Memristors can store data in the form of different resistances. In CPUs, these semiconductors can therefore simultaneously handle processor and memory functions – which neatly eliminates the traditional bottleneck in data exchange between these components. In addition, memristor-based CPUs are more energy-efficient than conventional ones. However, these semiconductors require a completely new computer architecture because they store information not as zeros and ones, but as different levels of resistance – in other words, in analog. Researchers hope that this technology will provide tremendous performance gains, especially in areas such as machine learning.
In recent years, renowned chip manufacturers have also tackled processors that specialize in the field of artificial intelligence. These chips, largely based on classical CPU technology, also promise performance gains, including in the field of image recognition and associated augmented reality applications. Whether and when the newly developed Memristor processors will be launched on the market is not yet foreseeable.