Transcript Lecture 13 - MemoryOrganization1

+
CS 325: CS Hardware and Software
Organization and Architecture
Memory Organization
+
Computer Systems Structure
+
Storage Hierarchy & Characteristics
+
Storage/Memory Hierarchy
+
Principle of Memory Hierarchy

To optimize memory performance for a given cost:

A set of technologies are arranged in a hierarchy that contains a
relatively small amount of fast memory and larger amounts of less
expensive, but slower memory.
+
Memory Hierarchy Importance

1980: no cache on CPU

1995: 2-level cache on CPU
1989: first Intel CPU with cache on chip
+
Memory Storage Characteristics

Location

Capacity

Unit of transfer

Access method

Performance

Physical type

Physical characteristics

Organization
+
Memory Storage - Location

CPU



Internal



Registers
L1, L2, L3, L4 Cache
Main Memory (System RAM)
BIOS (EEPROM)
External




Magnetic Disk (HDD)
Non Volatile Solid State (SSD)
Optical
Magnetic Tape
+
Memory Storage - Capacity

Word size




The natural unit of organization
Expected size of most data and instructions
Typically 32 bits or 64 bits
 Past: 16 bits
Typical Storage










L1 Cache: 32 – 64 KB per core
L2 Cache: 128 – 512 KB per core
L3 Cache: 2 – 8 MB (shared)
L4 Cache: 0 – 128 MB (video memory)
Main Memory (RAM): 4 – 32 GB (Typical Desktop)
HDD Cache: 16 – 64 MB
SDD: 64 – 512 GB
HDD: 200 – 2000 GB (Inexpensive, but extremely slow)
Optical:
 DVD: 4.7 – 17.08 GB
 Blu-ray: 25 – 100 GB
Magnetic Tape: 10 – 35 TB per cartridge (uncompressed)
+
Memory Storage – Unit of Transfer

Internal


External


Usually governed by data bus width
Usually a block which is much larger than a word
Addressable unit

Smallest unit which can be uniquely addressed

Byte internally (typically)
+
Memory Storage – Access Methods


Sequential (tape):

Shared read/write mechanism

Start at the beginning and read through in order

Access time depends on location of data and previous location of
magnetic tape
Direct (disk)

Shared read/write mechanism

Individual blocks have unique address

Access data by jumping to vicinity plus sequential search

Access time depends on location of data and previous location of
read/write mechanism
+
Memory Storage – Access Methods


Random (RAM):

Individual addresses identify locations exactly

Access time is independent of location or previous access
Associative (Cache):

Data is located by a comparison with contents of a portion of the
store

Access time is independent of location or previous access
+
Memory Storage – Performance

Latency/Access time



Time between presenting the address and getting the valid data
Memory Cycle time

Time may be required for the memory to “recover” before next
access

Cycle time is latency + recovery
Transfer Rate

Rate at which data can be moved

# of bits * (1/cycle time)
+
Performance – Transfer Rate
Example Problem

Assume we have a 32-Mbit SDRAM memory with 8 bits
simultaneously read and a cycle time of 250 ns.

How fast can data be moved out of memory?
8b * (1/250ns)
= 8b * (4x106/s)
= 32 Mbps
= 4 MBps
+
Memory Storage – Physical Types

Semiconductor




Magnetic



HDD
Tape
Optical




Cache
Main Memory (RAM)
SSD
CD
DVD
Blu-Ray
Others


Bubble
Hologram
+
Memory Storage – Physical
Characteristics

Volatility

Erasable

Power consumption/Heat
+
Memory Storage – Organization

Physical arrangement of bits into words

Not always obvious

Striped across multiple disks
+
Memory Storage – RAID

RAID: Redundant Array of Inexpensive Disks


Combines multiple disks into a logical unit for the purposes of

Data redundancy

Performance Improvement

Or both
Can be implemented by software or hardware

Software: OS controlled

Hardware: Physical RAID controller
+
Memory Storage – RAID Levels

RAID 0


RAID 1


Striped data without parity or mirroring. Performance increase.
Mirrored data without parity or striping. Fault tolerance.
RAID 5 (most common)

Striped data with distributed parity. Requires at least three disks.
Performance increase with Striped data and Fault tolerance. Can
still operate with one failed drive.
+
Memory Storage – Bottom Line
How
much?
 Capacity
How
fast?
 Performance
How
expensive?
+
Memory Storage – Hierarchy List

Registers

L1 Cache

L2 Cache

L3 Cache

Main Memory

Disk Cache

SSD

HDD

Optical

Tape
+
Memory Basics
+
Main Memory Basics

Memory: Where computer stores programs and data

Bit (binary digit): basic unit (8 bits = 1 byte)

Each memory cell (location) has an address numbered 0,…,n-1
(for n memory cells)

Possible address range limited by address size

m bits in address means 2m addresses

Memory cell size (typically 1 byte) grouped together into words
(typically 32 or 64 bits)

32 bit computer will typically have 32 bit registers and
instructions for manipulating 32 bit words

64 bit computer will be similar
+
Semiconductor Memory

Random Access Memory (RAM):

All semiconductor memory is random access

Directly accessed by address logic

Read/Write

Volatile

Requires constant power supply

Temporary storage

Static


Holds data
Dynamic

Periodically refreshes charge
+
Static RAM

Bits stored as on/off switches (transistors)

No charges to leak

Does not need refresh circuits

No refreshing needed when powered

Larger per bit

More expensive

Faster

Example:

Cache Memory:
+
SRAM Illustration

When write enable is high, output is the same as input.

Otherwise, output holds previous input value
+
Dynamic RAM

Bits stored as charge in capacitors (also uses transistors)


Charges leak from capacitors
Needs refreshing, even when powered

Needs refresh circuits

Smaller per bit

Less expensive

Slower

Asynchronous and Synchronous DRAMs

Example:

Main memory
+
DRAM Illustration

When write enable is high, output is the same as input.

Otherwise, output holds previous input value

Includes capacitor refresh circuits
+
Read Only Memory (ROM)

Permanent storage

Microprogramming

Library subroutines

Systems programs

Function tables
+
Measures of Memory Technology

Density

Latency and cycle time
+
Memory Density

Refers to memory cells per square area of silicon

Usually states as number of bits on standard chip size

Examples:



Memory cells typically structured in arrays



1 mb chip
4 mb chip
1Mb x 1 chip
256 Kb x 4 chips
Note: higher density chip generates more heat
+
Memory Packaging: Chips

16 Mbit chip (4M x 4)

WE = Write Enable

OE = Output Enable

RAS = Row Address Select

CAS = Column Adress Select

A0 – A10 = 11 address bits

D1 – D4 = Data to be read/written

NC = No connect, for even # of pins

Vcc = Power Supply

Vss = Ground Pin
+
Read-Write Performance

In many memory technologies, the time required to fetch
information from memory differs from the time required to
store information in memory.

This difference can be dramatic.

Performance is then determined by Read and Write
operations.