The AMD Deep Learning Processor Unit (DPU) is a series of soft IP dedicated for convolutional neural network acceleration. The DPUCAHX8L is a low-latency CNN inference IP for AMD Alveo™ cards with high bandwidth memory (HBM). At a high-level, the DPU is a microcoded compute engine which has an efficient, optimized instruction set, and which can support inference of most convolutional neural networks.