Chinese researchers have crafted an approach that makes data sorting faster and more energy‐efficient. By harnessing memristors—devices that mimic how our brains store information—and pairing them with a fresh sorting algorithm, they’ve tackled long‐standing bottlenecks in both computing and artificial intelligence.
The collaborative effort between scientists at Peking University and the Chinese Institute for Brain Research produced a hardware prototype that excels in tasks such as route optimisation and neural network inference. In tests reported by the South China Morning Post, the prototype boosted throughput by nearly 8 times while slashing energy use by more than 160 times and improving area efficiency over 32 times compared to traditional systems.
Published in Nature Electronics, the study explains that sorting remains a significant hurdle for applications in AI, database management, web search, and scientific computing. Traditional systems based on the Von Neumann architecture separate memory from processing, creating speed limits as data has to move back and forth. By contrast, a memristor‐based sort-in-memory approach could reduce or eliminate this gap.
Unlike ordinary resistors, memristors retain a memory of the electrical charges that pass through them, tweaking their resistance in the process. This dual function—as both storage and processor—means that data doesn’t need to shuttle between separate units. Instead, the system performs an iterative search to find minimum or maximum values without relying on direct comparisons, which saves both time and energy.
If you’ve ever found yourself frustrated by sluggish data processing, this method offers a welcome alternative. Beyond AI, its potential applications range from real-time smart traffic systems to quicker financial risk assessments, marking a promising step toward more efficient, integrated computing systems.
