This paper examines some aspects of the architecture of embedded DRAM (dynamic random access memory), its applications and its advantages over other conventional memory types. Embedded DRAM (eDRAM) – the concept of merging DRAM with logic on a single device – has become increasingly popular, thanks to the growth of existing and emerging high bandwidth applications such as graphics processing, backbone and access router data communications systems and base stations for mobile phones. A common requirement of all these designs is that they have to process very large amounts of data at very high speeds. Because of this, a fundamental design requirement is the ability to provide high performance, high speed memory access. One way of achieving this is to use system-on-chip (SoC) solutions that incorporate embedded DRAM, allowing wide on chip buses to connect logic to DRAM on the same die, rather than to external memory. Furthermore, integration of DRAM directly into an LSI device has the added benefits of minimising system power consumption, saving board space, reducing component count and, thanks to the elimination of external buses, reducing the effects of EMI. eDRAM is used in areas that require high memory bandwidth like in graphics accelerators and media-oriented vector processors like the Vector IRAM (VIRAM).