NHR PerfLab seminar talk on April 15, 2025

Speaker: Johannes Partzsch, TU Dresden

Title: Neuromorphic Hardware – A System Perspective

Slides: https://hpc.fau.de/files/2025/04/2025-04-PerfLab_NeuroHardware.pdf

Abstract

Neuromorphic computing has received interest recently for its potential to counteract the ever-increasing energy draw of AI applications. Taking inspiration from the human brain, it has shown significant improvements in energy efficiency on AI-related tasks. Still, only very recently, neuromorphic computing approaches have been able to solve more challenging computational problems. Key to this development is the availability of efficient, yet robust and flexible neuromorphic hardware systems on a larger scale.

In my talk, I will introduce some of the key principles of neuromorphic hardware systems and discuss their practicability. I will give an overview of the SpiNNaker2 system currently deployed at TU Dresden, which is one of the few large-scale neuromorphic systems existing today. Finally, I will give an outlook on potential future developments in neuromorphic hardware and how they link to recent trends in the field of artificial intelligence.

Short Bio:
Johannes Partzsch is group leader at the chair of Highly-Parallel VLSI Systems and Neuromicroelectronics (HPSN) at Technische Universität Dresden. He obtained his PhD in 2014 at the same institution, focusing on connectivity in neuromorphic systems. His research interests include neuromorphic hardware development, numerical function accelerators and edge AI hardware. He has been leading the ZEN project team, winning first prize in the BMBF AI hardware competition in 2021. He coordinates the construction of the SpiNNaker2 neuromorphic supercomputer "SpiNNcloud" at TU Dresden.

For a list of past and upcoming NHR PerfLab seminar events, see: https://hpc.fau.de/research/nhr-perfl...