Researchers have developed a mathematical method that can reduce the energy cost of storing and processing data in computers by thousands of times, reports infohub.kz.
With the growing popularity of artificial intelligence and cloud services, the energy consumption of data centers is rapidly increasing. Without the introduction of fundamentally new energy-efficient technologies, information systems could become one of the world's largest consumers of electricity. Scientists at the University of Edinburgh have proposed a solution to this problem for magnetic memory.
The new concept is based on the application of optimal control theory. The researchers mathematically calculated the parameters of ultrafast magnetic field pulses that switch digital bits with minimal energy loss. The method takes into account real technical constraints of devices, making it applicable in practice.
Computer simulations showed that this approach reduces the energy cost of switching states by several orders of magnitude compared to DRAM, STT-MRAM, and SOT-MRAM technologies. This brings devices closer to the Landauer limit—the physical minimum of energy required to process one bit of information.
Lead researcher Elton Santos notes that the developed mathematical framework is universal. In the future, the concept could be adapted to work with electric currents and ultrashort laser pulses.