This system can sort data for AI faster, more efficiently
What's the story
Chinese scientists have developed a data sorting system that is both faster and more energy-efficient than traditional methods. The new system uses memristors, electronic circuit components with memory-like properties, and a special sorting algorithm to enable more efficient data processing. The innovation might help overcome limitations in scientific computing, artificial intelligence (AI), and hardware design.
Technological advancement
New system uses memristors and special sorting algorithm
The team has created a memristor-based hardware sorting prototype to demonstrate tasks like route finding and neural network inference. This prototype showed both speed and energy efficiency improvements over conventional sorting methods, proving the potential of this new technology in various applications.
Performance limitation
Sorting is a major performance bottleneck
The team of scientists from Peking University and the Chinese Institute for Brain Research highlighted that sorting is a major performance bottleneck in many applications, including AI, databases, web search, and scientific computing. They also noted that current systems still rely on comparison operations which limits sorting performance. Their new memristor-based system could help overcome these limitations by enabling more efficient data processing.
Architectural constraints
Many computing systems are based on Von Neumann architecture
Most computing systems are based on Von Neumann architecture, that separates data storage and processing. This has led to the Von Neumann bottleneck, which limits the speed of data transfer between the main memory and processing unit. The researchers hope that the sort-in-memory using memristors could help overcome these limitations.