#hust

Workload-Driven HBF Substrate for Capacity-Scalable LLM Inference⭐8

Workload-driven HBF substrate enables capacity-scalable LLM inference with 3.2x bandwidth gains and 40% lower energy.