A semiconductor memory includes a substrate including a cell region, a first peripheral circuit region, and a second peripheral circuit region; a plurality of first lines disposed over the substrate across the cell region and the first peripheral circuit region; a plurality of second lines disposed over the first lines across the cell region and the second peripheral circuit region; and a first memory cell positioned at each of intersections between the first lines and the second lines, wherein the cell region includes a first cell region and a second cell region, the first cell region being disposed closer to the first and second peripheral circuit regions than the second cell region, and wherein a first portion of the second line that is in the first cell region has a greater resistance than a second portion of the second line that is in the second cell region.

A variable resistance memory device includes lower conductive lines on a substrate, upper conductive lines on the lower conductive lines to cross the lower conductive lines, and memory cells between the lower conductive lines and the upper conductive lines. The lower conductive lines are extended in a first direction and are spaced apart from each other in a second direction crossing the first direction. Each of the lower conductive lines include a first line portion extended in the first direction, a second line portion offset from the first line portion in the second direction and extended in the first direction, and a connecting portion connecting the first line portion to the second line portion.

A semiconductor device includes a first insulating layer; a second insulating layer arranged over the first insulating layer; a memory structure arranged within a memory region and including a resistance changing memory element within the first insulating layer; and a logic structure arranged within a logic region. In the memory region, the first insulating layer may contact the second insulating layer and in the logic region, the semiconductor device may further include a stop layer arranged between the first insulating layer and the second insulating layer.

It is described a mathematical solving circuit (100) comprising: a crosspoint matrix (M.sub.G) including a plurality of row conductors (L.sub.i), a plurality of column conductors (C.sub.j) and a plurality of analog resistive memories (Gij), each connected between a row conductor and a column conductor; a plurality of operational amplifiers (OA.sub.i) each having: a first input terminal (IN.sub.1i) connected to a respective row conductor (L.sub.i), a second input terminal (IN.sub.2i) connected to a ground terminal (GR) at least one operational amplifier (OA.sub.i) of the plurality being such to take the respective first input terminal (IN.sub.1i) to a virtual ground.

A method includes performing a quantity of write cycles on memory components. The method can further include monitoring codewords, and, for each of the codewords including a first error parameter value, determining a second error parameter value. The method can further include determining a probability that each of the codewords is associated with a particular one of the second error parameter values at the first error parameter value and determining a quantity of each of the codewords that are associated with each of the determined probabilities. The method can further include determining a statistical boundary of the quantity of each of the codewords and determining a correlation between the quantity of write cycles performed and the corresponding determined statistical boundary of the quantity of each of the codewords.

The present disclosure includes apparatuses and methods for buffer reset commands for write buffers. An example apparatus includes a memory and a controller coupled to the memory. The memory can include an array of resistance variable memory cells configured to store data corresponding to a managed unit across multiple partitions each having a respective write buffer corresponding thereto. The controller can be configured to update the managed unit by providing, to the memory, a write buffer reset command followed by a write command. The memory can be configured to execute the write buffer reset command to place the write buffers in a reset state. The memory can be further configured to execute the write command to modify the content of the write buffers based on data corresponding to the write command and write the modified content of the write buffers to an updated location in the array.

An adaptive memory management and control circuitry (AMMC) to provide extended test, performance, and power optimizing capabilities for a resistive memory is disclosed herein. In one embodiment, a resistive memory comprises a resistive memory array and an Adaptive Memory Management and Control circuitry (AMMC) that is coupled to the resistive memory array. The AMMC is configured with extended test, reliability, performance and power optimizing capabilities for the resistive memory.

Systems, methods and devices are disclosed for a smart compute memory circuitry that has the flexibility to perform a wide range of functions inside the memory via logic circuitry and an integrated processor. In one embodiment, the smart compute memory circuitry comprises an integrated processor and logic circuitry to enable adaptive System on a Chip (SOC) and electronics subsystem power or performance improvements, and adaptive memory management and control for the smart compute memory circuitry. A resistive memory array is coupled to the integrated processor.

A Bayesian neural network including an input layer, and, an output layer, and, possibly, one or more hidden layer(s). Each neuron of a layer is connected at its input with a plurality of synapses, the synapses of the plurality being implemented as a RRAM array constituted of cells, each column of the array being associated with a synapse and each row of the array being associated with an instance of the set of synaptic coefficients, the cells of a row of the RRAM being programmed during a SET operation with respective programming current intensities, the programming intensity of a cell being derived from the median value of a Gaussian component obtained by GMM decomposition into Gaussian components of the marginal posterior probability of the corresponding synaptic coefficient, once the BNN model has been trained on a training dataset.

Provided are a weight matrix circuit and a weight matrix input circuit. The weight matrix circuit includes a memory array including n input lines, m output lines, and n×m resistive memory devices each connected to the n input lines and the m output lines and each having a non-linear current-voltage characteristic, an input circuit connected to each of the input lines, and an output circuit connected to each of the output lines. The input circuit is connected to the resistive memory devices such that the weight matrix circuit has a linear current-voltage characteristic.