dynamic random-access memory

dynamic random-access memory
Name	dynamic random-access memory
Caption	A typical DDR4 SDRAM module
Inventor	Robert Dennard
Invented	1966
Manufacturer	Samsung, SK Hynix, Micron Technology

dynamic random-access memory is a type of volatile memory that stores each bit of data in a separate capacitor within an integrated circuit. The capacitor can be either charged or discharged, representing the two values of a bit. Because the charge in these capacitors leaks away, the data must be periodically refreshed, which is the origin of the term "dynamic." This technology forms the primary working memory for modern computing devices, from personal computers to supercomputers, and is a foundational component of the semiconductor industry.

History

The invention is credited to Robert Dennard at the IBM Thomas J. Watson Research Center in 1966, with a patent granted in 1968. Early commercial use began with Intel's 1103 chip in 1970, which replaced magnetic-core memory in systems like the HP 9800 series. The 1980s saw the rise of the Japanese semiconductor industry, with companies like NEC and Fujitsu driving innovation, while the PC revolution, led by the IBM Personal Computer, created massive demand. The transition from asynchronous DRAM to synchronous DRAM in the 1990s, pioneered by Samsung Electronics, marked a major performance leap. Subsequent generations, including DDR SDRAM, have been standardized by JEDEC and fueled continual advancement in systems from the PlayStation 2 to the Cray XT5.

Operation

Fundamental operation relies on the periodic reading and rewriting of data to counteract charge leakage from the storage capacitors. This refresh operation is typically managed by an external memory controller or an on-die counter circuit. A read operation is destructive; after sensing the charge state via a sense amplifier, the data must be written back immediately. The entire process is synchronized to a clock signal in modern types, with timing governed by protocols like those from JEDEC. The cycle of RAS and CAS signals is central to accessing specific memory cells within the array.

Structure and function

The basic storage cell, known as the 1T1C cell, consists of one transistor and one capacitor, fabricated on a silicon wafer using CMOS processes. Billions of these cells are arranged in a grid pattern, accessed by word lines and bit lines. The sense amplifier is critical for detecting minute charge differences. Major structural advancements include the development of trench capacitor and stacked capacitor technologies by firms like IBM and Toshiba. The physical packaging has evolved from DIP and SIMM modules to modern DIMM and SO-DIMM forms used in Apple's MacBook Pro and Dell XPS systems.

Types and generations

Major categories include Asynchronous DRAM, last used in early systems like the Commodore 64, and Synchronous DRAM, which became dominant in the 1990s. The main lineage is DDR SDRAM, with successive generations DDR2 SDRAM, DDR3 SDRAM, DDR4 SDRAM, and DDR5 SDRAM, each doubling data rates. Specialized types include Graphics DDR SDRAM for NVIDIA GeForce cards, Low Power DDR for Qualcomm Snapdragon smartphones, and High Bandwidth Memory for AMD Radeon GPUs and the Fujitsu A64FX processor in the Fugaku supercomputer.

Applications and market

It is the universal main memory in computing, essential for the operation of Microsoft Windows, Linux, and macOS on platforms from Lenovo laptops to AWS cloud servers. The market is dominated by the "Big Three" manufacturers: Samsung Electronics, SK Hynix, and Micron Technology, with production centered in South Korea and Taiwan. Demand is driven by data centers for Google and Meta, 5G smartphones from Apple and Samsung, and emerging technologies like artificial intelligence and autonomous vehicles. Pricing is famously cyclical, influenced by supply dynamics and trade policies between the United States and China.

Comparison with other memory types

In contrast to static random-access memory, it is slower and requires refresh but offers much higher density and lower cost per bit, making it suitable for main memory while SRAM is used for CPU cache in Intel Core processors. Compared to non-volatile memory like NAND flash used in SSDs from Western Digital, it is vastly faster but loses data without power. Emerging technologies like Storage-class memory, such as Intel's 3D XPoint, aim to bridge this performance-storage gap. It is also distinct from read-only memory, which is non-volatile and used for firmware like the BIOS in HP computers.

Category:Computer memory Category:American inventions Category:Semiconductor devices