Turbo Computing with Light Saves Time and Energy

Artificial intelligence, complex logistics, or energy systems management are often limited by computing time and energy consumption. A test system for optical computing has begun operation at the Karlsruhe Institute of Technology (KIT). Its aim is to make AI and other applications with high computing demands considerably faster and more energy-efficient. To meet the constantly increasing demand for computing power despite limited energy supplies, the technology processes information using light instead of electrical signals.

In the new system, photonic processors augment traditional supercomputers. In contrast to conventional chips, they process information with light instead of electric signals, making them especially efficient at performing certain computing operations – a promising approach for artificial intelligence and other data-intensive applications.

Now KIT researchers are working to find the applications with which the optical processors can make the most of their strengths, and how they can be integrated into existing computing infrastructures as accelerators. “Access to photonic processors gives us new ways to design applications with high computing demands,” said Professor Martin Frank, Director of KIT’s Scientific Computing Center (SCC). “We’re combining classical computers, quantum computers, and light-based systems with the aim of doing each part of a job on the most suitable hardware.”

Gaining Efficiency by Combining the Strengths of Different Systems

“Hybrid solution strategies are one focus of our research. If we can distribute complex calculations efficiently over the available types of hardware, we can get faster results overall while using less energy,” added Professor Ina Schaefer from KIT’s Institute of Information Security and Dependability (KASTEL). “And we’re presenting open-source results so other researchers can benefit from them.”

Germany’s Federal Ministry of Research, Technology and Space is funding the development of the test system with about EUR 4.3 million. The hardware was procured as part of the QuSol quantum computing project and was developed and produced by Q.ANT, a company based in Stuttgart.