AI021
CUDA Programming Guide
Advanced Memory Logic and Multi-GPU Systems
Learning Objectives
- Analyze the logic layer and performance characteristics of High Bandwidth Memory (HBM3).
- Evaluate memory consistency challenges in distributed multi-GPU environments.
- Master the architectural design of point-to-point and switched interconnect topologies like NVLink and Infinity Fabric.