AI Inference: CPU vs GPU Explained - Optimize Compute & Memory! #shorts

LLM inference has two phases: pre-fill and decode. The pre-fill phase is compute-heavy, while the decode phase is memory-heavy. CPUs can effectively handle the pre-fill, and GPUs work better for the decode. #LLM #inference #CPU #GPU #AI

Видео AI Inference: CPU vs GPU Explained - Optimize Compute & Memory! #shorts канала Red Hat AI