Scale-out Systolic Arrays

TL;DR

This paper introduces Scale-out Systolic Arrays (SOSA), a new multi-pod inference accelerator that optimizes array granularity, interconnect topology, and tiling to significantly improve throughput and efficiency in DNN inference workloads.

Contribution

The paper proposes a novel multi-pod systolic array design with optimized array size, scalable Butterfly interconnects, and a custom tiling scheme, outperforming existing accelerators.

Findings

Achieves up to 600 TeraOps/s effective throughput.

Outperforms state-of-the-art multi-pod accelerators by 1.5x.

Identifies suboptimal array sizes in current commercial accelerators.

Abstract

Multi-pod systolic arrays are emerging as the architecture of choice in DNN inference accelerators. Despite their potential, designing multi-pod systolic arrays to maximize effective throughput/Watt (i.e., throughput/Watt adjusted when accounting for array utilization) poses a unique set of challenges. In this work, we study three key pillars in multi-pod systolic array designs, namely array granularity, interconnect, and tiling. We identify optimal array granularity across workloads and show that state-of-the-art commercial accelerators use suboptimal array sizes for single-tenancy workloads. We, then evaluate the bandwidth/latency trade-offs in interconnects and show that Butterfly networks offer a scalable topology for accelerators with a large number of pods. Finally, we introduce a novel data tiling scheme with custom partition size to maximize utilization in optimally sized pods. We propose Scale-out Systolic Arrays, a multi-pod inference accelerator for both single- and multi-tenancy based on these three pillars. We show that SOSA exhibits scaling of up to 600 TeraOps/s in effective throughput for state-of-the-art DNN inference workloads, and outperforms state-of-the-art multi-pod accelerators by a factor of 1.5x.