Neural Network Benchmarks

AI, machine learning and deep neural networks are quickly becoming very popular for many different tasks. You can find examples of AI in quite a lot, including a multi-layer perceptron in the Zen architecture itself, helping to enhance predictions. They’re infinitely useful in assisting us, but at the moment larger models (and smaller ones) can be difficult to train. They can take between days and weeks depending on what precisely is being trained. Of course we generally offload that particular task to GPUs and even more purpose-built co-processors, though the CPU can augment that training process. Plus it’s interesting to see how well the 1800X performs here. Keep in mind that AMD doesn’t have an accelerated math library that’s optimized for Zen just yet. The Intel processors were benchmarked with Intel’s math library and their distribution of Python installed, which accelerates deep learning quite a bit. Intel’s CPUs benefit greatly from optimized and accelerated math libraries, so it stands to reason that AMD will as well.