LETI Opens Neuromorphic Computer Laboratory: 1000x Higher Energy Efficiency

The laboratory "Neuromorphic Electronics and Memory Computing" has started operating at the St. Petersburg State Electrotechnical University "LETI". The main goal of the division is to create energy-efficient computing systems that operate on the principles of biological neurons. According to the head of the laboratory, Natalya Andreeva, research by 2025 has made it possible to combine the efforts of specialists from materials science, IT, circuit design and ASIC design, which became the basis for the formation of a new scientific direction.

Neuromorphic computers differ from classic counterparts in their ability to perform parallel operations with minimal energy consumption — 1000 times less than that of graphics accelerators. Such systems can learn, adapt to tasks, and be used in machine learning, including artificial intelligence technologies. The laboratory will focus on the development of integrated blocks for autonomous neural networks with low energy consumption, as well as hardware and software for neuromorphic devices.

An important aspect will be the educational program for students in the field of "Electronics and Nanoelectronics". Students will have access to the study of modern components — from transistors to ultra-high-speed structures. This will ensure that graduates are competitive in the labor market, including cooperation with laboratory partners: the Intellectual Electronics — Valdai Innovation Center and the I.M. Sechenov Institute of the Russian Academy of Sciences.

LETI's developments put Russia in the ranks of countries exploring neuromorphic technologies. Unlike traditional GPUs, such systems are capable of processing data without constantly transferring information between the processor and memory, which speeds up calculations. Similar solutions are already being tested abroad: for example, Intel Loihi chips. However, the Russian approach focuses on the integration of a memristor base, which may become the key to creating compact and energy-efficient devices for robotics, IoT and neurointerfaces.

Read more materials on the topic:

TPU Proposes Using Scaffolds to Prevent Adhesions After Surgery

Technology for Producing Useful Nanomaterials from Toxic Waste Created in Novosibirsk

Scientists Develop Algorithm for Rapid Detection of Cardiac Fibrosis