Abstract: The embodiments described herein describe technologies for memory systems. One implementation of a memory system includes a motherboard substrate with multiple module sockets, at least one of which is populated with a memory module. A first set of data lines is disposed on the motherboard substrate and coupled to the module sockets. The first set of data lines includes a first subset of point-to-point data lines coupled between a memory controller and a first socket and a second subset of point-to-point data lines coupled between the memory controller and a second socket. A second set of data lines is disposed on the motherboard substrate and coupled between the first socket and the second socket. The first and second sets of data lines can make up a memory channel.

Abstract: An integrated circuit including a multiplier-accumulator execution pipeline including a plurality of multiplier-accumulator circuits to process the data, using filter weights, via a plurality of multiply and accumulate operations. The integrated circuit includes first conversion circuitry, coupled the pipeline, having inputs to receive a plurality of sets of data, wherein each set of data includes a plurality of data, Winograd conversion circuitry to convert each set of data to a corresponding Winograd set of data, floating point format conversion circuitry, coupled to the Winograd conversion circuitry, to convert the data of each Winograd set of data to a floating point data format. In operation, the multiplier-accumulator circuits are configured to: perform the plurality of multiply and accumulate operations using the data of the plurality of Winograd sets of data from the first conversion circuitry and the filter weights, and generate output data based on the multiply and accumulate operations.

Abstract: A memory module includes cache of relatively fast and durable dynamic, random-access memory (DRAM) in service of a larger amount of relatively slow and wear-sensitive nonvolatile memory. Local controller manages communication between the DRAM cache and nonvolatile memory to accommodate disparate access granularities, reduce the requisite number of memory transactions, and minimize the flow of data external to nonvolatile memory components.

Abstract: An integrated circuit device is disclosed including core circuitry and interface circuitry. The core circuitry outputs in parallel a set of data bits, while the interface circuitry couples to the core circuitry. The interface circuitry receives in parallel a first number of data bits among the set of data bits from the core circuitry and outputs in parallel a second number of data bits. The ratio of the first number to the second number is a non-power-of-2 value.

Abstract: Memory controllers, devices, modules, systems and associated methods are disclosed. In one embodiment, an integrated circuit (IC) memory component is disclosed that includes a memory core, a primary interface, and a secondary interface. The primary interface includes data input/output (I/O) circuitry and control/address (C/A) input circuitry, and accesses the memory core during a normal mode of operation. The secondary interface accesses the memory core during a fault mode of operation.

Abstract: A memory system includes a memory controller with a plurality N of memory-controller blocks, each of which conveys independent transaction requests over external request ports. The request ports are coupled, via point-to-point connections, to from one to N memory devices, each of which includes N independently addressable memory blocks. All of the external request ports are connected to respective external request ports on the memory device or devices used in a given configuration. The number of request ports per memory device and the data width of each memory device changes with the number of memory devices such that the ratio of the request-access granularity to the data granularity remains constant irrespective of the number of memory devices.

Abstract: Memory controllers, devices, modules, systems and associated methods are disclosed. In one embodiment, a memory controller is disclosed. The memory controller includes write queue logic that has first storage to temporarily store signal components of a write operation. The signal components include an address and write data. A transfer interface issues the signal components of the write operation to a bank of a storage class memory (SCM) device and generates a time value. The time value represents a minimum time interval after which a subsequent write operation can be issued to the bank. The write queue logic includes an issue queue to store the address and the time value for a duration corresponding to the time value.

Abstract: An integrated circuit includes a delay circuit and first and second interface circuits. The delay circuit delays a first timing signal by an internal delay to generate an internal timing signal. The first interface circuit communicates data to an external device in response to the internal timing signal. The second interface circuit transmits an external timing signal for capturing the data in the external device. An external delay is added to the external timing signal in the external device to generate a delayed external timing signal. The delay circuit sets the internal delay based on a comparison between the delayed external timing signal and a calibration signal transmitted by the first interface circuit.

Abstract: In a chip-to-chip signaling system includes at least one signaling link coupled between first and second ICs, the first IC has an interface coupled to the signaling link and timed by a first interface timing signal. The second IC has an interface coupled to the signaling link and timed by a second interface timing signal that is mesochronous with respect to the first interface timing signal. The second IC further has phase adjustment circuitry that adjusts a phase of the second interface timing signal using a digital counter implemented with Josephson-junction circuit elements.

Abstract: A memory system employs an addressing scheme to logically divide rows of memory cells into separate contiguous regions, one for data storage and another for error detection and correction (EDC) codes corresponding to that data. Data and corresponding EDC codes are stored in the same row of the same bank. Accessing data and corresponding EDC code in the same row of the same bank advantageously saves power and avoids bank conflicts. The addressing scheme partitions the memory without requiring the requesting processor to have an understanding of the memory partition.

Abstract: A system includes a memory controller and a memory device having a command interface, refresh circuitry, control logic, and a plurality of memory banks, each with a plurality of rows of memory cells. The command interface is operable to receive a refresh command from a memory controller and the refresh circuitry is configured to perform one or more refresh operations to refresh data stored in at least one bank of the plurality of memory banks during a refresh time interval in response to the refresh command from the memory controller. The control logic is to configure the command interface to enter a calibration mode during the refresh time interval, and the command interface is configured to perform a calibration operation in the calibration mode during the refresh time interval.

Abstract: A controller includes a link interface that is to couple to a first link to communicate bi-directional data and a second link to transmit unidirectional error-detection information. An encoder is to dynamically add first error-detection information to at least a portion of write data. A transmitter, coupled to the link interface, is to transmit the write data. A delay element is coupled to an output from the encoder. A receiver, coupled to the link interface, is to receive second error-detection information corresponding to at least the portion of the write data. Error-detection logic is coupled to an output from the delay element and an output from the receiver. The error-detection logic is to determine errors in at least the portion of the write data by comparing the first error-detection information and the second error-detection information, and, if an error is detected, is to assert an error condition.

Abstract: An integrated circuit including control/configure circuitry which interfaces with a plurality of interconnected (e.g., serially) multiplier-accumulator circuits and/or one or more rows of interconnected (e.g., serially) multiplier-accumulator circuits. The control/configure circuitry may include a plurality of control/configure circuits, each control/configure circuit interfaces with at least one multi-bit MAC execution pipeline, wherein each pipeline includes a plurality of interconnected (e.g., serially) multiplier-accumulator circuits. Each control/configure circuit may include one or more (or all) of (i) a configurable input data signal path to provide data to the MACs of the pipeline during the execution sequence(s), (ii) a configurable accumulation data path for the ongoing/accumulating MAC accumulation totals generated by the MACs during an execution sequence, and (iii) a configurable output data path for the output data generated by execution sequence (i.e.

Abstract: An integrated circuit comprising a plurality MAC processors, interconnected into a linear pipeline, configurable to process input data, wherein each MAC processor includes (A) a multiplier and (B) an accumulator circuit, and (C) a plurality of rotate input data paths, wherein each rotate input data path couples two sequential MAC processors of the linear pipeline including an input of the multiplier circuit of a first MAC processor of sequential MAC processors to an input of the multiplier circuit of the immediately following MAC processor of the associated sequential MAC processors of the pipeline—wherein each rotate input data path is configurable to provide rotate input data from a first MAC processor of sequential MAC processors of the linear pipeline to the immediately following MAC processor of the associated sequential MAC processors thereby forming a serial circular path via the plurality of rotate input data paths.

Abstract: A hybrid volatile/non-volatile memory module employs a relatively fast, durable, and expensive dynamic, random-access memory (DRAM) cache to store a subset of data from a larger amount of relatively slow and inexpensive nonvolatile memory (NVM). A module controller prioritizes accesses to the DRAM cache for improved speed performance and to minimize programming cycles to the NVM. Data is first written to the DRAM cache where it can be accessed (written to and read from) without the aid of the NVM. Data is only written to the NVM when that data is evicted from the DRAM cache to make room for additional data. Mapping tables relating NVM addresses to physical addresses are distributed throughout the DRAM cache using cache line bits that are not used for data.

Abstract: A semiconductor IC device comprises a timing circuit to transfer a timing signal, the timing circuit being configured to receive a first test signal and to effect a delay in the timing signal in response to the first test signal, the first test signal including a first timing event. The semiconductor IC device further comprises an interface circuit configured to transfer the data signal in response to the timing signal, the interface circuit being further configured to receive a second test signal and to effect a delay in the data signal in response to the second test signal, the second test signal including a second timing event that is related to the first timing event according to a test criterion.

Abstract: A memory module with multiple memory devices includes a buffer system that manages communication between a memory controller and the memory devices. Each memory device supports an access mode and a low-power mode, the latter used to save power for devices that are not immediately needed. The module provides granular power management using a chip-select decoder that decodes chip-select signals from the memory controller into power-state signals that determine which of the memory devices are in which of the modes. Devices can thus be brought out of the low-power mode in relatively small numbers, as needed, to limit power consumption.

Abstract: A method and system that provides for execution of a first calibration sequence, such as upon initialization of a system, to establish an operation value, which utilizes an algorithm intended to be exhaustive, and executing a second calibration sequence from time to time, to measure drift in the parameter, and to update the operation value in response to the measured drift. The second calibration sequence utilizes less resources of the communication channel than does the first calibration sequence. In one embodiment, the first calibration sequence for measurement and convergence on the operation value utilizes long calibration patterns, such as codes that are greater than 30 bytes, or pseudorandom bit sequences having lengths of 2N?1 bits, where N is equal to or greater than 7, while the second calibration sequence utilizes short calibration patterns, such as fixed codes less than 16 bytes, and for example as short as 2 bytes long.

Abstract: Improvements are disclosed for “leveling” or averaging out more evenly the number of activate/precharge cycles seen by the rows of a memory component, so that one or more particular rows are not excessively stressed (relative to the other rows). In one embodiment, a memory controller includes remapping facilities arranged to move data stored in a physical row from RPK to RPK? and modify the mapping from logical row RLK while minimizing impact on normal read/write operations. Remapping operations may be scheduled relative to refresh or other maintenance operations. Remapping operations may be conditionally deferred so as to minimize performance impact.

Abstract: An integrated circuit including a plurality of processing components to process image data of a plurality of image frames, wherein each image frame includes a plurality of stages. Each processing component includes a plurality of execution pipelines, wherein each pipeline includes a plurality of multiplier-accumulator circuits configurable to perform multiply and accumulate operations using image data and filter weights, wherein: (i) a first processing component is configurable to process all of the data associated with a first plurality of stages of each image frame, and (ii) a second processing component of the plurality of processing components is configurable to process all of the data associated with a second plurality of stages of each image frame. The first and second processing component processes data associated with the first and second plurality of stages, respectively, of a first image frame concurrently.