skip to main content

Introduction

Artificial intelligence (AI) and its highly active branch of machine learning (ML) are being applied across every major industry, bringing automated intelligent decision-making and processing.

Highlights

  • This report will help enterprises of all kinds that are using or plan to use AI systems to make the right choices about AI hardware acceleration.

Features and Benefits

  • Learn why training deep learning neural network systems is best performed on GPUs.
  • Identify new players in the field.

Key questions answered

  • How important is the availability of data processing tools and development software stack for the accelerator?
  • Are there benchmarks available for assessing the different AI accelerators?

Table of contents

Summary

  • Catalyst
  • Ovum view
  • Key messages

Recommendations

  • Recommendations for enterprises
  • Recommendations for vendors

There is no single best hardware AI accelerator

  • Data processing for ML applications
  • Choosing the right AI accelerator for the task
  • Defining Table 1 entries

Benchmarking AI hardware is essential to move the field forward

  • MLPerf is the first vendor and university benchmarking initiative for AI hardware

Appendix

  • Author