or Software

A benchmark may test graphics speed, year 2000 compliance of system or performance for integer or floating point operations

Page 27: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

SPECSPEC Standard Performance Evaluation

Corporation, a nonprofit corporation set up by many computer and microprocessor vendors to create a standard set of benchmark tests.

Most The most widely used set of tests SPEC95( CPU 95 test) measures: Integer operations (SPECint95) and Floating point operations

(SPECfp95).

Page 28: Indira Gandhi National Open University presents. A Video Lecture Course: Computer Platforms.

Sum UpSum Up defined the performance

factors for: The CPU The Memory The Bus The storage Devices

Defined the benchmark tests