RAM

Photo: Franck V. on Unsplash

In short: “Random Access Memory” — a computer’s volatile main memory, where currently running programs and data sit for fast access by the CPU.

In more detail: Unlike SSDs/HDDs, RAM loses its content when switched off (volatile memory), but is many times faster in return. More RAM allows more programs to be open at once, without the system having to swap out to slower mass storage.

In Depth

Why RAM is volatile

RAM stores data in a huge matrix of tiny capacitors (in the common DRAM design), each of which is either charged (1) or discharged (0) — this is exactly what makes RAM both fast and volatile: a capacitor can be electrically read and written extremely quickly, but loses its charge on its own within milliseconds without constant refreshing (“refresh”, several thousand times per second). Without continuous power supply, this refreshing no longer happens, which is why the content is irretrievably lost when switched off. This distinguishes DRAM from the rarer, more expensive SRAM (Static RAM, used among other things for CPU cache), which doesn’t need constant refreshing and is therefore even faster, but needs considerably more chip area per bit.

Random access as a speed advantage

The decisive speed advantage over SSDs/HDDs lies in “random access” (hence the name): RAM can access any storage cell in practically constant, extremely short time, regardless of where in memory it sits — mass storage, even SSDs, is by contrast noticeably slower for very many small, scattered accesses. If the available RAM isn’t enough, the operating system swaps rarely used data out into a “swap file”/pagefile on the slower mass storage — a system with too little RAM becomes noticeably slower as a result, since it constantly has to shuffle data back and forth between RAM and mass storage (“thrashing” in extreme cases).

Generations and clock rates

Like many hardware components, RAM generations have advanced considerably: DDR3 (from around 2007, typically 1333-2133 MT/s), DDR4 (from 2014, typically 2133-3200 MT/s, the standard for a long time) and DDR5 (from 2021, typically 4800-6400+ MT/s, with higher efficiency through internal splitting into two independent channels per module). The generations are electrically and physically incompatible with each other — a DDR4 module doesn’t fit into a DDR5 RAM slot and vice versa; the notches in the contact strips are deliberately placed at different positions, to prevent incorrect insertion.

Dual-channel and capacity

Most modern motherboards support “dual-channel” operation: if two (or four) identical RAM modules are used instead of a single large module, the CPU can access both channels in parallel and nearly double the effective memory bandwidth. That’s why it’s often recommended to buy, say, 2×8 GB instead of 1×16 GB. As a rough guide for capacity: 8 GB is now considered a tight minimum, 16 GB a solid everyday standard, and 32 GB or more makes sense for video editing, virtual machines or very memory-hungry applications.

See also: RAM Slots, SSDs, CPU, Motherboard/Mainboard