Full course coming soon

GPU Architecture for AI and Compute Workloads

7 chapters · 31 lessons

1. 1. Parallel Processing Fundamentals

   4 lessons

   Understand why GPUs excel at throughput and how parallelism differs from CPU execution models.

2. 2. GPU Memory Architecture

   5 lessons

   Learn how memory capacity, bandwidth, and bus width determine real-world performance.

3. 3. Data Movement and Bottlenecks

   4 lessons

   Identify and resolve performance bottlenecks caused by data transfer between CPU and GPU.

4. 4. Optimizing AI Inference Workloads

   5 lessons

   Apply GPU memory and compute strategies to model serving, embeddings, and retrieval systems.

5. 5. Training and Fine-Tuning Considerations

   4 lessons

   Understand GPU memory demands during training and how to work within hardware limits.

6. 6. Reading GPU Specifications Strategically

   5 lessons

   Evaluate GPU options by matching specs to workload requirements, not marketing claims.

7. 7. Profiling and Performance Tuning

   4 lessons

   Use profiling tools to measure GPU utilization, memory usage, and identify optimization opportunities.