Other

Green Mountain Semiconductor Inc.

Cypress Semiconductor Corporation

Being Advanced Memory Corporation

International Business Machines Corporation

Ryan A Jurasek

Aaron Willey

Wolfgang Hokenmaier

Donald W Labrecque

Van Butler

Yuanxing Li

Adam B Wilson

Jacob Bucci

Kotb Jabeur

Aaron D Willey

A novel architecture provides many of the advantages of the array and datapath architecture of DRAM products that do not utilize ECC (error correction code) functionality, while simultaneously allowing the flexible deployment of ECC error correction as needed. Aspects of the disclosure enable the minimization of write and read latency typically introduced by the implementation of ECC error correction. Sharing of circuit components between neighboring memory regions is also introduced, which allows for a reduction in circuit area as well as a reduction in loading on speed-critical data bus wiring, which improves overall performance. A very fast single error correct (SEC) and double error detect (DED) read-out for real-time system-level awareness is also provided.

Flexible, low-latency error correction architecture for semiconductor memory products

An MRAM bitline write control circuit including an MRAM array of a plurality of MTJ cells. Each MTJ cell is connected to a bitline between a bitline transfer gate and a transfer device. Each transfer device is connected to a sourceline and a sourceline transfer gate. A master bitline is connected to each bitline transfer gate. A first bitline control transistor is connected to VDD and to a source follower transistor that is connected to the master bitline and a gate connected to a write 0 bias voltage. A second bitline control transistor is connected to VSS and to the master bitline. A selected MTJ cell is biased to write a 0 when the transfer device, the bitline transfer gate and the source line transfer gate, associated with the selected MTJ cell, are enabled and the first bitline control transistor is enabled to connect the source follower transistor to VDD.

MRAM bit line write control with source follower

Techniques are disclosed for artificial neural network functionality within dynamic random-access memory. A plurality of dynamic random-access cells is accessed within a memory block. Data within the plurality of dynamic random-access cells is sensed using a plurality of sense amplifiers associated with the plurality of dynamic random-access cells. A plurality of select lines coupled to the plurality of sense amplifiers is activated to facilitate the sensing of the data within the plurality of dynamic random-access cells, wherein the activating is a function of inputs to a layer within a neural network, and wherein a bit within the plurality of dynamic random-access cells is sensed by a first sense amplifier and a second sense amplifier within the plurality of sense amplifiers. Resulting data is provided based on the activating wherein the resulting data is a function of weights within the neural network.

Artificial neural network functionality using dynamic random-access memory

Artificial neural network functionality within dynamic random-access memory

A highly configurable, extremely dense, high speed and low power artificial neural network is presented. The architecture may utilize DRAM cells for their density and high endurance to store weight and bias values. A number of primary sense amplifiers along with column select lines (CSLs), local data lines (LDLs), and sense circuitry may comprise a single neuron. Since the data in the primary sense amplifiers can be updated with a new row activation, the same hardware can be reused for many different neurons. The result is a large amount of neurons that can be connected by the user. Training can be done in hardware by actively varying weights and monitoring cost. The network can be run and trained at high speed since processing and/or data transfer that needs to be performed can be minimized.

Dynamic random-access memory with embedded artificial neural network functionality

An extremely dense, high speed, and low power content addressable DRAM is presented. To enable a parallel searching, a data word to be searched may be driven onto column select lines (CSLs) of a DRAM array. Although two or more primary sense amplifiers typically are not connected at the same time to the same local data line during operation of a DRAM, in various embodiments presented herein, some or all sense amplifiers in a DRAM can be activated simultaneously to enable maximum parallelism with local data line sharing being explicitly allowed. Using this architecture, a data word can be simultaneously searched in all banks and with multiple wordlines. Since no input/output transactions are required and no data needs to be driven from the bank during execution of a search, overall current, and thus power usage, can be reduced.

Content addressable dynamic random-access memory with parallel search functionality

Methods, circuits, and systems for phase change memories. A matching bit line, on which no data-containing PCM cells have been selected, is used to cancel out time-dependent current components due to parasitic capacitive and leakage resistance loading of bit lines. This can effectively allow direct comparison of the current from the phase change memory cell to the desired reference current, at a time before the voltage of the first bit line permits stable operations using DC comparison.

Phase change memory with bit line matching

Methods and systems for time-based cell decoding for PCM memory. Generally, the higher the PCM element resistance, the longer it takes for a read output to change state. PCM memory output is determined using differentiated timings of read outputs changing state, rather than differentiated values of read outputs. In some single-bit single-ended sensing embodiments, a reference, with resistance between the resistances corresponding to a pair of adjacent logical states, is stored in multiple reference cells; a 'vote' unit emits a clock signal when a majority of the reference cell read outputs transition at the vote unit. Timing units produce different binary outputs depending on whether a data read output or the clock signal changes state first at the timing unit. Time-based decoding provides advantages including improved temperature and drift resilience, improved state discrimination, improved reliability of multibit PCM, and fast and reliable sensing.

Phase change memory with flexible time-based cell decoding

Methods and systems for phase change memories and arrays with improved write characteristics. If a data word can be more efficiently written by e.g. exchanging SETs and RESETs, it is written as such on the fly, and e.g. a bit of overhead is written to indicate the transformation. This has a surprising synergy with phase change memory as SET operations usually take longer and consume more power than do RESET operations. In one sample embodiment of multilevel phase change memory, states intermediate between SET and RESET can be even less desirable to write than SETs, as they take more precision than do the extreme states of SET and RESET, so that a desirable transformation can be to exchange intermediate states for extreme states.

Systems, methods, and devices with write optimization in phase change memory

Methods and systems for multi-bit phase change memories. Using differential sensing for memory reads provides advantages including improved temperature and drift resilience, improved state discrimination and increased storage density.

Multilevel differential sensing in phase change memory

The present application discloses phase-change memory architectures and methods, in which an additional test is performed, after the normal power-valid signal, to assure that the phase-change memory components which are used for storing configuration data are able to operate correctly. Surprisingly, the inventor has discovered that this additional test is highly desirable when using phase-change memory for configuration data.

Growing community of inventors

Burlington, VT, United States of America

Ryan A Jurasek