
Weebit Nano seeks ‘selector’ for discrete memory market entry
Weebit Nano Ltd. (Hod Hasharon, Israel) is partnering with French research institute Leti – as it has done throughout its development phase – in a fifteen-month program, that should end with a demonstration of Weebit’s ReRAM cell working with the selector. Leti has been developing a selector for the discrete memory market for the past several years.
An initial three-month stage will be used to define the specific details of the selector.
A selector serves a key role in two-terminal memories in that it selects the memory cell being addressed and prevents sneak paths through the array bypassing the selected cell.
A number of technologies and voltage schemes already exist for this but with different complexities in production and energy efficiencies in operation. For example Sony is reportedly preparing to introduce a filamentary ReRAM with an ‘ovonic’ chalcogenide-based selector (see Will Sony launch cross-point nonvolatile memory?).
That would place productization of discrete SiOx memories in the 2021-2022 time frame.
Meanwhile Weebit claims it is on course to receive first orders for embedded non-volatile memory in late 2020 for applications such as AI, IoT and analog.
“Our recent work with XTX Technology and other potential partners accelerated our entry into the discrete memory market, and we are now scaling up our efforts in this segment in parallel to the continued work in the embedded space,” said Coby Hanoch, CEO of Weebit, in a statement. He added: “Our goal is to demonstrate the ReRAM cell intended for the discrete memory market by mid-2021.”
Weebit will also present its SPIRIT neuromorphic demo at the International Solid-State Circuit Conference (ISSCC) 2020 in San Francisco from 16 to 20 February. The demo uses Weebit’s silicon-oxide ReRAMs to perform inference tasks using CEA-Leti’s spiking neural network (SNN) algorithms. The memory cell implements synapses in a way that mimics human biological synapse activity and the memory array greatly increases the inherent parallel connectivity compared to more coarse-grained processor-plus-memory circuits. This in turn improves the energy efficiency of artificial neural networks and makes processing-in-memory an attractive approach.
Related links and articles:
News articles:
Will Sony launch cross-point nonvolatile memory?
China’s XTX adopts Weebit’s ReRAM
Weebit, Leti to demo SiOx ReRAM in neuromorphic application
Weebit pivots SiOx memory towards embedded – and revenue
