Systems, devices, and methods related to a Deep Learning Accelerator and memory are described. An integrated circuit may be configured to execute instructions with matrix operands and configured with: random access memory configured to store instructions executable by the Deep Learning Accelerator and store matrices of an Artificial Neural Network; a connection between the random access memory and the Deep Learning Accelerator; a first interface to a memory controller of a Central Processing Unit; and a second interface to a direct memory access controller. While the Deep Learning Accelerator is using the random access memory to process current input to the Artificial Neural Network in generating current output from the Artificial Neural Network, the direct memory access controller may concurrently load next input into the random access memory; and at the same time, the Central Processing Unit may concurrently retrieve prior output from the random access memory.

Apparatuses and methods related to memory bank power coordination in a memory device are disclosed. A method for memory bank power coordination may include concurrently performing a memory operation by a threshold number of memory regions, such as banks or subarrays, and executing a command to cause a budget area, such as a register, to perform a power budget operation associated with the memory operation. The threshold number of memory regions may be set based at least in part on a threshold power consumption value, and the number of memory regions to concurrently perform an operation may be controlled by a bank arbiter. A counter having a value representing the threshold number of memory regions may be decremented while performing an operation or incremented upon completion of an operation associated with one of the memory regions. A number of the memory regions may be selected to perform a processing-in-memory operation.

A method of controlling a memory device can include: receiving, by an interface, a write command from a host; beginning execution of a write operation on a first array plane of a memory array in response to the write command, where the memory array includes a plurality of memory cells arranged in a plurality of array planes; receiving, by the interface, a read command from the host; reconfiguring the write operation in response to detection of the read command during execution of the write operation; beginning execution of a read operation on a second array plane in response to the read command; and restoring the configuration of the write operation after the read operation has at least partially been executed.

A memory, a memory chip and a memory data access method are provided. The memory of the disclosure includes a plurality of memory chips. Each of the plurality of memory chips includes a first bank group, a second bank group and a read amplifier and a write amplifier. The first bank group includes a plurality of first memory banks. The second bank group includes a plurality of second memory banks. The read amplifier and the write amplifier are separately coupled to the first bank group and the second bank group. The read amplifier and the write amplifier are configured to independently access different bank groups.

Various implementations described herein are directed to a device having various circuitry for reading first data from a memory location in single-port memory and writing second data to the memory location in the single-port memory after reading the first data from the memory location. In some implementations, reading the first data and writing the second data to the memory location are performed in a single operation.

A non-volatile memory device includes a plurality of memory cells arranged in a matrix, a plurality of word lines extended in a row direction, and a plurality of bit lines extended in a column direction. Each of the memory cells is coupled to one of the word lines and one of the bit lines. The memory device further includes a word-line control circuit coupled to and configured to control the word lines, a first bit-line control circuit configured to control the bit lines and sense the memory cells in a digital mode, and a second bit-line control circuit configured to bias the bit lines and sense the memory cells in an analog mode. The first bit-line control circuit is coupled to a first end of each of the bit lines. The second bit-line control circuit is coupled to a second end of each of the bit lines.

Embodiments of operation methods of ferroelectric memory are disclosed. In an example, a method for reading ferroelectric memory cells is disclosed. The ferroelectric memory cells include a first set of ferroelectric memory cells and a second set of ferroelectric memory cells. In a first cycle, first data in a first ferroelectric memory cell of the first set of ferroelectric memory cells is sensed. In a second cycle subsequent to the first cycle, the sensed first data is written back to the first ferroelectric memory cell, and second data in a second ferroelectric memory cell of the second set of ferroelectric memory cells is simultaneously sensed.

A nonvolatile memory device includes a memory cell array and a row decoder. The memory cell array includes a plurality of mats. A first cell string of first mat is connected to a plurality of first word-lines, a first bit-line and a first string selection line. A second cell string of second mat is connected to a plurality of second word-lines, a second bit-line and a second string selection line. Each of the first and second cell strings includes a ground selection transistor, memory cells, and a string selection transistor coupled in series. The row decoder applies a first voltage to a third word-line among the plurality of first and second word-lines for a first period of time in a single mat mode and to apply a second voltage to the third word-line for a second period of time longer than the first period of time in a multi-mat mode.

A selection circuit includes: a first selection device coupled between a write IO line and a first node; a second selection device coupled between a read IO line and a second node; a third selection device controllable by a first address decode signal, and coupled between a first bit line and a third node; a fourth selection device controllable by a second address decode signal, and coupled between a second bit line and the third node; a first suppression device controllable by a write enable signal, and coupled between the second node and ground; a second suppression device controllable by a read enable signal, and coupled between the first node and ground; a first isolation device controllable by the write enable signal, and coupled between the first and third nodes; and a second isolation device controllable by the read enable signal, and coupled between the second and third nodes.

The present disclosure provides techniques for using a multiple-port buffer to improve a transaction rate of a memory module. In an example, a memory module can include a circuit board having an external interface, first memory devices mounted to the circuit board, and a first multiple-port buffer circuit mounted to the circuit board. The first multiple-port buffer circuit can include a first port coupled to data lines of the external interface, the first port configured to operate at a first transaction rate, a second port coupled to data lines of a first plurality of the first memory devices, and a third port coupled to data lines of a second plurality of the first memory devices. The second and third ports can be configured to operate at a second transaction rate, wherein the second transaction rate is slower than the first transaction rate.