Su Shi, a prototype chip of superconducting neuromorphological processor in China, released
CTOnews.com October 11, according to the official news of the Voice of the Chinese Academy of Sciences, the Institute of Computing Technology of the Chinese Academy of Sciences recently released the superconducting neuromorphological processor prototype chip Su Shi.
According to reports, "Su Shi" is a superconducting computing chip based on superconducting single flux quantum (CTOnews.com Note: SFQ) circuit. Superconducting SFQ circuit has the characteristics of ultra-high computing speed and ultra-low computing power consumption, which is expected to break through the bottleneck of traditional computing which is difficult to improve computing power under the conditions of unit volume and unit energy consumption.
In recent years, superconducting computing chip has become one of the research hotspots in the field of new principle computing. The team proposed the design method of neuron, weight, and on-chip network based on superconducting SFQ circuit, and developed the prototype chip of superconducting neural morphology processor.
The ▲ "Su Shi" neural morphology processing flow is based on the official website of the Chinese Academy of Sciences. Su Shi is fabricated using the 2 micron SIMIT-Nb03 superconducting integrated circuit process independently developed by the Shanghai Institute of Microsystems and Information Technology of the Chinese Academy of Sciences. The test verifies and obtains the correct reasoning results of the complete neural morphological computing network, which is the first time in the world to use superconducting computing chips to achieve this function.
The real "Su Shi" chip of ▲ "Su Shi" has precision variability and easy scalability. Using 100, 000 Josephson knots, the peak neuromorphological processing performance of nearly 1.4 trillion synaptic operations per second and the high energy efficiency of more than 32 trillion synaptic operations per watt can be obtained.
The related research results, entitled SUSHI: Ultra-High-Speed and Ultra-Low-Power Neuromorphic Chip Using Superconducting Single-Flux-Quantum Circuits, were selected into the top international conference on computer architecture MICRO 2023 (56th IEEE / ACM International Symposium on Microarchitecture).
