€3m ViTfox project to develop 50TOPS/W edge AI vision chip
A joint European-Korean project is developing ferroelectric memory devices for in-memory compute to cut the power consumption of edge AI vision chips.
The three year Vitfox project, coordinated by the Greek National Centre for Scientifc Research, will develop a detailed simulator to refine circuits as part of the design and fabrication of an Vision Transformer AI chip with an energy efficiency of 50TOPS/W. This will use 3D hafnium-zirconium oxide (HZO) ferroelectric memory devices and the chip will be used in a collaborative hardware-software platform
he platform will support two types of emerging memories, high-density 3D FeRAM developed in Korea and epitaxial ferroelectric tunnel junctions developed in Europe and the European Union is funding the project with €1.5m with the other half from the National Research Foundation of Korea (NRF).
- Innatera claims first mass market neuromorphic microcontroller
- Hailo ports transformer AI to embedded chip
The consortium of eight partners includes Fraunhofer IPMS in Germany, ETH Zurich, Silicon Austria Labs (SAL), Seoul National University as well as Hanyang, Soongsil and Kookmin Universities in Korea. Fraunhofer IPMS is working with NY Creates in the US to build ferroelectric devices on 300mm wafers.
Vision Transformers (ViTs) are neural networks that outperform traditional methods in image recognition. This project focuses on developing a ViT using ferroelectric oxide materials for AI edge applications.
“We aim to push the boundaries of current technology by developing hardware-software co-optimization platforms, novel materials, and integration methods that will not only enhance AI performance but also ensure sustainability in energy consumption, “ says Prof. Dr. Thomas Kämpfe, project leader at Fraunhofer IPMS, one of the partners in the consortium. “We want to significantly contribute to the semiconductor industry, addressing both the technical challenges of emerging memory technologies and the societal need for efficient computing solutions,“ he said.
This is one of four joint projects in a €12m programme co-funded by the Chips JU under Horizon Europe and the NRF. The other projects are developing brain-like circuits using two-dimensional materials to create energy-efficient AI systems (Energize), and a laser-based LIDAR system called Nehil that integrates various technologies for precise distance measurement. The Haetae project is developing photonic brain-like chips that efficiently process AI tasks and can adapt to new functions.
www.vitfox.eu; www.demokritos.gr/
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
