Multi-DNN Accelerators for Next-Generation AI Systems

TL;DR

This paper discusses the design challenges and architectural considerations for multi-DNN accelerators to support next-generation AI systems across cloud, mobile, and embedded platforms, emphasizing scalability and performance.

Contribution

It introduces new architectural insights and design principles for multi-DNN accelerators to efficiently handle increasing workloads in diverse AI applications.

Findings

Proposes scalable architecture models for multi-DNN processing

Identifies key bottlenecks in current accelerator designs

Suggests optimization strategies for performance and quality-of-service

Abstract

As the use of AI-powered applications widens across multiple domains, so do increase the computational demands. Primary driver of AI technology are the deep neural networks (DNNs). When focusing either on cloud-based systems that serve multiple AI queries from different users each with their own DNN model, or on mobile robots and smartphones employing pipelines of various models or parallel DNNs for the concurrent processing of multi-modal data, the next generation of AI systems will have multi-DNN workloads at their core. Large-scale deployment of AI services and integration across mobile and embedded systems require additional breakthroughs in the computer architecture front, with processors that can maintain high performance as the number of DNNs increases while meeting the quality-of-service requirements, giving rise to the topic of multi-DNN accelerator design.