π€ AI Summary
The surging demand for high-performance DRAM driven by generative AI and hyperscale data centers has led to supply constraints and escalating costs. This work proposes Fiber Memory, a novel architecture that uniquely integrates multi-core fiber, passive optical tap-and-amplify interfaces, co-packaged optics, and all-optical regeneration to repurpose optical fiber as an active recirculating delay line for efficiently storing immutable dataβsuch as large language model weights. By leveraging spatial multiplexing and parallel optical broadcasting, the architecture eliminates redundant weight replication across accelerators. At scales of ten thousand AI accelerators, Fiber Memory reduces energy consumption for weight delivery by over 70% compared to HBM3e-based solutions.
π Abstract
The rising pressure on DRAM availability and contract pricing reflects generative AI's massive high-performance memory requirements. This pressure is heavily compounded by hyperscale data center expansion, which now consumes a significant portion of global DRAM output. In this work, we propose a new architecture: Fiber Memory, which reimagines the role of optical fiber in a hyperscale data center, deploying it as an active, recirculating delay-line memory for immutable data, such as large language model (LLM) weights. We present a data-parallel optical broadcast delay-line memory architecture that accounts for fiber's physical realities. By incorporating space-division multiplexed multi-core fibers (MCFs), passive optical tap-and-amplify interfaces, co-packaged optics (CPO), and regional all-optical regeneration, our case study evaluation demonstrates that Fiber Memory can eliminate redundant weight storage across 10,000 AI accelerators and reduce weight-delivery energy by over 70% compared to traditional HBM3e configurations.