94 Part 1 ú Fundamentals Section 2 ú The Computer Space

was not a problem. In other cases a small random-access core memory was added to provide synchronization between the two memories.

The major application of M.cyclic is now for Ms, where the price per bit of disk memory has been decreasing at a rate of 22 percent per year and the price per bit of magnetic tape has been decreasing at 10 percent per year over the last 25 years.

Random-Access Memories (Mp. random)

Random-access memories were used late in the first generation, and they have remained the predominant memory during the second, third, and fourth generations. It is unlikely that their popularity will decline unless content-addressable memories can be constructed sufficiently cheaply (if then). The earliest first-generation random-access memories were electrostatic and depended on maintaining a charge on plates of an array of capacitors. The most common was the Williams tube (invented by F. H. Williams at the University of Manchester), which works id essence like a CRT, with the beam used to charge a capacitor array at the tube face [Williams and Kilburn, 1949]. Other schemes included an array of capacitors that could be selected by digital logic.

Late in the first generation Forrester [1951] invented the core memory, which rapidly became the predominant primary -memory component. In the fourth generation (1972) semiconductor memory (bipolar for speed, MOS for bulk) has become the dominant memory technology.