Machine Learning (ML) techniques in real-time control are emerging, with applications such as arc fault detection and motor fault detection. Artifical Intelligence (AI) accelerators on-chip are becoming common to run embedded AI models. However, ML performance on real-time control CPUs is also an important consideration. Machine Learning Benchmarks shows benchmarks of 3, 4, and 5-layer Computational Neural Networks (CNN) on a Cortex-M7 MCU and the C29 based F29H85x. The C29 is almost five times faster than the Cortex-M7, even with the latter operating at twice the CPU frequency.