Dynamic random access memory (DRAM) has been the memory device of choice for design engineers worldwide for more than 30 years. While the technology has had a phenomenal life, it is running out of steam. The problem is obvious once you think about it--how can you store the same bit-cell charge in an ever-decreasing silicon area?

DRAM designers did what any city planner would do--they went vertical and made capacitors that resembled cylinders rather than plates. But now, with capacitors resembling needles, the ability to continue scaling cost-effectively is unknown. As the industry migrates to technologies below 40 nm, a new type of memory that will allow semiconductors to continue scaling at their current rate will be required.

The good news is that there's a leading contender in resolving this critical challenge--floating-body memories. FBMs use a single transistor for the memory bit-cell, and eliminate the troublesome capacitor completely.

No capacitor required

How a memory can exist without a capacitor? In an FBM, the memory state is stored in the body region of a floating-body transistor, rather than in a separate storage capacitor, as is the case in a DRAM. FBMs take advantage of the floating-body effect inherent in all electrically isolated transistors. Today, virtually all FBMs require an SOI substrate to achieve electrical isolation for each transistor in the memory array. But over the next few years, as the semiconductor industry transitions to 3D devices such as FinFETs, many if not all semiconductor processes will be built on an SOI substrate or on bulk substrates that will exhibit a floating-body effect. As evidenced in the accompanying figures, this opportunity for broad deployment explains the recent and significant increase in development activity surrounding FBMs.