Accelerating LLM Inference Throughput via Asynchronous KV Cache Prefetching

Large Language Models (LLMs) exhibit pronounced memory-bound characteristics during inference due to High Bandwidth Memory (HBM) bandwidth constraints. In this paper, we propose an L2 Cache-oriented asynchronous KV Cache prefetching method to break through the memory bandwidth bottleneck in LLM inference through computation-load overlap. By strategically scheduling idle memory bandwidth during active computation windows, our method proactively prefetches required KV Cache into GPU L2 cache, enabling high-speed L2 cache hits for subsequent accesses and effectively hiding HBM access latency within computational cycles. Extensive experiments on NVIDIA H20 GPUs demonstrate that the proposed method achieves 2.15× improvement in attention kernel efficiency and up to 1.97× end-to-end throughput enhancement, surpassing state-of-the-art baseline FlashAttention-3. Notably, our solution maintains orthogonality to existing optimization techniques and can be integrated with current inference frameworks, providing a scalable latency-hiding solution for next-generation LLM inference engines.

Authors

• Yanhao Dong Alibaba Cloud
• Yubo Miao Alibaba Cloud
• Weinan Li Alibaba Cloud
• Xiao Zheng National University of Defense Technology
• Chao Wang
• Jiesheng Wu Alibaba Cloud
• Feng Lyu Central South University

Keywords

No keywords are indexed for this paper.