Products
Contact Us
English
Menu

Renesas Electronics' Balaji Kanigicherla talks in-memory computing for artificial intelligence

Balaji Kanigicherla, vice president and general manager of the Advanced Mixed Signal and ASIC Solutions Group at Renesas Electronics Corporation, emphasized the need to bridge the gap between memory and computing in the technology industry. This gap refers to the difference between computing requirements (especially in artificial intelligence applications) and storage system efficiency. Kanigicherla emphasized that while there have been significant innovations in computing power, the industry still faces challenges in effectively connecting compute and memory.

He discusses the inefficiencies of the von Neumann architecture, in which the processor and memory are decoupled. In artificial intelligence applications, this separation can create bottlenecks. Kanigicherla suggested that future solutions might involve integrating memory into the package and establishing memory hierarchies with different performance levels.

Collaboration with leading partners like Intel and AMD is critical to addressing these challenges. Renesas Electronics is actively working on improving DRAM technology, transitioning to higher-speed DDR5, and developing innovative interface ICs such as third-generation DDR5 RCD and CKD to enhance memory performance.

In addition, Kanigicherla also envisions a future in which technologies such as CXL (Compute Express Link) will play an important role. CXL can globally expand memory, enabling different processing units to seamlessly share memory, eliminating bottlenecks and improving efficiency.

In summary, Balaji Kanigicherla's insights underscore the industry's commitment to addressing the in-memory computing gap, with a focus on optimizing existing technologies and embracing future innovations like CXL.


Reference source:

embedded.com - Renesas Electronics' Balaji Kanigicherla: Need to bridge compute and memory

Sep 26, 2023|0410
Share: