A method includes determining a quantity of refresh operations performed on a block of a memory device of a memory sub-system and determining a quantity of write operations and a quantity of read operations performed to the block. The method also includes determining the block is read dominant using the quantity of write operations and the quantity of read operations and determining whether the quantity of refresh operations has met a criteria. The method further includes, responsive to determining that the block is read dominant and that the quantity of refresh operations has met the criteria, modifying trim settings used to operate the block of the memory device.

A non-volatile memory device includes arrays of non-volatile memory cells that are configured to the store weights for a recurrent neural network (RNN) inference engine with a gated recurrent unit (GRU) cell. A set three non-volatile memory arrays, such as formed of storage class memory, store a corresponding three sets of weights and are used to perform compute-in-memory inferencing. The hidden state of a previous iteration and an external input are applied to the weights of the first and the of second of the arrays, with the output of the first array used to generate an input to the third array, which also receives the external input. The hidden state of the current generation is generated from the outputs of the second and third arrays.

A non-volatile memory device includes arrays of non-volatile memory cells that are configured to the store weights for a recurrent neural network (RNN) inference engine with a gated recurrent unit (GRU) cell. A set three non-volatile memory arrays, such as formed of storage class memory, store a corresponding three sets of weights and are used to perform compute-in-memory inferencing. The hidden state of a previous iteration and an external input are applied to the weights of the first and the of second of the arrays, with the output of the first array used to generate an input to the third array, which also receives the external input. The hidden state of the current generation is generated from the outputs of the second and third arrays.

A magnetic multilayer film for a magnetic memory element includes an amorphous heavy metal layer having a multilayer structure in which a plurality of first layers containing Hf alternate repeatedly with a plurality of second layers containing a heavy metal excluding Hf; and a recording layer that includes a ferromagnetic layer and that is adjacent to the heavy metal layer, the ferromagnetic layer having a variable magnetization direction.

Disclosed are systems and methods that determine whether instances of data (e.g., forward activations, backward derivatives of activations) that are used to train deep neural networks are to be stored on-chip or off-chip. The disclosed systems and methods are also used to prune the data (discard or delete selected instances of data). A system includes a hierarchical arrangement of on-chip and off-chip memories, and also includes a hierarchical arrangement of data selector devices that are used to decide whether to discard data and where in the system the data is to be discarded.

Aspects of the present disclosure are directed to devices and methods for performing MAC operations using a memory array as a compute-in-memory (CIM) device that can enable higher computational throughput, higher performance and lower energy consumption compared to computation using a processor outside of a memory array. In some embodiments, an activation architecture is provided using a bit cell array arranged in rows and columns to store charges that represent a weight value in a weight matrix. A read word line (RWL) may be repurposed to provide the input activation value to bit cells within a row of bit cells, while a read-bit line (RBL) is configured to receive multiplication products from bit cells arranged in a column. Some embodiments provide multiple sub-arrays or tiles of bit cell arrays.

A dual-floating gates optoelectronic self-exciting synaptic memristor includes a bottom gate, a barrier layer coated on a surface of the bottom gate, a quantum dot layer coated on a surface of a middle portion of the barrier layer, two inverted L-shaped electron or hole tunneling layers coated on a surface of two end portions of the quantum dot layer respectively, two inverted L-shaped floating gate storage layers coated on the electron or hole tunneling layers respectively, two electron or hole blocking layers coated on the two floating gate storage layers respectively, an inverted L-shaped source electrode and an inverted L-shaped drain electrode coated on the two electron or hole blocking layers respectively, a photosensitive material layer coated on a surface of a middle portion of the quantum dot layer, and a top gate coated on the photosensitive material layer.

In some embodiments, an integrated circuit (IC) device includes an active semiconductor layer, a circuitry formed within the active semiconductor layer, a region including conductive layers formed above the active semiconductor layer, and a memory module formed in the region. The memory device includes a three-dimensional array of memory cells, each adapted to store a weight value, and adapted to generate at each memory cell a signal indicative of a product between the stored weight value and an input signal applied to the memory cell. The memory module is further adapted to transmit the product signals from the memory cell simultaneously in the direction of the active semiconductor layer.

A processing apparatus includes a bit-cell array including at least one bit-cell line including a plurality of bit-cells electrically connected to each other in series, wherein each of the plurality of bit-cells includes: a first magnetic resistor that is configured to store a first resistance value based on a movement of a location of a magnetic domain-wall; a second magnetic resistor that is configured to store a second resistance value, wherein the second resistance value is equal to or less than the first resistance value; a first switching element configured to switch an electrical signal applied to the first magnetic resistor; and a second switching element configured to switch an electrical signal applied to the second magnetic resistor.

A processing apparatus includes a bit-cell array including at least one bit-cell line including a plurality of bit-cells electrically connected to each other in series, wherein each of the plurality of bit-cells includes: a first magnetic resistor that is configured to store a first resistance value based on a movement of a location of a magnetic domain-wall; a second magnetic resistor that is configured to store a second resistance value, wherein the second resistance value is equal to or less than the first resistance value; a first switching element configured to switch an electrical signal applied to the first magnetic resistor; and a second switching element configured to switch an electrical signal applied to the second magnetic resistor.