ABSTRACT
The mainstream memory technologies, that is, static random-access memory (SRAM), dynamic RAM (DRAM), and Flash, have achieved tremendous accomplishments in the history of semiconductor industry. However, all these technologies are facing significant scaling challenges when the technology node enters 32 nm and beyond [1], such as high-leakage power consumption, degraded device reliability, and large process variations. Thus, many memory technologies have been invented to provide not only the better performance but also the improved reliability and scalability. Some examples include resistive RAM (RRAM), phase change RAM (PCRAM), and magnetic RAM (MRAM). Generally, they can be categorized as “universal memory,” which features high integration density, nonvolatility, zero standby power, and nanosecond access speed. Because of the small memory cell area, RRAM and PCRAM primarily target storage class memory applications, for example, replacing DRAM as main memory [2-4] and replacing Flash as solid-state drive (SSD) [5], respectively. MRAM, however, primarily focuses on embeddedmemory applications such as L2 and L3 caches [6,7] for its nanosecond read access speed (e.g., 11 ns for 64 Mb MRAM [8]).
