Performance Implications of Multi-Chiplet Neural Processing Units on Autonomous Driving Perception

TL;DR

This paper investigates how multi-chiplet Neural Processing Units can enhance autonomous driving perception systems by improving throughput and utilization, using Tesla Autopilot as a case study and proposing a novel workload scheduling strategy.

Contribution

It introduces a new scheduling approach for deploying perception workloads on multi-chiplet AI accelerators, demonstrating significant performance gains over monolithic designs.

Findings

82% increase in throughput

2.8x higher engine utilization

Effective workload deployment strategy

Abstract

We study the application of emerging chiplet-based Neural Processing Units to accelerate vehicular AI perception workloads in constrained automotive settings. The motivation stems from how chiplets technology is becoming integral to emerging vehicular architectures, providing a cost-effective trade-off between performance, modularity, and customization; and from perception models being the most computationally demanding workloads in a autonomous driving system. Using the Tesla Autopilot perception pipeline as a case study, we first breakdown its constituent models and profile their performance on different chiplet accelerators. From the insights, we propose a novel scheduling strategy to efficiently deploy perception workloads on multi-chip AI accelerators. Our experiments using a standard DNN performance simulator, MAESTRO, show our approach realizes 82% and 2.8x increase in throughput and processing engines utilization compared to monolithic accelerator designs.