Motivations and Introduction

1
Motivations and Introduction

• Phenomenal growth in computer industry/technology
  
• X2/18mo in 20yr.? multi-GFLOPs processors,
  largely due to
• Micro-electronics technology
• Computer Design innovations
• We have come a long way in a short time of 56
  years since the 1st general purpose computer in
  1946

2
Motivations and Introduction

• Past (Milestones)
• First electronic computer ENIAC in 1946 18,000
  vacuum tubes, 3,000 cubic feet, 20 2-foot
  10-digit registers, 5 KIPs (thousand additions
  per second)
• First microprocessor (a CPU on a single IC chip)
  Intel 4004 in 1971 2,300 transistors, 60 KIPs,
  200
• Virtual elimination of assembly language
  programming reduced the need for object-code
  compatibility
• The creation of standardized, vendor-independent
  operating systems, such as UNIX and its clone,
  Linux, lowered the cost and risk of bringing out
  a new architecture
• RISC instruction set architecture paved ways for
  drastic design innovations that focused on two
  critical performance techniques
  instruction-level parallelism and use of caches

3
Motivations and Introduction

• Present (State of the art)
• Microprocessors approaching/surpassing 10 GFLOPS
• A high-end microprocessor (lt10K) today is easily
  more powerful than a supercomputer (gt10million)
  ten years ago
• While technology advancement contributes a
  sustained annual growth of 35, innovative
  computer design accounts for another 25 annual
  growth rate ? a factor of 15 in performance
  gains!

4
Motivations and Introduction

• Present (State of the art)
• Three different computing markets (fig. 1.3)
• Desktop Computing - driven by price-performance
  (a few hundreds through over 10K)
• Servers availability driven (distinguished from
  reliability), providing sustained high
  performance (fig. 1.2)
• Embedded Computers fastest growing portion of
  the computer market, real-time performance
  driven, and need to minimize memory and power, as
  well as ASIC

5
Motivations and Introduction

• Present (State of the art)
• The Task of the Computer Designer (Fig. 1.4)
• Instruction Set Architecture (Traditional view of
  what Computer Architecture is), the boundary
  between software and hardware
• Organization, high-level aspects of a computers
  design, such as the memory system, the bus
  structure, the internal design of CPU, based on a
  given instruction set architectrue
• Hardware, the specifics of a machine, including
  the detailed logic design and the packaging
  technology of the machine.
• Future (Technology Trends)
• A truly successful instruction set architecture
  (ISA) should last for decades, however it takes
  an computer architects acute observation and
  knowledge of the rapidly changing technology, in
  order for the ISA to survive and cope with such
  changes

6
Motivations and Introduction

• Future (Technology Trends)
• IC logic technology transistor count on a chip
  grows at 55 annual rate (35 density growth rate
  10-20 die size growth) while device speed
  scales more slowly
• Semiconductor DRAM density grows at 60 annually
  while cycle time improves very slowly (decreasing
  one-third in ten years). Bandwidth per chip
  increases twice as fast as latency decreases
• Magnetic dish technology density increases at
  100 annual rate since 1990 while access time
  improves at about a third every ten years and
• Network technology both latency and bandwidth
  have been improving, with more focus on bandwidth
  of late the increasing importance of networking
  has led to faster improvement in performance than
  beforeInternet bandwidth doubles every year in
  the U.S.
• Scaling of transistor performance while
  transistor density increases quadratically with
  linear decrease in feature size, transistor
  performance increases roughly linearly with
  decrease in feature size?challenge opportunity
  for computer designer!
• Wires and power in IC propagation delay and
  power needs?

7
Motivations and Introduction

• Cost, Price and Their Trends
• Understanding cost and pricing structure of the
  industry and market is key to cost-sensitive
  design of computers
• The Learning Curve manufacturing costs decrease
  over time (Fig.1.51.6), best measured by change
  in yield ? helps project costs over products
  life

8
Motivations and Introduction

• Cost, Price and Their Trends
• Cost of an IC (Fig. 1.8)

9
Motivations and Introduction

• Cost, Price and Their Trends
• Cost of an IC die yield has been obtained
  empirically, where ? corresponds inversely to the
  number of masking levels (manufacturing
  complexity). For todays metal CMOS processes,
  its estimated at 4.0

10
Motivations and Introduction

• Distribution of Cost in a System
• Cost vs. Price (Fig. 1.10)

11
Motivations and Introduction

• Cost vs. Price (Fig. 1.10)
• Component cost(CC) original cost from a
  designers point of view
• Direct cost (DC, 20 of CC) making a product
  (labor cost, scrap, warranty, etc), not including
  service and maintenance
• Gross margin (GM, 33 of CCDC) indirect cost ?
  overhead RD, marketing, sales, manufacturing
  equipment maintenance, building rental, cost of
  financing, pretax profits, and taxes
• Average selling price (ASP) CC DC GM
• Average discount (AD, 33 of ASP) volume
  discounts by manufacturers
• List price ASP AD

12
Performances Quantitative Principles

• X is n times faster than Y ??
• Performance (throughput) is inversely
  proportional to execution time
• Definition of time
• wall-clock time response time or elapsed time
• CPU time the accumulated time during which CPU
  is computing
• user CPU time
• system CPU time
• An example from UNIX 90.7u 12.9s 239 65
• 90.7u user CPU time (seconds)
• 12.9s system CPU time
• 239(159 sec) elapsed time
• 65 percentage of CPU time

13
Performances Quantitative Principles

• Workload Representations (in decreasing
  accuracy)
• Real applications most accurate but inflexible
  and poor portability
• Modified/scripted applications scripts to
  stimulate (or highlight) certain features and to
  enhance portability
• Kernels extracted from real programs, good for
  isolating performance of individual features of a
  machine
• Toy benchmarks simple and run on almost all
  computers, good for beginning programming
  assignments
• Synthetic benchmarks artificially created to
  match an average execution profile, do not
  reward optimizations of behaviors in real
  programs but absent from benchmarks, and vice
  versa--thus can be misleading

14
Performances Quantitative Principles

• Benchmark Suites collection of kernels, real and
• benchmark programs, lessening the weakness of any
  one benchmark by the presence of others.(fig.
  1.11)
• Desktop Benchmark Suites
• CPU-intensive benchmarks SPEC (Standard
  Performance Evaluation Corporation) SPEC89 ?
  SPEC92 ? SPEC95 ? SPEC2000(11 int CINT 14 fp
  CFP2000, fig. 1.12) real programs modified for
  portability and highlighting CPU
• Graphics-intensive benchmarks SPECviewperf for
  systems supporting the OpenGL graphics library,
  SPECapc for applications with intensive use of
  graphics
• Server Benchmark Suites
• CPU-throughput benchmarks SPEC CPU2000 ?
  SPECrate
• I/O-intensive benchmarks SPECSFS for file
  server, SPECWeb for web server
• Transaction-processing (TP) benchmarks TPC
  (Transaction Processing Council) TCP-A (85) ?
  TCP-C (complex query) ? TCP-H (ad-hoc decision
  support)? TCP-R (business decision support) ?
  TCP-W (web-oriented)
• Embedded Benchmarks EEMBC (embassy suites,
  fig. 1.13)

15
(No Transcript)
16
(No Transcript)
17
(No Transcript)
18
(No Transcript)
19
(No Transcript)
20
(No Transcript)
21
(No Transcript)