Hailo-8 vs Intel Nervana Neural Network Processor [EOL] comparison

The Hailo-8 processor includes memory and runs at up to 26 tera operations per second (TOPS), and when running the ResNet-50 benchmark consumes 1.67 watts and achieve 2.8TOPS/W. Hailo offers a software development kit (SDK) that was co-developed with the hardware.

The Intel Nervana Neural Network Processor (NNP) is a purpose built architecture for deep learning. This new architecture provides the needed flexibility to support all deep learning primitives while making core hardware components as efficient as possible.

Deep learning is taking the next leap forward. Increasingly sophisticated and complex data, models, and techniques will allow AI to move beyond identifying information to understanding context – enabling a level of “common sense” for reasoning and decision-making.