 |
The stored program computer |
|
The system unit
 |
Grand Central Station of the PC
 |
Motherboard |
|
Passive Backplane |
|
|
Chips and Dips |
|
CPU and memory |
|
Data,
Address, and Control buses |
|
Power Supply |
|
Expansion slots and connectors |
|
Storage devices |
|
|
Computer Architecture and Chip Manufacturing Tutorials
|
|
Motherboard
|
Large printed circuit (PC) board into which other circuit boards and
components are plugged. |
|
Conduction paths for distribution of power, data, and control signals. |
|
Building as many components as possible onto the motherboard reduces
the overall cost of manufacturing the computer. |
|
Large board size increases problems associated with vibration, thermal
expansion, and replacement cost.
|
Components that might need replacement are mounted on separate
boards that plug into the motherboard. |
|
Upgrade (CPU, memory, secondary storage devices) |
|
Failure (modem) |
|
|
|
Passive Backplane
|
Used in industrial and military systems to
decrease the Mean Time To
Repair (MTTR). |
|
Rigid set of card slots interconnected without any active components. |
|
All active components are on cards that are plugged into the passive
backplane. |
|
Troubleshooting and Repair
|
Is done by replacing cards until the system resumes working
properly. |
|
Repair of a failed card is done later. |
|
|
|
Simple CPU and Memory sketch (PDF File,
46 kB) |
|
Central Processing Unit (CPU)
|
Control Unit
|
Instruction Cycle:
fetch instruction, decode, fetch operands, execute, store or
write-back |
|
Types of instruction sets:
|
RISC, CISC.
|
CISC: Complex Instruction Set
Computer. This is good for machines used for science and
engineering. |
|
RISC: Reduced Instruction Set
Computer. This is good for shuffling bits and bytes, such as
networking and word processing. Occasional
complex calculations can be done as a series of
RISC instructions. IBM pushed this concept,
as well as CISC. The 64-bit DEC Alpha processor uses
RISC. |
|
|
Fixed
length vs Variable length instruction set |
|
|
|
Arithmetic / Logic Unit: arithmetic, comparison, Boolean operations
|
Arithmetic
|
Simple CPU: Add, Subtract, Negate or
Change Sign, Increment, Decrement |
|
Cheap CPU: Multiply, Divide |
|
Expensive CPU: Floating Point Arithmetic |
|
|
Comparison
|
Equal, Not Equal |
|
Less Than, Greater Than |
|
Less Than or Equal, Greater Than or
Equal |
|
|
Boolean Operations
|
16 logic operations for one
or two variables.
|
[If large dots appear in the symbol columns of the pdf
file, you are missing some needed fonts. You can get
the fonts by downloading and installing the free Scientific
Viewer.] |
|
|
AND, OR, NOT |
|
NAND, NOR |
|
XOR: Exclusive OR |
|
|
|
Registers: 8, 16, 32...
|
Registers are special storage circuits in a CPU. |
|
Register size depends upon its purpose. |
|
Registers usually have associated logic circuits. |
|
Registers operate at the speed of the CPU, which is generally much
faster than memory fetch or store times. |
|
|
Simulated Simple CPU Demonstration
|
|
Computer Speed
|
System Clock
|
Clock
speed measured in Hertz (Hz). 1 Hz = 1 cycle per second. |
|
|
The number of clock pulses needed to execute an instruction
depends upon the instruction. For example, a multiply
instruction might take more than 20 times as many clock cycles as an
addition instruction to execute. The speed of execution
depends upon the mix of instructions used. The mix of
instructions are typically different for
|
Science, engineering, statistics, numerical analysis (floating
point, vector processing) |
|
Mathematics, artificial intelligence (symbolic processing and
formal logic) |
|
Business transaction data processing (simple arithmetic,
intensive storage and hard copy output) |
|
System software, message switching (boolean logic, intensive
device input and output, stack manipulation, bit manipulation) |
|
|
Measures of computer speed
|
MIPS: Millions of Instructions Per Second |
|
MFLOPS: Millions of Floating point Operations Per Second |
|
GHz: System clock speed. This is
only an indirect measure of system speed. |
|
|
|
Architecture
|
|
Processors
|
Silicon: the backbone of the semiconductor industry today. |
|
Gallium Arsenide (GaAs): order of magnitude faster than silicon,
and requires more electrical power. |
|
|
Comparison of Personal Computer Processors
|
Intel, AMD, Motorola, DEC, IBM
|
Alpha Processor: silicon-based 64-bit processor now by API
(Alpha Processor, Inc), with 256-bit memory bus |
|
Intel
Itanium Processor has
|
a
well developed instruction set for multimedia
support. |
|
an
interval timer useful for real time control
applications. |
|
|
Technical documentation for Intel processors are available
online from Intel. http://www.intel.com |
|
|
MMX: Multimedia Extensions. |
|
SSE: "Streaming-SIMD-Extensions" single-instruction, multiple-data instructions
|
Higher level language support: Intel C++/C and Fortran 95
compiler, to target IA-32 and IA-64 architectures. |
|
|
The motivation for packing more onto a wafer:
|
Reduces overall cost of system manufacturing. |
|
Increases reliability of a working system after placed into
service. |
|
Decreases distance between components, and thus increases
speed, compared to the same system built onto several chips. |
|
|
Buy the processor that meets your needs. Do not decide based
only on clock speed.
|
Decide what you want to use your
computer for. |
|
Check the processor requirements on
the software you want to run. |
|
Message switching (like Internet routers) does not require
sophisticated logic, but it does require speed. RISC
processors are a good choice. |
|
Office applications rarely require floating point
arithmetic. The Celeron, and clones, with integer and
fixed point arithmetic provide good performance at a reasonable
cost. RISC processors are a good choice. |
|
Science and engineering applications need floating point
arithmetic. Use a CISC processor. RISC processors are a
bad choice unless a separate floating point coprocessor is
available. |
|
|
|
Processor Installation and Upgrades
|
A computer is designed as a system.
Upgrading a processor,
or upgrading memory, does not necessarily result in an increase of
performance! You need to consult the technical documentation
for the motherboard to determine what makes sense. Mere plug
compatibility is not sufficient to guarantee that an upgrade will be
worthwhile. |
|
Do not handle processor or memory chips without wearing an
anti-static wrist strap, and connecting the other end to bare metal
of the chassis. You can get an anti-static wrist strap for
less than $10 at Radio Shack. It is well worth the cost. |
|
Chip for chip upgrade |
|
Piggyback: add additional new processor |
|
Daughterboard upgrade: processor is on a board |
|
Zero-Insertion Force (ZIF) socket |
|
|
Heat Dissipation
|
Excessive heat and power transients are the most important causes
of CPU failure. |
|
Conduction: Heat sinks, board edge: conduct heat away from
heat-generating components. |
|
Gas transport: Fans, Heat pipes, ventilation louvers
|
Do not block the ventilation holes of your chassis! |
|
If a fan stops, turn your computer off immediately. |
|
Radio Shack sells replacement fans. $5 to $15 for a fan
is cheaper than a CPU, and they are not hard to replace.
You can do it yourself. |
|
|
Liquid transport:
|
Water cooling, usually from card edges. In some
specialty applications, the water cooling can be mated with a heat
sink mounted to a component, such as a processor chip package. |
|
Used
on high performance systems, or in environments where
heat dissipation is difficult. |
|
|
|
Coprocessors
|
Digital Signal Processor (DSP) chips are used for high-end
applications in signal processing applications. |
|
Floating point coprocessors were common additions on pre-Pentium
era microcomputers. |
|
|
Parallel processing
|
Supercomputers are characterized by many general-purpose CPUs. |
|
The Transputer is an example of a multiple microprocessor system,
which was available in the early 1980s. It uses a language
called Occam. |
|
Modern mainframe computers often have a few processors. |
|
|
|
Data Representation
|
Analog and Digital
|
Analog
|
An analog signal can possibly represent an uncountable number of states. |
|
An ideal analog signal is not necessarily continuous. |
|
To the electrical engineer, in the real world, all signals are analog signals. |
|
We often represent digital signals using analog signals. |
|
|
Digital signals
|
A digital signal is a signal used to represent a countable number of states. |
|
It can have possibly an infinite number of states. |
|
Discrete signals are often the model idealized in digital communications. |
|
A real digital signal is an analog signal that is used to represent a discrete number of states. |
|
HDTV 8-VSB is an example of a real digital signal that is not a binary signal. |
|
|
Binary signals
|
A binary signal is a signal used to represent only two states. |
|
Binary signals are often the model in computer data representation. |
|
Square waves
|
Square waves are often the model of the binary signal. |
|
Square waves are approximated in electronics and optics by pulses. |
|
Square wave pulses have distortion, even with the best of engineering practice. |
|
Gibbs' phenomena |
|
Square waves do not retain their shape when propagating.
This is why you need a modem for long distance communication
between computers. |
|
|
A real binary signal is an analog signal that is used to represent two states. |
|
|
When most people say "digital", they mean "binary". |
|
|
In-depth tutorial: Number
Systems. If you are serious about the computer field, you need
to master this topic early in your training. It is not hard. |
|
bit, byte
|
bit: binary digit, used to represent two states
|
{ on, off }, { low voltage, high voltage }, { yes, no }, {
boy, girl },
{ 0,1 }, { -18, -2 }, { -5, 12 } |
|
NOT binary: Any set that does not have exactly 2 elements,
{-1, 0, 1}, {A, B, C} |
|
|
byte: a collection of bits necessary to represent one character.
|
IBM 7040 had 36-bit word with six 6-bit bytes. Only
36-bit words could be fetched from memory. Byte
manipulation occurred in the CPU. |
|
PDP-8 had byte-addressable memory |
|
IBM 360 had 8-bit byte-addressable memory |
|
|
Character Representation
|
|
|
Future of the byte.
|
Will the byte be redefined to be 16 bits? Wait and
see. No law says a byte must remain 8-bits. We changed
from 6 to 8. We can change again. |
|
"Octet" is now a common term for 8 bits, particularly
among people working with computer networks. |
|
Processor boards now have wider data buses. It could become
practical. |
|
16-bit unsigned integers are popular for loop counters. |
|
|
Word size: depends on the data type, as well as the processor.
|
The units of data processed by the
instruction set of a CPU.
|
Data handled by the Arithmetic / Logic Unit
determines the physical word sizes a processor can handle. |
|
The specific data types available depend upon the processor. |
|
Usually, a word is an integer multiple of the addressable unit of memory. |
|
|
Bit, logical (Boolean)
|
Because of units of fetching data from memory, some machines
and languages define Boolean or Logical data type to be 1 byte
long [not restricted to 1 bit] |
|
|
Character Data
|
Byte |
|
Language-dependent string formats
|
Fixed Length String |
|
Variable Length String
|
Includes a byte count, or |
|
Includes a null character string
terminator. |
|
|
|
|
Numeric
|
negative number formats: signed magnitude, unsigned (one's
complement, two's complement) |
|
fixed point (Radix point is fixed. In decimal,
the radix point is called a decimal point. In
other base number systems, it is called a radix point.)
|
Integer: Fixed Point format with no fraction. |
|
Unsigned: only positive numbers |
|
Short integer (16 bits), long integer (32 bits) |
|
Binary Coded Decimal (BCD) |
|
|
Floating point (used to represent real, complex,
quaternion numbers)
|
The binary equivalent of Scientific Notation is called
"Floating Point". Floating point arithmetic requires
significantly more effort than arithmetic needed for business data
processing. Floating point hardware adds to the cost of the
processor. |
|
Anatomy of a floating point number:
|
Machine dependent. IBM mainframes and Intel
processors are different. |
|
Analog to Scientific Notation: Characteristic
(exponent), Mantissa (fraction) 1.023 x 1023 |
|
Number System Base: Binary or Hexadecimal |
|
Normalization: binary or hexadecimal |
|
Implicit Bit: used only with Binary normalization |
|
|
Use of Floating Point to represent numbers
|
Real: floating point, double precision floating
point, extended precision, quadruple precision floating point. |
|
Complex: complex (ordered pair of floating point numbers), double precision complex,
extended precision, quadruple precision complex). |
|
|
Tutorials and Specifications:
|
|
|
|
|
|
Memory
|
Address: Identifies a location in memory
|
|
Access time: Time from request until data is
received
|
Access Times |
|
| Device |
Access times |
| RDRAM |
1.25 ns |
| SRAM: Static RAM
(not refreshed as often
as DRAM) |
5 - 15 ns |
| SDRAM |
7.5 ns |
| DRAM: Dynamic RAM
(must be refreshed) |
50 - 70 ns |
| EPROM |
55 - 250 ns |
| ROM |
25 - 250 ns |
|
|
Abbreviations:
|
1 kB
= 210 bytes = 1024 bytes
~ 103 bytes,
1 MB = 220
bytes ~ 106 bytes,
1 GB = 230 bytes ~ 109
bytes, 1 TB = 240 bytes ~ 1012 bytes |
|
1 ms = 10-3 s, 1 μs = 10-6 s, 1 ns = 10-9 s |
|
|
RAM: Random Access Memory
|
DRAM: dynamic RAM: requires re-energizing to retain contents.
Volatile memory. |
|
SRAM: static RAM: does not require refreshing. Volatile memory.
Faster than DRAM. |
|
1T-SRAM: Single transistor SRAM. It is a single transistor
DRAM cell that given the speed performance of SRAM, but at lower
power consumption, cost, and size. |
|
SDRAM: Synchronous DRAM: synchronized to the system clock. Volatile
memory. |
|
Magnetoelectronic Memory: Nonvolitale memory. Honeywell
demonstrated this in 1997. http://www.eetimes.com/story/OEG19991021S0033 |
|
|
RAM Requirements
|
Operating systems and major application packages are memory
hungry. Today's big system will be tomorrow's expected system,
and the day after tomorrow's obsolete system. Get the most RAM
you can at time of purchase of your computer. RAM is cheap
today. |
|
|
Memory Configuration
|
|
Board Configuration
|
SIMM: single inline memory module |
|
DIMM: dual inline memory module |
|
|
Cache:
|
Memory that is not part of primary storage (main memory) that is
used as a buffer between the device and the memory bus, servicing
the needs of a particular device. |
|
Idea:
speed matching. Provide the access time required
between devices of different speeds. |
|
CPU Cache, the text calls it "memory cache": L1, L2
cache
|
Primary cache, L1 Cache, internal cache: 8kB - 64 kB, fast
memory, built on the same chip as the CPU |
|
Advanced Transfer Cache:
|
L2 cache built onto the processor chip: 256 kB |
|
|
External cache:
|
L3 Cache, 64 kB - 4 MB, slower that L2 cache. |
|
Older L2 cache on a separate chip. |
|
|
|
Disk Cache |
|
|
Read Only Memory
|
ROM: Read Only Memory
|
Firmware: software stored in ROM. |
|
|
Programmable ROM: PROM |
|
EEPROM: Electronically erasable, programmable ROM. |
|
|
CMOS: Complementary Metal Oxide Semiconductor
|
This is a type of chip technology, and is used to implement many
circuits. |
|
CMOS requires less power than a transistor, which makes it good
for use in battery operated devices. |
|
Used in the IBM USB NET Camera Beige VGA 640 X 480 CMOS IMAG video
camera as part of the optical sensor. |
|
Fairchild Semiconductor is one manufacturer of CMOS chips. |
|
On PCs, a common use is to store configuration information.
|
type of disk drives, # cylinders, # heads |
|
keyboard, monitor |
|
time, date |
|
other information needed at startup time |
|
|
Like semiconductor RAM, it requires power to retain information.
Battery. |
|
|
Ports: interface or point of attachment
|
Connector types
|
male, female, gender changer. |
|
Number of pins and shape. Different connector types used for
different purposes helps sailor-proof the design. You usually
cannot plug a device into the wrong connector. |
|
|
Serial: RS-232, RS-422, RS-423.
|
1
data path or channel |
|
Webopedia: RS-422 supports multipoint connections whereas RS-423 supports only point-to-point connections.
http://webopedia.com/ |
|
|
Parallel
|
multiple
data paths or channels |
|
Centronics |
|
IEEE 1284: EPP (Enhanced Parallel Port), ECP (Extended
Capabilities Port) |
|
|
Universal Serial Bus (USB):
|
General
|
127 devices, daisy chained.
The device requiring the fastest service should be
plugged in closest to the system unit in the daisy chain
sequence. USB hub. |
|
USB ports are
self-powered. USB devices take power from
the USB port. Adding USB devices will reduce time between
battery charging on a laptop computer. Benefit: you have
fewer things to directly plug in to the wall. |
|
USB devices are hot-swappable. You do not need to turn
power off before connecting or disconnecting a USB device. |
|
|
USB-1
|
Data transfer rate:
1.5 and 12 mega-bits per second (Mb/s or mbps) [1.5 megabytes
per second]. |
|
|
Hi-Speed
USB (commonly called USB-2) http://www.usb.org
|
Data transfer
rates: 1.5, 12, and 480 mega-bits per second (Mb/s or
mbps) [60
mega-bytes per second]. 40 times faster than USB-1. |
|
USB-1 devices work with USB-2 ports, but only at USB-1 speed. |
|
Requires Windows 98, or newer. |
|
|
Example uses and benefits
|
Hook mouse to keyboard, reducing wire maze. |
|
|
|
Special Purpose:
|
IEEE 1394 (Fire Wire):
|
400
mega-bits per second [50 mega-bytes per second] |
|
63
devices, daisy chained |
|
digital video camera, digital
VCR, color printer, scanner, digital camera, DVD drive MIDI (muTeX, musicTeX, MTX, musicflex;
ftp.gmd.de) |
|
|
MIDI Manufacturers Association tutorial: http://www.midi.org/about-midi/abtmidi.htm
|
|
SCSI:
Small Computer System Interface
|
SCSI-I, 8-bid wide: 8 devices, |
|
optional 16-bit wide SCSI-II or SCSI-III: 16 devices |
|
Uses
50-pin Centronics connector |
|
|
IDE
(Integrated Device Electronics) = ATA (AT Attachment), 16-bit
interface, 2 devices limit
|
ATAPI: AT Attachment, Packet Interface |
|
ATAPI-5 is 66 MHz, ATAPI-6 is 100 MHz |
|
|
|
|
IrDA: Infrared Data Association:
line of sight transmission. |
|
Bluetooth:
radio link. |
|
|
System Port Numbers |
|
A port is also a specific process to which an Internet or other network message is to be forwarded
when it arrives at a server. A process is identified by a port
number. |
|
|
Buses
|
CPU bus, Data bus, address bus, control bus. These may have different
properties in speed and width.
|
The CPU bus transfers data between components of the CPU, such as
between the accumulator and registers. |
|
The address bus transmits addresses between the CPU's program
counter, address register, and the memory address registers. |
|
The data bus transmits data between the CPU and the memory buffer
register of main memory and other devices. |
|
The control bus transmits control, keying and timing signals
between computer components. |
|
|
Bus width
|
To transmit 64 bits on a 32 bit wide bus, you need 2 transmission
cycles. |
|
|
PC bus speeds: 100 MHz,
133 MHz, 266 MHz, 400 MHz. |
|
System bus |
|
Expansion bus
|
See figure 4-41, page 4.31. |
|
ISA (Industry Standard Architecture) bus: Slow speed devices:
mouse, modem, sound card, low speed network card. |
|
Local bus
|
PCI (Peripheral Component Interconnect) bus: video and sound
cards, SCSI cards, high-speed network interface cards. |
|
PCI (Peripheral Component Interconnect) expansion slots, short white. |
|
PCI-X, doubles the bandwidth of the PCI bus to enable more than 1-Gbyte-per-second throughput. However, PCI-X cards would maintain backward compatibility with
current PCI cards. http://www.picmg.org/ |
|
|
The Accelerated Graphics Port (AGP): Intel design. Direct
memory access by video card. |
|
Universal Serial Bus (USB) and 1394 bus |
|
PC Card bus |
|
|
Accelerated Graphics Port (AGP) |
|
|
Bays: External, Internal |
|
Power Supply
|
Voltage selection: 120 V, 240 V selects transformer taps. Permits
international sales. |
|
Internally, the computer generally uses 12 VDC and 5 VDC. 120
VAC --> Transformer --> 5 and 12 VAC --> rectifier -->
rippled DC --> filters --> smoother DC --> computer use. |
|
Problems in the electrical power to your computer:
|
Lightning strikes. Unplug equipment, including phone
lines, during lightning storms. Do not trust surge protectors
or UPS units to save equipment from this problem. |
|
Voltage spikes and fluctuations. Often from electric motors
starting, turning on lights (especially fluorescent lamps). |
|
Voltage noise. |
|
Voltage loss. |
|
|
The best present you can buy your computer is a continuous UPS. Battery, Uninterruptible Power Supply (UPS).
Two varieties: standby and continuous. To be effective, the
switch-in time needs to be less than 1/4 of a clock cycle. Typical
switch-in time is 5 ms. Standby systems are adequate for resistive
and inductive loads (lighting, heating, motors). For signal
processing equipment, including computers, the continuous UPS is more
appropriate. http://computer.howstuffworks.com/surge-protector8.htm
 |
|
|
Mobile
Computers and Devices: Laptop and Notebook Computers,
PDAs
|
Inadequate heat dissipation is the primary problem that limits life of
a laptop computer. Do not block ventilation ports. |
|
Battery life is a critical issue for laptops intended for battery
use. For laptops intended to be used without being plugged in, a
passive matrix display lowers battery drain. The active matrix
display requires more power. |
|
Port replicator |
|
Docking station |
|
Page 4.03: The three common reasons for computer hardware failure are
lightning strikes (affecting both power and phone line), momentary power loss,
and overheating. An online Uninterruptible Power Supply (UPS) is the best
solution to the first two problems. To prevent overheating, be careful to
not block ventilation holes of the chassis. If a fan stops, turn your
computer off until you have replaced the fan. Radio Shack sells
replacement fans. Clean the fan blades annually.
Page 4.04: In an industrial or military setting, an alternative to the
motherboard is the passive backplane architecture. The great advantage of
the passive backplane approach is a greatly reduced Mean Time To Repair.
The time a technician arrives at a broken computer until it is running again is
expected to be less than 5 minutes using the passive backplane design. It
is hard to replace a motherboard in less than 30 minutes. This time
difference can be significant in mission critical systems.
Page 4.05: The operating system manages most of a computer's operations.
The CPU is the hardware that carries out the instructions. The operating
system does the goal setting, scheduling, maintaining critical data, etc.
The operating system is the executive manager of computer hardware resources.
Pages 4.05 - 4.06: Change "machine cycle" to "instruction
cycle". A "machine cycle" can be one or two clock pulses,
depending on the system design. The number of machine cycles required to
execute an instruction depends upon the instruction. For example, an
integer multiply can take more that twenty times as many clock cycles an an
integer addition.
Page 4.12: Heat Pipe: This is used to distribute cooling air in a
notebook computer because the components are packed too tightly for free
convection to move air naturally around all components with enough velocity to
extract the required amount of heat. Notebook computers also use heat
sinks.
Page 4.17: RAM can be volatile or nonvolatile. RAM
means “Random Access Memory”. It
does not assume a particular technology for implementing it. Semiconductor
RAM is volatile. MRAM by IBM and Infineon Technologies, and magnetic core RAM
are not.
While magnetic core RAM is not in common use, there are some applications
where it is still attractive.
Page 4.16: Magnetic core and magnetic bubble memory are nonvolatile types of memory.
Magnetoelectronic memory is nonvolatile RAM.
See IEEE Spectrum, FEB 2000, pg. 33-40.
Bubble memory is not RAM. Information circulates around one or more loops rapidly.
It acts more like a disk in that information is sequentially stored in
blocks, but a bubble memory has no moving parts.
These types of memory are rarely used with personal or business
computers.