Showing papers on "Memory refresh published in 2005"

DRAMsim: a memory system simulator

Abstract: As memory accesses become slower with respect to the processor and consume more power with increasing memory size, the focus of memory performance and power consumption has become increasingly important. With the trend to develop multi-threaded, multi-core processors, the demands on the memory system will continue to scale. However, determining the optimal memory system configuration is non-trivial. The memory system performance is sensitive to a large number of parameters. Each of these parameters take on a number of values and interact in fashions that make overall trends difficult to discern. A comparison of the memory system architectures becomes even harder when we add the dimensions of power consumption and manufacturing cost. Unfortunately, there is a lack of tools in the public-domain that support such studies. Therefore, we introduce DRAMsim, a detailed and highly-configurable C-based memory system simulator to fill this gap. DRAMsim implements detailed timing models for a variety of existing memories, including SDRAM, DDR, DDR2, DRDRAM and FB-DIMM, with the capability to easily vary their parameters. It also models the power consumption of SDRAM and its derivatives. It can be used as a standalone simulator or as part of a more comprehensive system-level model. We have successfully integrated DRAMsim into a variety of simulators including MASE [15], Sim-alpha [14], BOCHS[2] and GEMS[13]. The simulator can be downloaded from www.ece.umd.edu/dramsim.

Storage device employing a flash memory

Abstract: A semiconductor disk wherein a flash memory into which data is rewritten in block unit is employed as a storage medium, the semiconductor disk including a data memory in which file data are stored, a substitutive memory which substitutes for blocks of errors in the data memory, an error memory in which error information of the data memory are stored, and a memory controller which reads data out of, writes data into and erases data from the data memory, the substitutive memory and the error memory. Since the write errors of the flash memory can be remedied, the service life of the semiconductor disk can be increased.

Raid controller using capacitor energy source to flush volatile cache data to non-volatile memory during main power outage

Abstract: A write-caching RAID controller includes a CPU that manages transfers of posted-write data from host computers to a volatile memory and transfers of the posted-write data from the volatile memory to a redundant array of storage devices when a main power source is supplying power to the RAID controller. A memory controller transfers the posted-write data received from the host computers to the volatile memory and transfers the posted-write data from the volatile memory for transfer to the redundant array of storage devices as managed by the CPU. The memory controller flushes the posted-write data from the volatile memory to the non-volatile memory when main power fails, during which time capacitors provide power to the memory controller, volatile memory, and non-volatile memory, but not to the CPU, in order to reduce the energy storage requirements of the capacitors. During main power provision, the CPU programs the memory controller with information needed to perform the flush operation, such as the location and size of the posted-write data in the volatile memory and various flush operation characteristics.

Nonvolatile semiconductor memory device capable of uniformly inputting/outputting data

Abstract: At the time of an operation of writing data to a specific memory cell in a memory block, a semiconductor memory device applies a write voltage for a predetermined period and, after that, performs a verifying operation by using a sense amplifier circuit and a comparator. When it is found as a result of the verifying operation that writing to the memory cell is insufficient, the writing operation is performed again by an instruction of a memory control circuit. At this time, the memory control circuit adjusts a write voltage.

Flash Memory Storage System and Method

Abstract: A flash memory storage system includes a memory array containing a plurality of memory cells and a controller for controlling the flash memory array. The controller dedicates a first group of memory cells to operate with a first number of bits per cell and a second, separate group of memory cells to operate with a second number of bits per cell. A mechanism is provided to apply wear leveling techniques separately to the two groups of cells to evenly wear out the memory cells.

Apparatus and method for reading an array of nonvolatile memory cells including switchable resistor memory elements

Abstract: A non-volatile memory cell includes a switchable resistor memory element in series with a switch device. An array of such cells may be programmed using only positive voltages. A method for programming such cells also supports a direct write of both 0 and 1 data states without requirement of a block erase operation, and is scalable for use with relatively low voltage power supplies. A method for reading such cells reduces read disturb of a selected memory cell by impressing a read bias voltage having a polarity opposite that of a set voltage employed to change the switchable resistor memory element to a low resistance state. Such programming and read methods are well suited for use in a three-dimensional memory array formed on multiple levels above a substrate, particularly those having extremely compact array line drivers on very tight layout pitch.

NAND flash memory management

Abstract: A memory controller is utilized to overcome NAND flash memory's propensity for comprising bad blocks of memory. The memory controller utilizes minimal hardware and is essentially transparent to a device requesting access to the NAND memory. A NAND flash memory device is configured to comprise a set of main blocks of memory and a set of auxiliary blocks of memory. Each block is divided into pages of memory and each page includes metadata. The metadata includes a block status indicator, indicating whether a block is good or bad. When receiving a request to access a page in the NAND flash memory, if the block in which the page resides is good, that block is accessed. If the block is bad, auxiliary memory is searched until a block containing the address of the bad block in its metadata is found. The found block is accessed in lieu of the bad block.

Memory system for legacy hosts

Abstract: A non-volatile memory device is provided with a controller and includes method that controls memory operations and to emulate the memory and communication characteristics of a legacy memory device. In this way, the memory device is compatible with a host that was originally designed to operate the legacy memory device. In particular, the controller performs the emulation to the host taking into account differences such as multibit memory, error correction requirement, memory support of overwrites, and erasable block sizes.

Storage device and information processing system

Abstract: A storage device able to make a redundant write operation of unselected data unnecessary and able to optimize an arrangement of pages to a state having a high efficiency for rewriting, wherein the storage device has a first memory unit, a second memory unit having a different access speed from the first memory, and a control circuit, wherein the control circuit has a function of timely moving the stored data in two ways between the first memory unit and the second memory unit having different access speeds in reading or rewriting.

Semiconductor integrated circuit device and magnetic memory device capable of maintaining data integrity

Abstract: Magnetic memory devices integrated together with a logic circuit on a common semiconductor chip are arranged to have layouts mirror-symmetrical (mirror inversion) with respect to an axis parallel to a magnetization-hard axis of a magneto-resistance element of a magnetic memory cell in the magnetic memory device. The logic circuit is arranged between the magnetic memory devices. The magnetic memory device capable accurately of maintaining integrity in logical level between write data and read data is achieved.

Adaptive programming technique for a re-writable conductive memory device

Abstract: A programming circuit is provided. As a conductive memory cell is programmed, its resistance changes. The provided programming circuit monitors the changing resistance while programming the memory cell. The programming circuit can be used to only program the memory cell for as long as programming is actually needed. Additionally, the programming circuit can be used to only program the memory cell when it has a value that needs to be changed.

Flash memory control method, flash memory system using the control method and flash memory device using the control method

Abstract: A method of correcting errors of a flash memory comprises steps of modifying the data of a group of memory units, each having a plurality of flash memory cells adapted to erasing data therefrom and writing data therein, checking for the presence or absence of an error of not properly modifying the data of the group of memory units and determining the completion of proper modification of the data of the group of memory units provided that an error is detected and the error can be corrected.

Swapping to Remote Memory over InfiniBand: An Approach using a High Performance Network Block Device

Abstract: Traditionally, operations with memory on other nodes (remote memory) in cluster environments interconnected with technologies like Gigabit Ethernet have been expensive with latencies several magnitudes slower than local memory accesses. Modern RDMA capable networks such as InfiniBand and Quadrics provide low latency of a few microseconds and high bandwidth of up to 10 Gbps. This has significantly reduced the latency gap between access to local memory and remote memory in modern clusters. Remote idle memory can be exploited to reduce the memory pressure on individual nodes. This is akin to adding an additional level in the memory hierarchy between local memory and the disk, with potentially dramatic performance improvements especially for memory intensive applications. In this paper, we take on the challenge to design a remote paging system for remote memory utilization in InfiniBand clusters. We present the design and implementation of a high performance networking block device (HPBD) over InfiniBand fabric, which serves as a swap device of kernel virtual memory (VM) system for efficient page transfer to/from remote memory servers. Our experiments show that using HPBD, quick sort performs only 1.45 times slower than local memory system, and up to 21 times faster than local disk. And our design is completely transparent to user applications. To the best of our knowledge, it is the first work of a remote pager design using InfiniBand for remote memory utilization

DRAM chip device well-communicated with flash memory chip and multi-chip package comprising such a device

Abstract: An SDRAM memory chip device comprises a non-volatile memory controller for operating a non-volatile memory, e.g., a NAND-flash, and a FIFO memory buffer. The FIFO memory buffer serves to operate background store and load operations between a FIFO buffer array and the non-volatile memory, while a host system such as a CPU exchanges data with the SDRAM work memory. The SDRAM memory chip device, therefore, has at least two additional pins as compared with conventional SDRAM standard for generating a set of additional commands. These commands are employed by the FIFO memory buffer to manage the data transfer between the FIFO buffer and each of the non-volatile memory and the volatile SDRAM memory. Two further pins reflecting the flash memory status provide appropriate issuance of load or store signals by the host system.

Power throttling in a memory system

Abstract: A memory system is disclosed. The memory system includes a memory controller coupled to one or more memory modules, at least one of the memory modules including a buffer. The memory controller is configured to convey a command to at least one of the memory modules in response to detecting that no memory requests addressed to the at least one of the memory modules have been received during a specified window of time. In response to the command, the buffer of the at least one of the memory modules is configured to enter a reduced power state. The specified window of time may be either a specified number of memory refresh intervals or buffer sync intervals. The memory controller maintains a count of memory refresh or buffer sync intervals.

Sense amplifier circuitry and architecture to write data into and/or read from memory cells

Abstract: A technique of, and circuitry for sampling, sensing, reading and/or determining the data state of a memory cell of a memory cell array (for example, a memory cell array having a plurality of memory cells which consist of an electrically floating body transistor). In one embodiment, sense amplifier circuitry is relatively compact and pitched to the array of memory cells such that a row of data may be read, sampled and/or sensed during a read operation. In this regard, an entire row of memory cells may be accessed and read during one operation which, relative to at least architecture employing multiplexer circuitry, may minimize, enhance and/or improve read latency and read access time, memory cell disturbance and/or simplify the control of the sense amplifier circuitry and access thereof. The sense amplifier circuitry may include write back circuitry to modify or 're-store' the data read, sampled and/or sensed during a read operation and/or a refresh operation in the context of a DRAM array. The sense amplifier circuitry of this embodiment restores and/or refreshes data in an entire row of volatile and/or destructive read type memory cells in parallel. This architecture may minimize, enhance and/or improve write back and read latency parameters, relative to at least architecture employing multiplexer circuitry. Also, data that has been read, sampled and/or sensed by the sense amplifier circuitry during a read operation may be modified before being written back to one or more of the memory cells of the selected row of the array of memory cells.

High-density memory module utilizing low-density memory components

Abstract: A memory module comprises a plurality of memory components. Each memory component has a first bit width. The plurality of memory components are configured as one or more pairs of memory components. Each pair of memory components simulates a single virtual memory component having a second bit width which is twice the first bit width.

Redistribution of memory to reduce computer system power consumption

Abstract: A method, apparatus and system are disclosed for redistributing memory allocation to portions of dynamic random access memory (DRAM) and dual in-line memory module (DIMM) devices that are underutilized, in order to balance memory usage more evenly amongst active devices so as to limit the amount of power and the thermal load consumed by an individual memory component. The disclosed method, apparatus and system are capable of identifying and tracking memory usage to minimize power consumption in a way that lessens the detrimental effects of “throttling” or reduced power modes for memory devices.

Power management via DIMM read operation limiter

Abstract: A method and system for enabling directed temperature/power management at the DIMM-level and/or DRAM-level utilizing intelligent scheduling of memory access operations received at the memory controller. Hot spots within the memory subsystem, caused by operating the DIMMs/DRAMs above predetermined/preset threshold power/temperature values for operating a DIMM and/or a DRAM, are avoided/controlled by logic within the memory controller. The memory controller logic throttles the number/frequency at which commands (read/write operations) are issued to the specific DIMM/DRAM based on stored parameter values and tracking of outstanding operations issued to the memory subsystem devices.

Non-volatile memory and method with power-saving read and program-verify operations

Abstract: A non-volatile memory device capable of reading and writing a large number of memory cells with multiple read/write circuits in parallel has features to reduce power consumption during read, and program/verify operations. A read or program verify operation includes one or more sensing cycles relative to one or more demarcation threshold voltages to determine a memory state. In one aspect, selective memory cells among the group being sensed in parallel have their conduction currents turned off when they are determined to be in a state not relevant to the current sensing cycle. In another aspect, a power-consuming period is minimized by preemptively starting any operations that would prolong the period. In a program/verify operation cells not to be programmed have their bit lines charged up in the program phase. Power is saved when a set of these bit lines avoids re-charging at every passing of a program phase.

Improving energy efficiency by making DRAM less randomly accessed

Abstract: Existing techniques manage power for the main memory by passively monitoring the memory traffic, and based on which, predict when to power down and into which low-power state to transition. However, passively monitoring the memory traffic can be far from being effective as idle periods between consecutive memory accesses are often too short for existing power-management techniques to take full advantage of the deeper power-saving state implemented in modem DRAM architectures. In this paper, the authors proposed a new technique that will actively reshape the memory traffic to coalesce short idle periods - which were previously unusable for power management - into longer ones, thus enabling existing techniques to effectively exploit idleness in the memory.

Fully buffered DIMM variable read latency

Abstract: Memory control that access memory devices having different read latencies is described. In on embodiment, a memory controller may include read latency logic to identify and match received read data with read commands to the memory devices based on values indicative of the read latency for the memory devices. In another embodiment, the memories may include read delay control to insert an amount of delay into the time a memory device takes in responding to a read command.

Dynamic command and/or address mirroring system and method for memory modules

Abstract: A memory module includes a memory hub that couples signals to memory devices mounted on opposite first and second surfaces of a memory module substrate. The memory devices are mounted in mirrored configuration with mirrored terminals of memory devices on opposite surfaces being interconnected. A memory hub mounted on each module alters the configuration of address and/or command signals coupled to the memory devices depending upon whether the memory devices on the first surface of the substrate or the memory devices on the second surface of the substrate are being accessed. Alternatively, the configuration of the address and/or command signals coupled to mirrored memory devices may be altered by a register mounted on the substrate that is coupled to the memory devices or by a memory controller coupled directly to memory devices on one or more memory modules.

Streaming reads for early processing in a cascaded memory subsystem with buffered memory devices

Abstract: A memory subsystem completes multiple read operations in parallel, utilizing the functionality of buffered memory modules in a daisy chain topology. A variable read latency is provided with each read command to enable memory modules to run independently in the memory subsystem. Busy periods of the memory device architecture are hidden by allowing data buses on multiple memory modules attached to the same data channel to run in parallel rather than in series and by issuing reads earlier than required to enable the memory devices to return from a busy state earlier. During scheduling of reads, the earliest received read whose target memory module is not busy is immediately issued at a next command cycle. The memory controller provides a delay parameter with each issued read. The number of cycles of delay is calculated to allow maximum utilization of the memory modules' data bus bandwidth without causing collisions on the memory channel.

Fast path memory read request processing in a multi-level memory architecture

Abstract: A circuit arrangement and method selectively reorder speculatively issued memory read requests being communicated to a lower memory level in a multi-level memory architecture. In particular, a memory read request that has been speculatively issued to a lower memory level prior to completion of a cache lookup operation initiated in a cache memory in a higher memory level may be reordered ahead of at least one previously received and pending request awaiting communication to the lower memory level. By doing so, the latency associated with the memory read request is reduced when the request results in a cache miss in the higher level memory, and as a result, system performance is improved.

Memory cell having improved read stability

Abstract: A memory cell for use in a memory array includes a storage element for storing a logical state of the memory cell, a write circuit and a read circuit. The write circuit is operative to selectively connect a first node of the storage element to at least a first write bit line in the memory array in response to a write signal for selectively writing the logical state of the memory cell. The read circuit includes a substantially high impedance input node connected to the storage element and an output node connectable to a read bit line of the memory array. The read circuit is configured to generate an output signal at the output node which is representative of the logical state of the storage element in response to a read signal applied to the read circuit. The memory cell is configured such that the write circuit is disabled during a read operation of the memory cell so as to substantially isolate the storage element from the first write bit line during the read operation. A strength of at least one transistor device in the storage element is separately optimized relative to a strength of at least one transistor device in the write circuit and/or the read circuit.

Various apparatuses and methods for reduced power states in system memory

Abstract: A method, apparatus, and system are described in which a memory controller may have two or more registers to create and track zones of memory in a volatile memory device. The memory controller controls a power consumption state of a first zone of memory in the volatile memory device and a second zone of memory within the first volatile memory device on an individual basis; and one or more memory arrays contained within the first volatile memory device.

Memory command delay balancing in a daisy-chained memory topology

Abstract: A methodology for a daisy-chained memory topology wherein, in addition to the prediction of the timing of receipt of a response from a memory module (DIMM), the memory controller can effectively predict when a command sent by it will be executed by the addressee DIMM. By programming DIMM-specific command delay in the DIMM's command delay unit, the command delay balancing methodology according to the present disclosure “normalizes” or “synchronizes” the execution of the command signal across all DIMMs in the memory channel. With such ability to predict command execution timing, the memory controller can efficiently control power profile of all the DRAM devices (or memory modules) on a daisy-chained memory channel. A separate DIMM-specific response delay unit in the DIMM may also be programmed to provide DIMM-specific delay compensation in the response path, further allowing the memory controller to accurately ascertain the timing of receipt of a response thereat, and, hence, to better manage further processing of the response.

Stacked semiconductor memory device

Abstract: The stacked semiconductor memory device of the present invention has the object of reducing the cost of developing a wide variety of memory devices and includes: a memory cell array chip that is equipped with memory cell arrays, an interface chip that is stacked with the memory cell array chip and that is provided with a memory configuration switching circuit for changing the input/output bit configuration of the memory cell arrays, and a plurality of interchip wires for connecting the memory cell array chip and the interface chip.

Memory hub and method for memory sequencing

Abstract: A memory module includes a memory hub coupled to several memory devices. The memory hub includes at least one performance counter that tracks one or more system metrics—for example, page hit rate, prefetch hits, and/or cache hit rate. The performance counter communicates with a memory sequencer that adjusts its operation based on the system metrics tracked by the performance counter.