Computer memory is the collection of technologies and structures used by a computer to store information so that processors and other components can access it. In the broadest sense, memory includes many layers, from extremely fast processor registers and CPU caches through main system RAM and various forms of persistent memory. In ordinary consumer usage, however, the word memory most often refers specifically to RAM.
The broader technical concept is more extensive because a computer uses a hierarchy of storage locations with different capacities, speeds, costs and persistence characteristics. Memory can hold instructions, data, addresses, intermediate results, operating-system state and other information needed for computation. The defining idea is that information is represented in a form that the computer's processing system can access according to the architecture and memory-management mechanisms of the machine.
Computer memory is built from physical technologies that represent binary information using electrical, magnetic or other physical states. Modern processor registers are implemented directly within CPU circuitry, CPU caches generally use SRAM, and main system memory is normally implemented using DRAM. Persistent memory technologies include ROM-derived technologies, flash memory and other non-volatile technologies. These technologies differ because physical properties impose different trade-offs. SRAM is fast but relatively expensive and dense compared with DRAM, making it suitable for processor caches. DRAM stores more bits per area but requires periodic refreshing and is therefore used for larger main-memory capacities. Flash retains information without power and provides high density, making it suitable for persistent storage rather than ordinary CPU working memory.
The memory hierarchy exists because no single technology provides simultaneously enormous capacity, extremely low latency, high bandwidth, low cost and persistence. CPU registers are very small but immediately available to execution units. Caches are larger and slower than registers but much faster than main memory. RAM provides substantially more capacity while remaining fast enough for general-purpose computation. Persistent storage provides much greater capacity and retains data without power but has considerably higher latency.
The operating system and hardware attempt to exploit locality so frequently accessed information remains in faster layers. This hierarchy means that the effective performance of a program depends not only on processor speed but also on how efficiently its instructions and data move through registers, caches, RAM and storage-related mechanisms.
RAM is the principal form of main memory in most modern computers. Dynamic RAM stores information in cells associated with capacitive charge and requires refresh operations to preserve the stored state. DDR SDRAM is the dominant family of system-memory technology in contemporary general-purpose computers, with generations such as DDR4 and DDR5 providing increasing bandwidth and other improvements. RAM is volatile, meaning that ordinary system RAM loses its contents when power is removed. During operation, the operating system places active program instructions, application data, caches, buffers and other working information in RAM. The CPU accesses this memory through the processor's memory subsystem, while virtual-memory mechanisms determine how virtual addresses map to physical RAM.
Memory is not limited to RAM. Registers hold operands, addresses and control information directly within the processor and are typically the fastest accessible storage locations available to software execution. CPU caches, commonly divided into levels such as L1, L2 and L3, retain frequently accessed instructions and data so the processor can avoid waiting for main memory on every access. Graphics processors have their own memory hierarchies and may use dedicated graphics memory or shared system memory. Storage controllers, network adapters and other devices can also contain buffers or local memory. These structures are all memory in a broad technical sense, but they differ in purpose, visibility, capacity, volatility and access mechanism.
Computer memory is managed at several levels. The processor architecture defines registers, address spaces and memory-access instructions, while the operating system manages processes, virtual address spaces, physical pages and allocation policies. Hardware implements caches, memory controllers and translation structures that operate beneath the operating system's abstractions. Applications typically do not control physical RAM directly; instead, they request memory through operating-system and runtime mechanisms. The system may move or share pages, cache files, reclaim unused memory and map devices or files into address spaces. This layered management allows software to use memory without needing to know the exact physical organisation of the machine.
Memory capacity and memory performance are separate concepts. A computer can have a large amount of RAM but relatively low memory bandwidth, or a smaller amount of RAM with high bandwidth. Latency, bandwidth, access patterns, cache behaviour, memory-channel configuration and processor architecture all influence actual performance. More RAM primarily increases the amount of information that can remain available without reclaiming or paging memory; it does not automatically make every operation proportionally faster.
Similarly, faster RAM can improve performance in workloads that are memory-bandwidth or latency sensitive, but applications may instead be limited by CPU computation, storage, graphics processing or other components. Memory capacity therefore needs to be considered in relation to workload rather than treated as a universal performance measure.
The historical development of computer memory includes many technologies that have since disappeared from mainstream systems, including mercury delay lines, magnetic-core memory and various specialised semiconductor designs. Semiconductor memory eventually became dominant because integrated circuits could provide increasingly dense, reliable and affordable storage. Modern systems continue to combine several technologies because their physical characteristics remain different. The word memory consequently has both a narrow and broad meaning: consumer specifications commonly use it as shorthand for RAM, whereas computer architecture uses memory to encompass a hierarchy of registers, caches, main memory and other accessible storage structures. Distinguishing these meanings is essential when discussing computer memory accurately.
