Researchers at the Laboratory of Nanoscale Electronics and Structures (LANES) at the Ecole Polytechnique Fédérale de Lausanne (EPFL) in Switzerland have developed the world's first large-scale in-memory processor, a university press release said. The processor aims to redefine energy usage in data processing and make it energy efficient.
Modern-day information technology systems are well known for producing large amounts of heat. Heat reduction is a more efficient way of using energy and will also help the world reduce carbon emissions, as it aims to go greener in the coming few decades. To minimize this unwanted heat, one must go to the root of the problem, the von Neumann architecture.
Contemporary computing architecture separates the information processing center from the storage ones, and most of the energy is spent in shuttling bytes of information between the memory and the processor. First described at Princeton by John von Neumann and others in 1945, the architecture made sense in a world where memory storage and processing devices were separate.
With the physical separation of components, processors must retrieve data before performing computations. This involves moving electric charges, charging and discharging capacitors, and transiting currents, all leading to energy dissipation through heat.
Researchers at EPFL have been working on an in-memory processor that can perform the dual role of data storage and processing. Instead of silicon, we have used molybdenum disulfide (MoS2), another semiconductor.