Dimarcia Corporation

• Japanese
• Contact Us
• Site Map

• Home(Dimarcia Corporation)
• About Dimarcia
• About Marching Memory
• About the Inventor
• Our Corporate Philosophy
• Our Business
• CSR

Marching Memory MM is a next-generation computer memory invented and developed by both Professor Tadao NAKAMURA at Stanford University, who is also Dimarcia Senior Adviser, and Professor Michael J. Flynn at the same university. It has a tremendous performance with 1,000 to 8,000 times faster memory access speed compared with that of conventional DRAM (Dynamic Random Access Memory).

Operations Principle

MM has been designed alternatively to the usual DRAM devices so as to avoid the memory bottleneck by reconsidering the functions and structure of memory. Conventional DRAM devices store data in static memory units (locations as words, for example) addressable each with an address. However, MM has a scheme of dynamic memory units within it, where the data are marching column by column.

Conventional DRAMs’ access means is to locate a destination memory unit roaming a memory “town” along with the streets and/or avenues. The structure of this uses lots of wires, like a jangle of wires, to construct these streets and avenues, which means many steps in long time to locate the destination and much more energy consumption due to parallel signal transfer through lots wires. On the other hand, MM has a special scheme of memory construction, where data are marching column by column so that accesses are in parallel and automatically accessible at the end of the MM. The circuitry of the MM seems likely to be very simple and in fact there are no wires to transfer data even in parallel within it.

The most important is how to access memory with tremendously high speed and with less energy consumption in order to avoid the memory bottleneck at a practicable level of computer systems.

In conventional computer systems, memory access as read/write processing requires much longer access time compared with CPU processing speed related to CPU clock. As a result, we cannot obtain enough performance as a whole due to memory bottleneck. On the other hand, MM has no memory bottleneck thanks to its outstanding operating speed of more than 1,000 times faster than that of the conventional memory like DRAMs.

In addition, MM’s capacity size is 100 times as large as that of conventional DRAMs also with 100 times the reliability, only two hundredth times the energy consumption and one third times the fabricating cost owing to its much simpler memory structure.