AI-Optimized Semiconductor Architectures: The Convergence of Deep Learning and Custom Hardware Design

Authors: Karthik Wali

DOI: https://doi.org/10.37082/IJIRMPS.v8.i4.232593

Short DOI: https://doi.org/g9q392

Country: USA

Full-text Research PDF File:   View   |   Download

Abstract: The integration of AI with the semiconductor design has transformed computing paradigms in a way that benefits from specifically optimised design for deep learning. This paper aims at discussing Artificial Intelligence optimized semiconductor structures where authors apply deep learning approaches to optimize hardware systems. Thus, through the employment of present-day NAS techniques and reinforcement learning, the synthesis of semiconductor circuit diagrams and their layout, power-consumption, and computing capabilities can be automatically and optimally designed. We consider different HW accelerators for AI applications: Google’s TPUs, neural networks based on GPUs from NVIDIA, and application-specific programmable devices known as FPGAs. Finally, it is crucial to discuss further the role of AI in the design of semiconductors, growth strategies and trends, methods, and challenges and prospective improvements based on case studies, comparisons and benchmarks. The findings showed that various optimisations using AI lead to increased performance-per-watt efficiency, speed of computation, and the ability to support AI workloads. Last of all, we explore how AI maintains its imprint on semiconductor architectures and other factors concerning the future of the computing platform.

Keywords:

Paper Id: 232593

Published On: 2020-07-03

Published In: Volume 8, Issue 4, July-August 2020

Share this