Joep Boonstra, Synopsys Scientist at Synopsys, presents the “Introduction to Data Types for AI: Trade-offs and Trends” tutorial at the May 2025 Embedded Vision Summit.
The increasing complexity of AI models has led to a growing need for efficient data storage and processing. One critical way to gain efficiency is using smaller and simpler data types. In this presentation, Boonstra explores the trade-offs in data types for AI. He introduces the most commonly used data types, including compact integer and floating-point formats, and highlights their advantages and limitations, as well as their impact on model accuracy and the complexity of the quantization process.
Boonstra examines the main trends in this space, including emerging techniques such as microscaling, and considers the benefits of advanced compression techniques. He concludes by summarizing the key considerations for AI data type selection, including maximizing system throughput and balancing compute and memory bandwidth.
See here for a PDF of the slides.
