Good things come in small packages: Should we build AI clusters with Lite-GPUs?

TL;DR

This paper explores the potential of large clusters of small, Lite-GPUs with advanced optics for scalable, cost-effective, and efficient AI workloads, challenging traditional large GPU designs.

Contribution

It proposes a new cluster design using Lite-GPUs with co-packaged optics, highlighting benefits and addressing system challenges for scalable AI computing.

Findings

Lite-GPU clusters reduce manufacturing costs.

High-bandwidth optics enable efficient communication among Lite-GPUs.

Lite-GPU clusters improve scalability and power efficiency.

Abstract

To match the blooming demand of generative AI workloads, GPU designers have so far been trying to pack more and more compute and memory into single complex and expensive packages. However, there is growing uncertainty about the scalability of individual GPUs and thus AI clusters, as state-of-the-art GPUs are already displaying packaging, yield, and cooling limitations. We propose to rethink the design and scaling of AI clusters through efficiently-connected large clusters of Lite-GPUs, GPUs with single, small dies and a fraction of the capabilities of larger GPUs. We think recent advances in co-packaged optics can enable distributing AI workloads onto many Lite-GPUs through high bandwidth and efficient communication. In this paper, we present the key benefits of Lite-GPUs on manufacturing cost, blast radius, yield, and power efficiency; and discuss systems opportunities and challenges around resource, workload, memory, and network management.