Spintec unveils high-potential memristor for neuromorphic computing

Spintec recently completed a proof-of-concept of a promising memristor that could be used as an artificial synapse in neuromorphic architectures.
This non-volatile memory’s resistance can have multiple intermediate values between a minimum and a maximum.
Thanks to recent advances, the storage layer’s magnetization can be oriented and stabilized in all directions of the layer plane, rather than just parallel or antiparallel to a reference layer.
The memristor stands out for its low power consumption, low variability, and write endurance, making it ideal for machine learning applications with their billions of potentially energy-intensive operations.
The next step is to implement the component in circuit architectures for testing.

Contact: bernard.dieny@cea.fr

Facebook

Twitter