AQPIM: Breaking the PIM Capacity Wall for LLMs with In-Memory Activation Quantization
This work addresses the memory capacity bottleneck of PIM architectures for LLMs with long-context scenarios, offering a practical solution for efficient deployment.
AQPIM introduces a PIM-aware activation quantization framework using Product Quantization to reduce memory footprint and computational overhead for LLMs, achieving 3.4× speedup over a SOTA PIM approach and drastically reducing GPU-CPU communication that accounts for 90-98.5% of decoding latency.
Processing-in-Memory (PIM) architectures offer a promising solution to the memory bottlenecks in data-intensive machine learning, yet often overlook the growing challenge of activation memory footprint. Conventional PIM approaches struggle with massive KV cache sizes generated in long-context scenarios by Transformer-based models, frequently exceeding PIM's limited memory capacity, while techniques like sparse attention can conflict with PIM's need for data locality. Existing PIM approaches and quantization methods are often insufficient or poorly suited for leveraging the unique characteristics of activations. This work identifies an opportunity for PIM-specialized activation quantization to enhance bandwidth and compute efficiency. We explore clustering-based vector quantization approaches, which align well with activation characteristics and PIM's internal bandwidth capabilities. Building on this, we introduce AQPIM, a novel PIM-aware activation quantization framework based on Product Quantization (PQ), optimizing it for modern Large Language Models (LLMs). By performing quantization directly within memory, AQPIM leverages PIM's high internal bandwidth and enables direct computation on compressed data, significantly reducing both memory footprint and computational overhead for attention computation. AQPIM addresses PQ's accuracy challenges by introducing several algorithmic optimizations. Evaluations demonstrate that AQPIM achieves significant performance improvements, drastically reducing of GPU-CPU communication that can account for 90$\sim$98.5\% of decoding latency, together with 3.4$\times$ speedup over a SOTA PIM approach.