TI's integrated TinyEngine NPU can run AI models with up to 90 times lower latency and more than 120 times lower energy utilization per inference than similar MCUs without an accelerator. New ...