Present and Future, Challenges of High Bandwith Memory (HBM)
Kwiwook Kim, Myeong-Jae Park
Abstract
In recent years, the demand for fast and high-capacity memory has surged due to the emergence of generative AI models such as GPT. To address this need, High Bandwidth Memory (HBM) has rapidly risen as a critical memory solution. This paper aims to elucidate why HBM plays a pivotal role in the AI industry and discusses the imminent challenges that need to be overcome in the development of HBM memory. By exploring the significance of HBM in the context of AI applications and outlining the pressing challenges ahead, this paper contributes to understanding the importance of HBM in meeting the memory requirements of advancing AI technologies.
Topics & Concepts
Computer scienceContext (archaeology)Cognitive sciencePsychologyBiologyPaleontologyFerroelectric and Negative Capacitance DevicesAdvanced Memory and Neural Computing3D IC and TSV technologies