Globalfoundries U.S. Inc.

Arm Limited

Globalfoundries Inc.

Purdue Research Foundation

Akhilesh Ramlaut Jaiswal

Ajey Poovannummoottil Jacob

Mudit Bhargava

Bipul C Paul

Kaushik Roy

Steven R Soss

Yusheng Bian

Amogh Agrawal

William Taylor

Danny Pak-Chum Shum

George McNeil Lattimore

David C Pritchard

Michal Rakowski

Indranil Chakraborty

Nishtha Gaul

Mustafa Ali

An in-memory vector addition method for a dynamic random access memory (DRAM) is disclosed which includes consecutively transposing two numbers across a plurality of rows of the DRAM, each number transposed across a fixed number of rows associated with a corresponding number of bits, assigning a scratch-pad including two consecutive bits for each bit of each number being added, two consecutive bits for carry-in (C), and two consecutive bits for carry-out-bar (), assigning a plurality of bits in a transposed orientation to hold results as a sum of the two numbers, for each bit position of the two numbers: computing the associated sum of the bit position; and placing the computed sum in the associated bit of the sum.

In-memory bit-serial addition system

Disclosed are methods, systems and devices for operation of memory device. In one aspect, volatile memory bitcells and non-volatile memory bitcells may be integrated to facilitate transfer of stored values between the volatile and non-volatile memory bitcells.

Method, system and device for integration of volatile and non-volatile memory bitcells

Disclosed is a memory structure with reference-free single-ended sensing. The structure includes an array of non-volatile memory (NVM) cells (e.g., resistance programmable NVM cells) and a sense circuit connected to the array via a data line and a column decoder. The sense circuit includes field effect transistors (FETs) connected in parallel between an output node and a switch and inverters connected between the data line and the gates of the FETs, respectively. To determine the logic value of a stored bit, the inverters are used to detect whether or not a voltage drop occurs on the data line within a predetermined period of time. Using redundant inverters to control redundant FETs connected to the output node increases the likelihood that the occurrence of the voltage drop will be detected and captured at the output node, even in the presence of process and/or thermal variations. Also disclosed is a sensing method.

Memory with high-accuracy reference-free multi-inverter sense circuit and associated sensing method

The present disclosure relates to semiconductor structures and, more particularly, to optical neuro-mimetic devices and methods of manufacture. The structure includes: a plurality of photodetectors and electrical circuitry that converts photocurrent generated from the photodetectors into electrical current and then sums up the electrical current to mimic neural functionality.

Optical neuro-mimetic devices

The present disclosure relates to a structure including a latch circuit, a first non-volatile field effect transistor (FET) connecting to a first side of the latch circuit and a bit line, and a second non-volatile field effect transistor (FET) connecting to a second side of the latch circuit and a complementary bit line.

Non-volatile transistor embedded static random access memory (SRAM) cell

Disclosed are embodiments of an integrated circuit structure (e.g., a processing chip), which includes an array of integrated pixel and memory cells configured for deep in-sensor, in-memory computing (e.g., of neural networks). Each cell incorporates a memory structure (e.g., DRAM structure or a ROM structure) with a storage node, which stores a first data value (e.g., a binary weight value), and a sensor connected to a sense node, which outputs a second data value (e.g., an analog input value). Each cell is selectively operable in a functional computing mode during which the voltage level on a bit line is adjusted as a function of both the first data value and the second data value. Each cell is further selectively operable in a storage node read mode. Furthermore, depending upon the type of memory structure (e.g., a DRAM structure), each cell is selectively operable in a storage node write mode.

Array of integrated pixel and memory cells for deep in-sensor, in-memory computing

An image sensor includes an array of optically switchable magnetic tunnel junctions (MTJs) arranged in columns and rows. The image sensor has first lines of transparent conductive material and second lines of conductive material. Each first line is in contact with the free layers of the MTJs in a corresponding row. Each second line is electrically connected to the fixed layers MTJs in a corresponding column. The first lines are concurrently exposable to radiation. The first and second lines are selectively biasable. In a global reset operation, biasing conditions are such that all MTJs are switched to an anti-parallel state. In a global sense operation, biasing conditions are such that, depending upon the intensity of radiation received at those portions of the first lines in contact with MTJs, the MTJs may switch to a parallel state. In selective read operations, biasing conditions are such that stored data values in the MTJs can be read.

Image sensor incorporating an array of optically switchable magnetic tunnel junctions

Disclosed is a cell that integrates a pixel and a two-terminal non-volatile memory device. The cell can be selectively operated in write, read and functional computing modes. In the write mode, a first data value is stored the memory device. In the read mode, it is read from the memory device. In the functional computing mode, the pixel captures a second data value and a sensed change in an electrical parameter (e.g., voltage or current) on a bitline connected to the cell is a function of both the first and second data value. Also disclosed is an IC structure that includes an array of the cells and, when multiple cells in a given column are concurrently operated in the functional computing mode, the sensed total change in the electrical parameter on the bitline for the column is indicative of a result of a dot product computation.

Integrated pixel and two-terminal non-volatile memory cell and an array of cells for deep in-sensor, in-memory computing

The disclosure provides a circuit structure and method for memory storage using a memory cell and magnetic random access memory (MRAM) stack. A circuit structure includes a memory cell having a first latch configured to store a digital bit, a first diode coupled to the first latch, and a first magnetic random access memory (MRAM) stack coupled to the first latch of the memory cell through the first diode. The first MRAM stack includes a first layer and a second layer each having a respective magnetic moment. The magnetic moment of the second layer is adjustable between a parallel orientation and an antiparallel orientation with respect to the magnetic moment of the first layer. Further, the magnetic anisotropy of the second layer can be modified through application of an applied voltage (VCMA effect). A spin Hall electrode is directly coupled to the first MRAM stack.

Circuit structure and method for memory storage with memory cell and MRAM stack

Disclosed is a reference circuit having an even number m of groups of m parallel-connected magnetic tunnel junctions (MTJs). The MTJs in half of the groups are programmed to have parallel resistances (RP) and the MTJs in the other half are programmed to have anti-parallel resistances (RAP). Switches connect the groups in series, creating a series-parallel resistor network. The total resistance (R) of the network has low variability and is essentially equal to half the sum of a nominal RP plus a nominal RAP and can be employed as a reference resistance (RREF). Under specific biasing conditions the series-parallel resistor network can generate a low variability reference parameter (XREF) that is dependent on this RREF. Also disclosed are an integrated circuit (IC) that includes the reference circuit and a magnetic random access memory (MRAM) structure, which uses XREF to determine stored data values in MRAM cells and associated methods.

Low variability reference parameter generation for magnetic random access memory

Disclosed is a cell that integrates a pixel and a three-terminal non-volatile memory device. The cell can be selectively operated in write, read and functional computing modes. In the write mode, a first data value is stored in the memory device. In the read mode, it is read from the memory device. In the functional computing mode, the pixel captures a second data value and a sensed change in an electrical parameter (e.g., voltage or current) on a bitline connected to the cell is a function of both the first and second data value. Also disclosed is an IC structure that includes an array of the cells and, when multiple cells in a given column are concurrently operated in the functional computing mode, the sensed total change in the electrical parameter on the bitline for the column is indicative of a result of a dot product computation.

Integrated pixel and three-terminal non-volatile memory cell and an array of cells for deep in-sensor, in-memory computing

Disclosed are techniques for forming and operating magnetic memory devices.

Method, system and device for magnetic memory

One illustrative MRAM device disclosed herein includes a first bit cell and a second bit cell. The first bit cell comprises a first access transistor and a first MTJ stack. The first MTJ stack comprises a first pinned layer and a first free layer, wherein the first pinned layer is connected to the first access transistor. The second bit cell comprises a second access transistor and a second MTJ stack. The second MTJ stack comprises a second pinned layer and a second free layer, wherein the second free layer is connected to the second access transistor.

Growing community of inventors

West Lafayette, IN, United States of America

Akhilesh Ramlaut Jaiswal