AMD Investor Day Presentation Deck

AMD AI ENGINE: ADAPTIVE DATAFLOW PROCESSOR L1 Deep Neural Network (DNN) L2 L3 L4 L5 L6 DNN Runs Optimally on AIE L2 L1 Adaptive Interconnect L3 ↓ L4 L6 L5 Data "flows" from layer to layer, connections between layers are often “sparse" Dataflow architecture, sparsity, efficient datatypes deliver high performance and low power High-Performance, Energy Efficient, and Customizable for Al Workloads

View entire presentation