|
The closer memory is to computer processor cores, the faster the information is processed - this rule is well illustrated by the hierarchy of cache of modern processors, which often has a three-tier structure. Memory is most often placed in separate modules that are installed in the motherboard, and increase the capacity limits of the corresponding interface is specified by a set of economic and technical constraints.
The idea of placing memory in the immediate vicinity of the CPU is not new, but existing technologies do not permit the integration of memory to processors . The first step in this direction, according to Intel engineers was made. At IDF 2011 the company demonstrated and created in collaboration with Micron chips a prototype of a multilevel structure, implying a future integration of memory on a single chip with computing devices.
The development was dubbed as Hybrid Memory Cube. By placing the memory chips in a few levels, the engineers have achieved a significant reduction in energy consumption (up to 7 times compared to current DDR3) and increase the capacity of up to one terabit per second (10 times higher than DDR3). Related Products :
|