Nidish Kamath, director of product management for Rambus Memory Controller IP, discusses the company's HBM, GDDR, and LPDDR solutions that address AI training and inference workload requirements. Nidish explains the differences between AI training and inference, highlighting the unique compute and memory needs of each. He will delve into the architectural innovations of HBM, including its 2.5D and 3D memory architecture, and the evolution of HBM memory to meet AI demands. Additionally, Nidish will cover advancements in GDDR and LPDDR memory technologies, emphasizing their roles in AI inference and mobile devices.
