Lightning Talk: Bigger and Smarter: Scale up Your AI Accelerators
DescriptionAI hardware today is facing tremendous challenges in memory bandwidth and data movement. Even with advanced design techniques, such as in-memory computing (IMC), its scaling trend is still lagging behind the ever-increasing demand of AI algorithms. More disruptive innovation, such as 2.5D/3D heterogeneous integration (HI), will be critical to scaling up the system and speeding up the computation. In this talk, we will review the fundamental limitations of current AI accelerators, especially interconnection and memory access. Then we will explore the enormous potential of chiplet-based 2.5D/3D systems, as well as future research needs to drive such paradigm shift.
Event Type
Research Manuscript
TimeWednesday, July 12th1:30pm - 1:40pm PDT
Location3012, 3rd Floor
AI/ML Architecture Design