DRAM, short for dynamic random-access memory, is the high-density memory that serves as main memory in almost every computer, phone, and server. It stores each bit as an electric charge on a tiny capacitor, with a single transistor acting as the switch that reads or writes that charge. Because a capacitor holds a bit using far fewer components than the cross-coupled transistors of SRAM, DRAM packs many more bits into the same silicon area, which is what makes it cheap enough to build memories measured in gigabytes.
The invention is credited to Robert Dennard at IBM, who filed for a patent in 1967 that was granted on June 4, 1968, as US Patent 3,387,286, titled “Field-Effect Transistor Memory.” The patent describes a memory array in which, in the words of its abstract, “each cell is formed, in one embodiment, using a single field-effect transistor and a single capacitor.” That one-transistor, one-capacitor cell is the heart of DRAM and remains the basic structure used today.
The word “dynamic” points to the memory’s central quirk: the charge on each capacitor leaks away over time. Left alone, a DRAM cell would lose its bit within milliseconds. To prevent that, the memory must be refreshed - every cell is read and rewritten on a regular schedule, typically every few dozen milliseconds. This refresh requirement is the price of the dense single-transistor design, and it is why DRAM is “dynamic” while the more complex but self-sustaining SRAM is “static.”
DRAM is also slower to access than SRAM, because reading the small charge on a capacitor through a sense amplifier takes time, and the act of reading is destructive and must be followed by a rewrite. This speed gap is precisely why computers layer SRAM caches in front of DRAM main memory. In the memory hierarchy described by Hennessy and Patterson, DRAM occupies the role of large, cheap, moderately slow main store, with faster cache levels above it and disk below.
The economics of DRAM have shaped computing for half a century. Dennard’s single-transistor cell let memory density roughly track the relentless shrinking of transistors, so the amount of RAM affordable in a typical machine grew enormously decade over decade. Standardized by industry bodies and manufactured in vast volumes, DRAM became a commodity, and its cost-per-bit curve is one of the quiet forces behind the steady growth in what software could assume about available memory.
For all its dominance, DRAM has not escaped its physical roots. Refresh consumes power and bandwidth, the capacitor charge grows harder to manage as cells shrink, and effects such as disturbance between closely packed rows have become security and reliability concerns. Yet the basic cell Dennard patented in 1967 still underlies the main memory of essentially every general-purpose computer in use.