Dynamic random-access memory

Computer memory used to be built like furniture. Early semiconductor memory cells held each bit with several transistors, while magnetic core memory filled cabinets with tiny woven rings. Dynamic random-access memory changed the economics by accepting something earlier designers had treated as failure: stored charge leaks away. Robert Dennard's 1967 IBM design used one transistor and one capacitor per bit, then relied on periodic refresh to restore the fading signal. Memory became dynamic because forgetting, briefly and controllably, turned out to be cheaper than permanent vigilance.

That bargain only became possible once the `mosfet` existed as a manufacturable switch. MOS devices made it practical to gate access to a tiny capacitor without spending the area budget of older multi-transistor cells. `resource-allocation` captures the core insight. DRAM sacrificed stability in each individual cell so that the chip could devote far more surface area to density. Instead of paying six-transistor prices for every bit, designers paid a refresh tax at the system level and won orders of magnitude more memory per chip.

`niche-construction` then made the leaky cell usable. A raw capacitor that forgets is not a product. DRAM required sense amplifiers, multiplexed row-and-column addressing, refresh circuitry, and controller logic that revisited cells before their charge decayed too far. The invention was therefore not just a clever cell. It was a habitat built around a fragile component. IBM supplied the originating idea, but `mos-dram` turned that architecture into an industrial standard once Intel's 1103 reached the market in 1970 and proved that dense semiconductor memory could displace magnetic core in commercial machines.

From there the cascade ran straight into computing's next expansion. Cheap working memory let the `microprocessor` become more than a calculator brain. Programs could grow larger, operating systems could keep more state in active memory, and personal computers could become general-purpose machines rather than tightly constrained controllers. Later devices such as the `digital-camera` also depended on DRAM's density because image sensors produce data in bursts that must be buffered, processed, and moved before permanent storage catches up. DRAM sat in the middle of that flow as the fast, disposable workspace of the digital age.

`path-dependence` explains why DRAM remained so central even as its weaknesses never disappeared. It still needed refresh. It still lost data when power failed. It still demanded elaborate manufacturing and yield discipline. Yet once computer architectures, software expectations, and supply chains were organized around abundant volatile memory, the whole industry learned to build around those costs instead of replacing the technology outright. By the 1980s, memory leadership had become a strategic contest among American, Japanese, and later Korean producers. IBM invented the architecture, Intel commercialized the first big win, and later `samsung-electronics` mastered scale so completely that DRAM became one of the defining commodity battles of modern electronics.

Dynamic RAM did not triumph because it was elegant in isolation. It triumphed because it moved complexity to the place semiconductor manufacturing could afford it. A memory cell that forgot on its own sounded like a defect. In practice it was the compromise that made modern computing affordable.