Random-access memory (RAM) is memory in which any location can be accessed in a fixed, short amount of time, independent of which location. The opposite would be sequential-access memory (like tape) where accessing a location far down the medium takes longer than accessing one near the start.

The term “random access” is misleading — it doesn’t mean access is random. It means access time is the same regardless of location. You can hop around freely without paying a position-dependent cost.

In modern systems, “RAM” usually refers specifically to the volatile semiconductor main memory of a computer — the DRAM chips on a motherboard. But cache, registers, and even some SSDs are also “random access” in the original sense. Convention is what it is.

Properties

A typical RAM chip:

  • Typically word-addressable internally, exposed as byte-addressable to the processor (each address picks one 8-bit byte). The internal access width varies in practice — DRAM rows are hundreds to thousands of bits wide, and SDRAM bursts return multiple words per command — so the “word-addressable internally” picture is a simplification.
  • Address space of locations for address bits.
  • Access time measured in nanoseconds for current devices — 10 ns for SRAM, 100 ns for DRAM.
  • Volatile: data is lost when power is removed.

Memory is organized so that any location can be retrieved with the same control sequence — apply the address, assert Read, get the data. No “seek” needed.

Two main flavors of semiconductor RAM

  • SRAM — fast (~ ns), uses transistors per cell, expensive per bit. Used for processor registers and cache.
  • DRAM — slower (~ ns), uses transistor + capacitor per cell, much cheaper per bit. Used for main memory.

Both are volatile and both offer true random access. The trade-off is the speed/density tension: more transistors per cell means each cell takes more silicon, so chips of a given size hold less data — but they can access it faster.

What it is not

  • Not ROM (which is non-volatile and not normally writable).
  • Not cache (which is RAM but with hardware logic that auto-fills it from a slower tier).
  • Not secondary storage like SSDs (which use flash, not the standard cell types).

The line between “main memory” and the rest of the Memory hierarchy used to be sharp; with NVDIMM and persistent memory technologies, it’s blurring.