Description
Instructor(s)/Supervisor(s)/Coordinator(s): Yuan XIEThis graduate-level course explores cutting-edge computer architecture in the era of artificial intelligence, with a focus on the hardware foundations that enable modern machine learning workloads. The course will give introduction on AI-centric CPU architecture (using RISC-V as example), GPGPU architecture, and the specialized AI hardware architectures, examining how these architectures balance throughput, efficiency, and scalability for AI workloads. A major emphasis is placed on algorithm-hardware co-design, highlighting how AI models and hardware should be co-optimized for maximum performance and energy and thermal efficiency. We also analyze advanced technology trends shaping the future of AI systems, including 3D integration, emerging memory technologies, and near-memory or in-memory computing. Students will critically evaluate the trade-offs in designing hardware for AI acceleration and explore how technology scaling impacts architectural choices. By the end, students will be able to understand, analyze, and conceptualize AI-centric architectures from both a systems and design perspective. Key Topics covers RISC-V Architecture for AI, GPGPU architecture and parallel execution models, Specialized AI hardware accelerators, Algorithm-hardware co-design strategies, Technology-driven architecture: 3D stacking, HBM, NVM, Near-memory and in-memory computing for AI, Performance/energy/thermal trade-offs in AI system design.