Week 2: Silicon Architecture – GPUs, TPUs, and Custom Chips
Understanding the hardware that powers AI at scale
Time Estimate: 3-4 hours
Topics Covered
- GPU vs TPU vs custom ASIC architectures
- Tensor cores and specialized AI accelerators
- Memory bandwidth bottlenecks and optimization
- NVIDIA Blackwell and Google TPU architectures
- Future of AI hardware design
Featured Speaker
JH
Jensen Huang
CEO & Founder, NVIDIA
Learn from industry leaders who are building the future of AI infrastructure and applications.
Video Resources
📹 Video content will be added here by Agent 2
Videos include keynotes, technical talks, and tutorials from industry leaders.
Reading Materials
📚 Reading list will be added here by Agent 3
Research papers, blog posts, and technical documentation.
🛠️ Hands-On Lab
Design a GPU Cluster Network
Intermediate 3 hoursObjective
Design a Clos network topology for a 256-GPU cluster and understand datacenter networking fundamentals.
Prerequisites
- Basic networking knowledge (IP, bandwidth)
- Understanding of network topologies
- Spreadsheet or diagramming tool
Setup Instructions
- Install draw.io or use Lucidchart
- Review Clos network topology basics
- Open the provided network calculator spreadsheet
Tasks
- Design a 3-tier Clos network for 256 GPUs with 8 GPUs per server
- Calculate bisection bandwidth requirements
- Compare InfiniBand (400 Gbps) vs RoCE (100 Gbps) costs
- Estimate switch count and cabling requirements
- Calculate total network cost and identify bottlenecks