Computer memory: concepts, types, and how it supports computing

The working memory of the computer is an area structured by addresses (in tabular form), which can hold binary words of fixed size. Due to binary addressing, working memory practically always has a 'binary' size (based on powers of 2), since otherwise areas would remain unused.

The main memory of modern computers is volatile, which means that all data is lost after the power supply is switched off - the main reason for this is the technology of DRAMs. However, available alternatives such as MRAM are still too slow for use as main memory. For this reason, computers also contain fixed memory in the form of hard disks or SSDs, on which the operating system and the application programs and files are retained when the power is switched off.

The most common design for use in computers is the memory module. A distinction must be made between different types of RAM. While memory in the form of ZIP, SIPP or DIP modules were still common designs in the 1980s, SIMMs with FPM or EDO RAM were predominantly used in the 1990s. Today, DIMMs with e.g. SD, DDR SD, DDR2 SD, DDR3 SD or DDR4 SDRAMs are primarily used in computers.

The first computers had no working memory, only some registers, which were built with the same technology as the calculator, i.e. tubes or relays. Programs were hardwired ("plugged in") or stored on other media, such as punched tape or punched cards, and were executed directly after being read.

"In 2nd generation computing systems, drum memories served as the main memory" (Dworatschek). In addition, in the early days, experiments were also carried out with rather exotic approaches, such as run-time memories in mercury baths or in glass rod spirals (fed with ultrasonic waves). Later, magnetic core memories were introduced that stored information in small ferrite cores. These were strung in a cross-shaped matrix, with one address line and one word line crossing in the middle of each ferrite core. The memory was non-volatile, but the information was lost when it was read and then immediately written back by the drive logic. As long as the memory was not written to or read from, no current flowed. It is several orders of magnitude bulkier and more expensive to manufacture than modern semiconductor memories.

Typical mainframes in the mid-1960s were equipped with 32 to 64 kilobytes of main memory (for example, the IBM 360-20 or 360-30), and in the late 1970s (for example, the Telefunken TR 440) with 192,000 words of 52 bits each (net 48 bits), or over 1 megabyte.

The core memory as a whole provided sufficient space, in addition to the operating system, to first load the program currently being executed from an external medium into the main memory and to hold all the data. Programs and data are in this model from the point of view of the processor in the same memory, the most common today Von Neumann architecture was introduced.

With the introduction of microelectronics, working memory was increasingly replaced by integrated circuits (chips). Each bit was stored in a bistable switch (flip-flop), which requires at least two, but with control logic up to six transistors and consumes a relatively large chip area. Such memories always consume power. Typical sizes were integrated circuits (ICs) of 1 KiBit, with eight ICs addressed together at a time. Access times were a few 100 nanoseconds, faster than processors clocked around one megahertz. On the one hand, this allowed the introduction of processors with very few registers, such as the MOS Technology 6502 or the TMS9900 from Texas Instruments, which performed most of their calculations in main memory. On the other hand, it made it possible to build home computers whose video logic used part of the main memory as screen memory and could access it in parallel with the processor.

At the end of the 1970s, dynamic random access memories were developed that store the information in a capacitor and require only one additional field-effect transistor per memory bit. They can be built very small and require very little power. However, the capacitor loses the information slowly, so the information has to be rewritten again and again at intervals of a few milliseconds. This is done by external logic that periodically reads the memory and rewrites it back (refresh). Due to higher integration in the 1980s, this refresh logic could be built cheaply and integrated into the processor. Typical sizes in the mid-1980s were 64 KBit per IC, with eight chips addressed together at a time.

The access times of the dynamic RAMs were also a few 100 nanoseconds for inexpensive designs and have changed little since then, but the sizes have grown to a few GBit per chip. Processors are now clocked in the gigahertz range rather than the megahertz range. Therefore, in order to reduce the average access time, caches are used and both the clock rate and the width of the connection of the main memory to the processor are increased (see Front Side Bus).

In June 2012, it was announced that a new smaller and more powerful design for main memory was to be developed with the so-called Hybrid Memory Cube (HMC), in which a stack of several dies is to be used. The Hybrid Memory Cube Consortium was founded specifically for this purpose and ARM, Hewlett-Packard and Hynix are among the members.