Registers and RAM: Crash Course Computer Science #6

Summary

The video explains how computers store information, starting with a circuit that remembers one bit and building up to a larger, addressable memory system.

Main ideas and concepts

• RAM is working memory. It stores information a computer is currently using, such as a game’s state. In the type of RAM discussed here, stored information is lost when power is removed. Persistent memory, by contrast, can retain information without power.
• Feedback makes memory possible. Unlike circuits that send signals only forward, a circuit with feedback can preserve a value. An OR gate with its output fed back can hold a 1, while an AND gate with feedback can hold a 0. Combining these behaviors in an AND-OR latch creates a circuit that can be set to 1, reset to 0, or retain its previous value when neither input is active.
• Writing and reading: Writing puts data into memory; reading retrieves it. A gated latch adds a single data input and a write-enable signal. When write enable is off, changes to the data input do not affect the stored value. When it is on, the latch can store the input.
• Registers group bits together. Placing latches side by side creates a register, which holds one number. Its width is the number of bits it contains. The video notes that register widths have grown from 8 bits in early computers to 64 bits in many modern ones.
• A grid makes larger memories practical. Arranging latches in rows and columns lets the system select a cell by its row and column, rather than wiring every latch independently. In a 16-by-16 grid, each location has a 4-bit row address and a 4-bit column address. A shared data line can serve the selected cell, while read- and write-enable signals control access.
• Multiplexers select locations. A multiplexer uses an address to activate the corresponding row or column. In the example, two 1-to-16 multiplexers select among the 16 rows and 16 columns.
• Memory modules scale by combining components. Eight 256-bit memory units can operate together to store an 8-bit value at each of 256 addresses, providing 256 bytes of addressable memory. Larger memories use the same principle: smaller memory structures are packaged into increasingly large arrangements.
• RAM means “random access.” Any memory location can be accessed directly, in any order. More locations require more address bits; the video gives 32-bit addresses as enough to address roughly a billion bytes.
• Memory is built in layers of abstraction. A modern RAM module contains many nested arrays of memory cells. The video demonstrates this with a 1980s-era module that holds about 1 megabyte and notes that modern modules can hold gigabytes.
• Different technologies store bits differently. The video identifies SRAM as the latch-based memory built in the lesson. It also mentions DRAM, Flash memory, and NVRAM, which use different circuits or components to store bits.

Method: Building up from a single bit

1. Use feedback in logic gates to make a circuit that can retain a 0 or a 1.
2. Combine the behaviors into a latch with set, reset, and hold states.
3. Add a data input and write-enable signal to create a gated latch.
4. Place multiple latches side by side to form a register.
5. Arrange latches in a row-and-column matrix to reduce wiring and select individual cells.
6. Use address bits and multiplexers to select the desired row and column.
7. Combine memory units so each address holds a multi-bit value.
8. Repeat this scaling and packaging process to build larger RAM systems.

Speakers and sources featured

• Carrie Anne Philbin — host and narrator.
• Crash Course Computer Science — the program presenting the lesson.
• Referenced material and examples: the prior episode’s ALU and logic gates; examples involving games and Minesweeper; and a visual example of a RAM stick and its memory modules.