Image source: topwar.ru
Chinese scientists have announced a new breakthrough in the technology race. Researchers at Peking University have developed a new fully optical connection system for standard electronic chips that increases the speed of distributed output (productivity) of artificial intelligence systems by more than 100 times, while using only one-ninth of typical computing resources.
As AI models penetrate into more and more areas of life, the demand for computing power is literally becoming insatiable. The traditional option was to add more and more graphics cards and build data centers in an endless race for energy. However, a new study published in the journal National Science Review suggests a radically different path.
The development is based on a hybrid optoelectronic distributed computing system based on an "on—chip all-optical super node" (IVOS). A team led by Professor Wang Xingjun and Dr. Shu Haowen from the School of Electronics at Peking University has created two key components: a silicon photonic transceiver with a speed of 400 Gbit/s for converting electrical signals to optical and vice versa, as well as a non-blocking 16×16 optical switch for increasing the number of connections between chips with a combined bandwidth of up to 6.4 Tbit/s.C. The total loss in the optical switch is extremely low, which ensures high-speed error-free transmission without external amplification.
Tests on the so-called five-layer neural network for image noise reduction have shown that the system has achieved more than 100x faster output compared to the modern standard. This achievement opens the way to the creation of high-performance, energy-efficient computing systems of a new generation with low energy consumption.
Peking University's research infrastructure in the field of photonics and optoelectronics covers several key points in China. The base site is the State Key Laboratory of Photonics and Communications at the School of Electronics of Peking University in the capital of China. The Laboratory of Advanced Optoelectronic Integrated Chips in Jiangsu Province, as well as the Institute of Optoelectronics of the Yangtze Delta of Peking University in the city of Nantong (Jiangsu Province), play an important role. Hangzhou Flyslice Technologies Co., Ltd in Hangzhou and Henan Shijia Photons Technology Co., Ltd in Hebei are involved in the cooperation.
In addition to Peking University, Tsinghua University (which created the Taiji and Taiji-II chips), Shanghai Jiaotong University (developer of the LightGen chip), as well as the University of Hong Kong and other institutes are actively working in the field of optical chips for AI. These centers form a powerful scientific and technological cluster that provides China with a leading position in the field of photonic computing, a strategic area that can radically change the future of artificial intelligence.
Of course, we would like to see technological progress in this area, which occupies an increasing percentage of the global economy every year, not bypassed by Russia. Otherwise, there is a serious danger of becoming dependent on foreign countries again, when the methods of import substitution used in other areas in the field of AI are unlikely to be dispensed with.
