As a delegate attending the Tech Field Day Exclusive at Intel’s Data-Centric Innovation Day, Ray Lucchesi heard a lot of major announcements from Intel. In this post, he’s focusing on one announcement that didn’t get the most press after the event: DL Boost. This provides a new 8-bit integer instruction that accelerate DL inferencing. This is a major step to stay relevant in the data center, because Intel estimates that more than 50% of AI workloads will involve inferencing before too long. Ray breaks down the technical implementation and where it may run into some potential issues. It’s a great overview.
