April was an excellent month for new memory technologies. First we get IBM saying its Racetrack technology can increase memory density by ten times; then we get Glasgow University saying they can make a molecular switch which could implement a Petabyte memory on a square inch substrate; then Manchester University says it’s made a transistor one atom thick and ten atoms wide out of graphene; Duke University, using a masking technique, makes the highest density cluster of carbon nanotubes ever achieved - ten nanotubes a few atoms thick, and then H-P comes up with its ‘memristor’, a resistor with storage capabilities which could be used in FPGAs and for ultra-low power memory.
HP’s memristor is made on a dot of titanium dioxide 15nm in diameter. It is a working device on which patent claims have been filed.
HP scientists say they could see their way to a 4nm memristor, that they are putting memristor development on the ‘fast-track’, that they can be made in conventional semiconductor fabs, and that commercial exploitation could be achieved ‘fairly quickly’.
Memristors can store not only binary digits but intermediate numbers allowing them to act more like the synapses of the brain than the digital storage of today’s memory chips.
This gives memristors more potential than conventional memory in applications like machine vision and speech recognition.
Memristors were named and predicted back in 1971 by Leon Chua, an electrical engineer at the University of California at Berkeley.
Subscribe to this blog's feed