CAMformer: Associative Memory is All You Need

TL;DR

CAMformer introduces an associative memory-based accelerator for transformers, significantly improving energy efficiency, throughput, and area while maintaining accuracy by reinterpreting attention as an associative memory operation.

Contribution

It proposes a novel analog associative memory architecture for transformers, enabling constant-time attention computation and substantial hardware efficiency improvements.

Findings

Over 10x energy efficiency gain

Up to 4x higher throughput

6-8x smaller area compared to existing accelerators

Abstract

Transformers face scalability challenges due to the quadratic cost of attention, which involves dense similarity computations between queries and keys. We propose CAMformer, a novel accelerator that reinterprets attention as an associative memory operation and computes attention scores using a voltage-domain Binary Attention Content Addressable Memory (BA-CAM). This enables constant-time similarity search through analog charge sharing, replacing digital arithmetic with physical similarity sensing. CAMformer integrates hierarchical two-stage top-k filtering, pipelined execution, and high-precision contextualization to achieve both algorithmic accuracy and architectural efficiency. Evaluated on BERT and Vision Transformer workloads, CAMformer achieves over 10x energy efficiency, up to 4x higher throughput, and 6-8x lower area compared to state-of-the-art accelerators--while maintaining near-lossless accuracy.