Abstract: A memory controller and buffers on memory modules each operate in two modes, depending on the type of motherboard through which the controller and modules are connected. In a first mode, the controller transmits decoded chip-select signals independently to each module, and the motherboard data channel uses multi-drop connections to each module. In a second mode, the motherboard has point-to-point data channel and command address connections to each of the memory modules, and the controller transmits a fully encoded chip-select signal group to each module. The buffers operate modally to correctly select ranks or partial ranks of memory devices on one or more modules for each transaction, depending on the mode.

Abstract: A memory controller includes logic circuitry to generate a first data address identifying a location in a first external memory array for storing first data, a first tag address identifying a location in a second external memory array for storing a first tag, a second data address identifying a location in the second external memory array for storing second data, and a second tag address identifying a location in the first external memory array for storing a second tag. The memory controller includes an interface that transfers the first data address and the first tag address for a first set of memory operations in the first and the second external memory arrays. The interface transfers the second data address and the second tag address for a second set of memory operations in the first and the second external memory arrays.

Abstract: In a multirank memory system in which the clock distribution trees of each rank are permitted to drift over a wide range (e.g., low power memory systems), the fine-interleaving of commands between ranks is facilitated through the use of techniques that cause each addressed rank to properly sample commands intended for that rank, notwithstanding the drift. The ability to perform such “microthreading” provides for substantially enhanced memory capacity without sacrificing the performance of single rank systems. This disclosure provides methods, memory controllers, memory devices and system designs adapted to these ends.

Abstract: An integrated-circuit memory component receives, as part of respective first and second memory read transactions, a first column access command that identifies a first volume of data and a second column read command that identifies a second volume of data, the second volume of data being constituted by not more than half as many data bits as the first volume of data. In response to receiving the first column access command, the integrated-circuit memory component transmits the first volume of data as N parallel bit-serial data signals over N external signaling links. In response to receiving the second column access command, the integrated-circuit memory component transmits the second volume of data as M parallel bit-serial data signals over M of the N external signaling links, where M is less than N.

Abstract: The described embodiments provide a system for controlling an integrated circuit memory device by a memory controller. During operation, the system sends a memory-access request from the memory controller to the memory device using a first link. After sending the memory-access request, the memory controller sends to the memory device a command that specifies performing a timing-calibration operation for a second link. The system subsequently transfers data associated with the memory-access request using the second link, wherein the timing-calibration operation occurs between sending the memory-access request and transferring the data associated with the memory-access request.

Abstract: A memory module can be programmed to deliver relatively wide, low-latency data in a first access mode, or to sacrifice some latency in return for a narrower data width, a narrower command width, or both, in a second access mode. The narrow, higher-latency mode requires fewer connections and traces. A controller can therefore support more modules, and thus increased system capacity. Programmable modules thus allow computer manufacturers to strike a desired balance between memory latency, capacity, and cost.

Abstract: A compute system includes an execution unit (e.g. of a CPU) with a memory controller providing access to a hybrid physical memory. The physical memory is “hybrid” in that it combines a cache of relatively fast, durable, and expensive memory (e.g. DRAM) with a larger amount of relatively slow, wear-sensitive, and inexpensive memory (e.g. flash). A hybrid controller component services memory commands from the memory controller component and additionally manages cache fetch and evict operations that keep the cache populated with instructions and data that have a high degree of locality of reference. The memory controller alerts the hybrid controller of available access slots to the cache so that the hybrid controller can use the available access slots for cache fetch and evict operations with minimal interference to the memory controller.

Abstract: The embodiments described herein describe technologies for non-volatile memory persistence in a multi-tiered memory system including two or more memory technologies for volatile memory and non-volatile memory.

Abstract: The embodiments described herein describe technologies of dynamic random access memory (DRAM) components for high-performance, high-capacity registered memory modules, such as registered dual in-line memory modules (RDIMMs). One DRAM component may include a set of memory cells and steering logic. The steering logic may include a first data interface and a second data interface. The first and second data interfaces are selectively coupled to a controller component in a first mode and the first data interface is selectively coupled to the controller component in a second mode and the second data interface is selectively coupled to a second DRAM component in the second mode.

Abstract: A hybrid memory includes cache of relatively fast and durable dynamic, random-access memory (DRAM) in service of a larger amount of relatively slow and wear-sensitive flash memory. An address buffer on the module maintains a static random-access memory (SRAM) cache of addresses for data cached in DRAM.

Abstract: Disclosed is a dynamic random access memory (DRAM) that includes a plurality of data rows and a plurality of tag rows. The DRAM includes a communication interface to receive a first group of address bits. The DRAM includes one or more comparators to generate a tag match indication and one or more set bits based on the first group of address bits and a first group of tag information bits from the plurality of tag rows. The one or more comparators are further to combine, based on the tag match indication, the one or more set bits and the first group of address bits to generate a second group of address bits.

Abstract: A memory controller and buffers on memory modules each operate in two modes, depending on the type of motherboard through which the controller and modules are connected. In a first mode, the controller transmits decoded chip-select signals independently to each module, and the motherboard data channel uses multi-drop connections to each module. In a second mode, the motherboard has point-to-point data channel and command address connections to each of the memory modules, and the controller transmits a fully encoded chip-select signal group to each module. The buffers operate modally to correctly select ranks or partial ranks of memory devices on one or more modules for each transaction, depending on the mode.

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: A stacked semiconductor device is disclosed that includes a plurality of semiconductor dies. Each die has oppositely disposed first and second surfaces, with pads formed on each of the surfaces. A plurality of through-vias connect respective pads on the first surface to respective pads on the second surface. The through-vias include a first group of through-vias coupled to respective I/O circuitry on the semiconductor die and a second group of through-vias not coupled to I/O circuitry on the semiconductor die. The plurality of semiconductor dies are stacked such that the first group of through-vias in a first one of the plurality of semiconductor dies are aligned with respective ones of at least a portion of the second group of through-vias in a second one of the plurality of semiconductor dies.

Abstract: During system initialization, each data buffer device and/or memory device on a memory module is configured with a unique (at least to the module) device identification number. In order to access a single device (rather than multiple buffers and/or memory devices), a target identification number is written to all of the devices using a command bus connected to all of the data buffer devices or memory devices, respectively. The devices whose respective device identification numbers do not match the target identification number are configured to ignore future command bus transactions (at least until the debug mode is turned off.) The selected device that is configured with a device identification number matching the target identification number is configured to respond to command bus transactions.

Abstract: A compute system includes an execution unit (e.g. of a CPU) with a memory controller providing access to a hybrid physical memory. The physical memory is “hybrid” in that it combines a cache of relatively fast, durable, and expensive memory (e.g. DRAM) with a larger amount of relatively slow, wear-sensitive, and inexpensive memory (e.g. flash). A hybrid controller component services memory commands from the memory controller component and additionally manages cache fetch and evict operations that keep the cache populated with instructions and data that have a high degree of locality of reference. The memory controller alerts the hybrid controller of available access slots to the cache so that the hybrid controller can use the available access slots for cache fetch and evict operations with minimal interference to the memory controller.

Abstract: A semiconductor memory system includes a first semiconductor memory die and a second semiconductor memory die. The first semiconductor memory die includes a primary data interface to receive an input data stream during write operations and to deserialize the input data stream into a first plurality of data streams, and also includes a secondary data interface, coupled to the primary data interface, to transmit the first plurality of data streams. The second semiconductor memory die includes a secondary data interface, coupled to the secondary data interface of the first semiconductor memory die, to receive the first plurality of data streams.

Abstract: Memory controllers, devices, modules, systems and associated methods are disclosed. In one embodiment, an integrated circuit (IC) memory controller is disclosed. The IC memory controller includes a first controller command/address (C/A) interface to transmit first and second read commands for first and second read data to a first memory C/A interface of a first bank group of memory. A second command/address (C/A) interface transmits third and fourth read commands for third and fourth read data to a second memory C/A interface of a second bank group of memory. Receiver circuitry receives the first and second read data via a first data link interface and the third and fourth read data via the second data link interface. For a first operating mode, the first and second read data are received after respective first delays following transmission of the first and second read commands and at a first serialization ratio.

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: An integrated circuit comprising a plurality of multiplier-accumulator circuits connected in series to form a linear pipeline to process first data, via performing a plurality of concatenated multiply and accumulate operations, and generate MAC output data, wherein each multiplier-accumulator circuit of the plurality of multiplier-accumulator circuits includes (i) a multiplier to multiply first data by a multiplier weight data and generate a product data, and (ii) an accumulator, coupled to the multiplier of the associated multiplier-accumulator circuit, to add second data and the product data of the associated multiplier to generate sum data. The integrated circuit further includes an activation circuit, connected to the output of the linear pipeline of the plurality of multiplier-accumulator circuits, to receive the MAC output data and process the MAC output data, via a non-linear activation function, to generate MAC pipeline output data.