Various embodiments provide methods for configuring a phase-change random-access memory (PCRAM) structures, such as PCRAM operating in a single-level-cell (SLC) mode or a multi-level-cell (MLC) mode. Various embodiments may support a PCRAM structure being operating in a SLC mode for lower power and a MLC mode for lower variability. Various embodiments may support a PCRAM structure being operating in a SLC mode or a MLC mode based at least in part on an error tolerance for a neural network layer.

A multi-level cell (MLC) one-selector-one-resistor (1S1R) three-dimensional (3D) cross-point memory system includes at least one MLC 1S1R structure including a stacked arrangement of a phase change memory (PCM) cell and a threshold switch selector. An electrically conductive bit line is in electrical communication with the OTS selector, and an electrically conductive word line is in electrical communication with the PCM cell. A controller is in electrical communication with the bit line and the word line. The controller is configured to select at least one voltage pulse from a group of different voltage pulses comprising a read pulse, a partial set pulse, a set pulse, a partial reset pulse, and a reset pulse, and configured to deliver the selected at least one voltage pulse to the at least one MLC 1S1R structure.

Methods, systems, and devices for data-based polarity write operations are described. A write command may cause a set of data to be written to a set of memory cells. To write the set of data, a write operation that applies voltages across the memory cells based on a logic state of data to be written to the memory cells may be used. During a first interval of the write operation, a voltage may be applied across a memory cell based on a logic state of a data bit to be written to the memory cell. During a second interval of the write operation, a voltage may be applied across the memory cell based on an amount of charge conducted by the memory cell during the first interval.

A device may include a first die having a first circuit and a second die having a second circuit. The die may be separated by a material layer. The material layer may include multiple through-silicon vias (TSVs) for electrically coupling the first die to the second die. A first TSV of the TSVs may electrically couple the first circuit to the second circuit and a second TSV of the TSVs may include a redundant TSV that electrically bypasses the first TSV to couple the first circuit to the second circuit if a fault is detected in the first TSV.

The present disclosure is directed to a method for storing information in a coded manner in non-volatile memory cells. The method includes providing a group of non-volatile memory cells of non volatile memory. The memory cell is of the type in which a stored logic state, which can be logic high or logic low, can be changed through application of a current to the cell and the state in the memory cell is read by reading a current provided by the cell. The group of non-volatile memory cells include a determined number of non-volatile memory cells which is greater than two. The group of non-volatile memory cells store a codeword formed by the values of said stored states of the cells of the group taken according to a given order. Given a set of codewords obtainable by the stored values in the determined number of non-volatile memory cells in a group, the method includes storing the information in at least two subsets of said set of codewords comprising each at least a codeword. Each codeword in a same subset has a same Hamming weight. Each codeword belonging to one subset has a Hamming distance equal or greater than two with respect to each codeword belonging to another subset.

Embodiments of three-dimensional (3D) memory devices with embedded dynamic random-access memory (DRAM) and methods for forming the 3D memory devices are disclosed. In an example, a method for operating a 3D memory device is disclosed. The 3D memory device includes an input/output circuit, an array of embedded DRAM cells, and an array of 3D NAND memory strings in a same chip. Data is transferred through the input/output circuit to the array of embedded DRAM cells. The data is buffered in the array of embedded DRAM cells. The data is stored in the array of 3D NAND memory strings from the array of embedded DRAM cells.

Three-dimensional (3D) memory devices with 3D phase-change memory (PCM) and methods for forming and operating the 3D memory devices are disclosed. In an example, a 3D memory device includes a first semiconductor structure including an array of NAND memory cells, and a first bonding layer including first bonding contacts. The 3D memory device also further includes a second semiconductor structure including a second bonding layer including second bonding contacts, a semiconductor layer and a peripheral circuit and an array of PCM cells between the second bonding layer and the semiconductor layer. The 3D memory device further includes a bonding interface between the first and second bonding layers. The first bonding contacts are in contact with the second bonding contacts at the bonding interface.

There is described a two-terminal multi-level memristor element synthesised from binary memristors, which is configured to implement a variable resistance based on unary or binary code words. There is further described a circuit such as a synapse circuit implemented using a multi-level memristor element.

Systems, methods and apparatus to program a memory cell to have a threshold voltage to a level representative of one value among more than two predetermined values. A first voltage pulse is driven across the memory cell to cause a predetermined current to go through the memory cell. The first voltage pulse is sufficient to program the memory cell to a level representative of a first value. To program the memory cell to a level representative of a second value, a second voltage pulse, different from the first voltage pulse, is driven across the memory cell within a time period of residual poling in the memory cell caused by the first voltage pulse.

An output, representing synaptic weights of a neural network can be received from first memory elements. The output can be compared to a known correct output. A random number can be generated with a tuned bias via second memory elements. The weights can be updated based on the random number and a difference between the output and the known correct output.