27 November 2020

Exploiting parallelism in NN workloads to realize scalable, high performance NN acceleration hardware

Conventional wisdom says to execute a huge computation task like AI you need a single huge processor. In our latest white paper, Tony King-Smith explains how by understanding the inherent parallelism in NN algorithms used in automotive AI systems, multiple smaller processors can deliver the same throughput with superior scalability, flexibility and robustness.