Produkt-Neuheit

Sensor systems based on conventional computer architectures store and process data in different locations, which requires moving and processing the measurement data in different parts of the system, resulting in higher consumption of time and energy (“Von-Neumann-Bottleneck”). Neuromorphic computing solves this problem by omitting the laborious data transfer between processor and memory, which minimizes latency times as well as power consumption for complex calculation and transfer processes. This makes the technology interesting especially for AI-applications.
Within the TEMPO project, the Fraunhofer EMFT scientists have developed an ASIC with computing-in-memory architecture, combining computation and memory. Utilizing the neuromorphic computing approach it enables local computation of the measurement data, resulting in shorter response times and higher energy-efficiency.