Communications of the ACM
New Type of DRAM Could Accelerate AI
The transistors in the capacitorless DRAM developed by U.S.-based researchers.
Credit: University of Notre Dame
Teams of U.S. and Belgian researchers have proposed a new kind of dynamic random-access memory (DRAM) that can benefit artificial intelligence (AI) applications. This DRAM is made from oxide semiconductors and embedded within the layers above a processor, holding bits hundreds or thousands of times longer than commercial DRAM.
The two transistors only, no capacitor (2ToC) embedded DRAM is fashioned from just two transistors and no capacitor, since a transistor's gate is a natural capacitor, and lets the charge representing the bit to be stored there.
Georgia Institute of Technology's Arijit Raychowdhury says the benefits for AI include the ability to read from a 2ToC DRAM cell without destroying the data and having to rewrite it.
Meanwhile, researchers at Belgium's Interuniversity Microelectronics Center introduced their own 2T0C embedded DRAM with an indium gallium zinc oxide semiconductor, which was particularly effective in controlling leakage that drains the bit.
From IEEE Spectrum
View Full Article
Abstracts Copyright © 2021 SmithBucklin, Washington, DC, USA
No entries found