Semiconductor leaders advance Edge AI and 3D DRAM architectures

The semiconductor and memory industries are advancing new architectures to meet the increasing demands of Artificial Intelligence. Intel and Innodisk have developed a reference platform for Edge AI applications using Intel Core Ultra Series 3 processors. This platform utilizes heterogeneous computing, distributing tasks across the CPU, GPU, and NPU to optimize performance and reduce latency for devices like robots and cameras at the network edge.

Simultaneously, major memory manufacturers including Samsung Electronics, SK Hynix, and Micron are shifting toward three-dimensional (3D) memory structures to overcome the ‘memory wall’ that limits AI infrastructure efficiency. Samsung has introduced its ‘CUBE’ strategy, focusing on capacity, utilization, bandwidth, and efficiency through vertical transistor architectures. This approach aims to significantly increase memory density while reducing heat generation. SK Hynix is pursuing monolithic 3D DRAM integration by connecting memory directly with logic circuits.