Dresden starts with revolutionary supercomputer for AI of the future!

Dresden starts with revolutionary supercomputer for AI of the future!

The Technical University of Dresden (TUD) has taken a significant step in the development of energy-efficient AI systems with the commissioning of its new supercomputer “SpiNNcloud”. This state-of-the-art system, developed under the leadership of Prof. Christian Mayr at the Chair of Highly Parallel VLSI Systems and Neuromicroelectronics, is based on the novel SpiNNaker2 chip, which includes 35,000 chips and over five million processor cores. The main goal of this initiative is to create neuromorphic computing systems that are modeled on the human brain and utilize brain-like principles such as distributed memory and event-driven processing. These approaches enable a significant reduction in energy consumption while increasing performance and flexibility tu-dresden.de reported.

The SpiNNaker2 chip, developed as part of the prestigious EU flagship Human Brain Project, promises real-time processing with latencies under one millisecond. The technology adapts dynamically to complex and changing environments, making it particularly attractive for applications in areas such as smart cities, autonomous driving and the tactile internet. Hector Gonzalez, CEO of SpiNNcloud, describes the implementation of this system as an important milestone for AI development.

Technology and Architecture

The SpiNNaker2 chip offers impressive technical features: With 153 ARM cores, 19 MB on-chip SRAM and 2 GB DRAM, including dedicated hardware for machine learning and neuromorphic accelerators, the system is optimized for resource-intensive tasks. By using the 22nm FDSOI manufacturing process, the chip achieves a tenfold increase in neural simulation capacity per watt compared to the previous SpiNNaker1 generation. These advances were made by open-neuromorphic.org highlighted and enable the use of both traditional and event-based deep neural networks.

The system architecture remains flexible and uses a software-based approach. The independent ARM cores are arranged in a Globally Asynchronous Locally Synchronous (GALS) configuration. This improves energy efficiency and enables fast, real-time processing, which is particularly important for edge AI applications.

Applications and future prospects

SpiNNcloud will be established as part of the AI ​​competence center ScaDS.AI Dresden/Leipzig, which bundles regional big data competencies. The integration of AI into end devices is being driven forward by innovative approaches in neuromorphic computing. Fraunhofer IIS has initiated a project to develop scalable, configurable neuromorphic processor units, which will form the basis for new hardware solutions. These should not only be suitable for resource-intensive AI tasks, but also increase energy efficiency and, in particular, take into account the increasing demand for sustainable technologies, such as iis.fraunhofer.de explained.

With over 60 scientists and 200 employees, ScaDS.AI is funded by the federal government and the Free State of Saxony. Industry partners such as the RAFI Group, Cloud & Heat and Racyics support the project with technological expertise. All in all, the SpiNNcloud represents a groundbreaking development in the field of neuromorphic computing and sets new standards in terms of the energy efficiency and performance of AI systems.