Memory: From Relays to DRAM

Memory: From Relays to DRAM

This page traces the evolution of memory technology: how information was stored, from mechanical relays to modern transistor-based RAM. It explains why each step was necessary and what trade-offs were involved.

What Is Memory?

In computer science, memory is any container that holds a pattern (0/1) until it is changed.

• Everyday example: a light switch. The switch remembers its position until flipped.
• In computers: voltage (electrical charge) represents a bit.
• Memory is essential → without it, a CPU would have no place to store instructions or data.

Early Physical Memory

Before electronics, patterns were stored in physical form:

• Punch cards – holes punched into paper represent bits. Once punched, they stay.
• Delay lines – sound waves or torsion pulses in a medium (wire, mercury tube) hold information briefly.
• Drum memory – rotating magnetic drum; data stored on the surface, accessible once per rotation.

Limitation → access was sequential or slow. Not true “random access.”

Relays: Electrically Controlled Switches

• A relay = coil + mechanical contacts.
• Electricity controls whether the contacts are open or closed.
• Multiple relays can:

 * Store bits,
 * Perform logic (AND, OR, NOT),
 * Form early computers (e.g., Zuse Z3 in 1941).

Advantages → electrical control, reusable, programmable.

Drawbacks → mechanical movement = slow (few Hz), wear-out over time.

Vacuum Tubes: Purely Electronic Switches

• Tubes switch or amplify signals without moving parts.
• Much faster than relays.
• Enabled the first electronic computers (ENIAC, 1940s).

Advantages → high speed.

Drawbacks → expensive, power-hungry, unreliable for main memory.

Magnetic Core Memory

• Small ferrite rings (cores) store bits by magnetizing clockwise or counter-clockwise.
• Non-volatile (keeps information without power).
• Widely used 1955–1965; even in spacecraft until the 1980s (radiation resistant).

Advantages → reliable, non-volatile.

Drawbacks → bulky, expensive compared to semiconductors.

Transistors: The Turning Point

• Replaced tubes in the 1950s.
• Tiny, cheap, low power.
• Today’s memories are based on field-effect transistors (FETs).

Key effect → allowed integration of thousands → billions of transistors on a single chip.

Memory Cells

A memory cell stores a single bit. There are two main designs:

Static RAM (SRAM)

• Uses 6 transistors (6T cell).
• Stable latch stores one bit indefinitely (as long as powered).
• Fast (nanoseconds), but:

 * Large (takes space),
 * Power-hungry,
 * Expensive.

Used mainly for caches (L1/L2/L3).

Dynamic RAM (DRAM)

• One transistor + capacitor (1T1C).
• Stores charge as “1”, absence as “0”.
• Needs refresh (capacitor leaks in nanoseconds).
• Slower than SRAM, but:

 * Very dense,
 * Much cheaper.

Used for main memory (MB → GB → TB).

Black-box model of a cell

• Select line chooses the cell.
• Read/Write line decides the operation.
• Data in/out carries the bit.

Scaling Up Over Time

• 1963: first integrated memory cell.
• Early 1970s: SN7489 chip, 64 bits.
• 1986: first 1-Mbit DRAM chip.
• 2018: 128-Gbit SDRAM.
• Continuous miniaturization (Moore’s Law) → exponential growth.

Case Study: How Much RAM Do We Need?

• Atari 2600 (1980): only 128 bytes RAM.
• Entire game logic + graphics + sound fit into 4 KB ROM cartridges.
• Clever programming (reusing data, procedural graphics) made this possible.

Summary

• Memory evolved from physical storage to electronic semiconductors.
• Each step traded cost, speed, density, and reliability.
• Modern computers use a hierarchy:

 * SRAM caches (fast, small, expensive),
 * DRAM main memory (slower, large, cheap),
 * Non-volatile storage (SSDs, HDDs).