물리colloquium

주제 : Scaling and Innovations of DRAM Technology for the AI-Driven Future

발 표 요 약 :

As artificial intelligence (AI) reshapes computing demands across industries, dynamic random-access memory (DRAM) stands at the forefront of enabling next-generation AI workloads. This talk explores the scaling challenges and innovation pathways critical to advancing DRAM technology in alignment with the exponential growth of AI-driven applications.

In Part I, we examine the physical and architectural limitations of conventional DRAM scaling — including cell capacitor miniaturization, leakage current, and signal integrity — and discuss emerging solutions such as 3D stacking, High Bandwidth Memory (HBM), and computational memory architectures.

In Part II, we explore innovative memory solutions enabled by system-level co-design approaches that enhance bandwidth, energy efficiency, and latency. We also examine how AI-specific workloads — particularly those involving large-scale matrix operations and real-time inference — are driving new DRAM design paradigms, including near-memory computing, in-memory computing, and future advanced packaging solutions.

Together, these insights offer a roadmap for DRAM evolution that balances performance, scalability, and cost to meet the demands of the AI-driven future.

󰏠 연 사 : 박형진 박사 (SK Hynix (DRAM PI))

󰏠 일 시 : 2025년 12월 04일(목요일) 오후 4시 30분

󰏢 장 소 : 자연대 4호관 215호 세미나실

우주소립자연구소

양자기술센터

중성미자정밀연구센터

입자-광자 초정밀 측정 고급인력양성사업팀

전 남 대 학 교 물 리 학 과