Micron Technology Incorporated

Yang Lu

Kang-Yong Kim

Sujeet V Ayyapureddi

Markus H Geiger

John Christopher Sancon

Edmund Gieske

Ameen D Akel

Hyun Yoo Lee

Nathaniel J Meier

Amitava Majumdar

Cagdas Dirik

Mark K Hadrick

Robert Michael Walker

Christopher Heaton Stoddard

Zhenming Zhou

Jiangli Zhu

Danilo Caraccio

Tingjun Xie

Qing Liang

Elliott Clifford Cooper-Balis

Francesco Douglas Verna-Ketel

Yu-Sheng Hsu

Elliott C Cooper-balis

Creston M Dupree

Smruti Subhash Jhaveri

Marco Sforzin

Paolo Amato

Hyunyoo Lee

Keun Soo Song

Xinyu Wu

Erik T Barmon

Ke Wei Chan

Niccolo′ Izzo

Niccolo Izzo

This document describes apparatuses and techniques for die-based rank management for a memory system. In various aspects, a die-based rank controller (controller) can determine which memory dies of a memory device are not functional to store data and correlate rank selections of a memory system to ranks of other memory dies (e.g., functional memory dies). The controller may store information that indicates the correlation or mapping of the rank selections to the ranks of the other memory dies to enable access to those ranks of the memory system. In some aspects, the controller receives a command to access the memory device with a rank selection, and the controller enables access to a corresponding rank based on the information. By so doing, aspects of die-based rank management enable memory packages with non-functional memory dies to be used instead of discarded, which can increase production utilization or lower manufacturing costs.

Die-based rank management

Described apparatuses and methods relate to adaptive wordline refresh for a memory system that may support a nondeterministic protocol. To help manage power delivery networks in a memory system, a memory device can include logic that can stagger activation of multiple wordlines that are to be activated or refreshed approximately simultaneously. The logic circuitry can be coupled between wordlines that are to be activated and delay propagation of the activation signal. Thus, a first group of wordlines (e.g., 'before' the logic circuitry) are activated by the signal, but activation of a second group of wordlines (e.g., “after” the logic circuitry), is delayed, reducing the peak current draw. Additional logic circuitries can be coupled between the wordlines to divide the wordlines into multiple groups, thereby staggering the activation of wordlines that are activated by a refresh command, which can reduce the peak current draw and power consumption.

Adaptive wordline refresh

Described apparatuses and methods enable a system including at least one memory device to load different address scramble patterns on dies of the memory device. The address scramble patterns may include the logical-to-physical conversion of rows in the memory device or the memory dies. In aspects, the apparatuses and methods can change the address scrambles at different intervals, such as after a power reset or when the data stored on the memory device is invalid, not current, flushable, or erasable. The described aspects may reduce effectiveness of usage-based disturb attacks used by malicious actors to discover a layout of a type of particular memory device or memory die.

Dynamic address scramble

There are provided systems and methods that include at least one memory that has a plurality of memory cells. The cells may be disposed in rows and columns. The device can further include a controller that is communicatively coupled to the at least one memory, and the controller may be configured by its hardware topology and its instruction set and/or by a communicatively coupled processor or higher-level system or subsystem to maintain data integrity in the at least one memory and/or to prevent or mitigate malicious access patterns that may compromise the at least one memory. The controller may be configured to execute a deterministic protocol in conjunction with or sequentially to a probabilistic protocol to achieve one or more of the above-noted functions.

Row access strobe (RAS) clobber and row hammer failure mitigation

In various examples, refreshing a bank can include receiving a refresh command, wherein the refresh command comprises selector bits and receiving mode register bits from the mode registers. Refreshing a bank can also include refreshing a number of banks from the plurality of banks utilizing the mode register bits and the selector bits.

Bank selection for refreshing

Described apparatuses and methods relate to a bank-level self-refresh for a memory system. A memory device can include logic that implements self-refresh operations in the memory device. The logic may perform self-refresh operations on a set of banks of the memory device that is less than all banks within the memory device. The set of banks of the memory device may be determined such that the peak current in a power distribution network of the memory device is bounded when the self-refresh operation is performed. Accordingly, bank-level self-refresh can reduce a cost of the memory device of a memory system by enabling use of a less complicated power distribution network. The bank-level self-refresh may also be implemented with different types of refresh operations. Amongst other scenarios, bank-level self-refresh can be deployed in memory-expansion environments.

Bank-level self-refresh

Described apparatuses and methods relate to selectively disabling a die that may be included in a multiple-die package without necessarily disabling all the remaining dies within the package. A nonvolatile circuit, such as one or more fuses, may be included within individual dies and/or otherwise incorporated within the package. The nonvolatile circuit maintains a value for the die that is indicative of the operability of the die. Die disablement logic is operatively coupled to the nonvolatile circuit and can disable the die based on the value indicating that the die is unusable. The disabling of the die by the die disablement logic may be controlled by an override signal that allows the disabling or prevents the logic from disabling the die. Thus, the die disablement logic can prevent a defective die from functioning, but the die disablement logic may be overridden for testing or debugging.

Die disablement

There are provided systems and methods that include at least one memory that has a plurality of memory cells. The cells may be disposed in rows and columns. The device can further include a controller that is communicatively coupled to the at least one memory, and the controller may be configured by its hardware topology and its instruction set and/or by a communicatively coupled processor or higher level system or subsystem to maintain data integrity in the at least one memory and/or to prevent or mitigate malicious access patterns that may compromise the at least one memory. The controller may be configured to execute a deterministic protocol in conjunction with or sequentially to a probabilistic protocol to achieve one or more of the above-noted functions.

Row access strobe (RAS) clobber and row hammer failures using a deterministic protocol

An apparatus can include memory devices and a memory controller coupled to the memory devices via memory channels. The memory channels can disable a first memory channel associated with a first memory die in a respective memory chip of a memory device and perform a memory operation via a second memory channel involving a second memory die in the respective memory chip.

Memory channel disablement

Described apparatuses and methods facilitate bus training with multiple dice, such as multiple memory dice. A controller can communicate with multiple dice over a bus to perform bus training by sending a test pattern and receiving in return a feedback pattern indicative of the bits detected by the dice. Because suitable signal timing can differ between dice, even those using the same bus, the controller may attempt to train each die separately from the others. In some situations, however, individualized training may be infeasible. To accommodate such situations, logic associated with two or more dice can combine the bits as detected from the test pattern into a combined feedback pattern. A timing parameter that is jointly suitable for multiple dice can be determined, and the bus training may be concluded, responsive to the combined feedback pattern matching the test pattern. The multiple dice may be stacked or linked.

Bus training with interconnected dice

Described apparatuses and methods facilitate bus training with multiple dice, such as multiple memory dice. A controller can communicate with multiple dice to perform bus training by sending a test pattern and receiving in return a feedback pattern indicative of the bits detected by the dice. Because suitable signal timing can differ between dice, even those using the same bus, a controller may train each die separately from the others. In some situations, however, individualized training may be infeasible. To accommodate such situations, logic associated with two or more dice can combine, using at least one logical operation, bits as detected from the test pattern into a combined feedback pattern. A timing parameter that is jointly suitable for multiple dice can be determined, and the bus training may be concluded, responsive to the combined feedback pattern matching the test pattern. The multiple dice may be stacked or linked.

An energy-efficient and area-efficient, mitigation of errors in a memory media device that are caused by row hammer attacks and the like is described. The detection of errors is deterministically performed while maintaining, in an SRAM, a number of row access counters that is smaller than the total number of rows protected in the memory media device. The reduction of the number of required counters is achieved by aliasing a plurality of rows that are being protected to each counter. The mitigation may be implemented on a per-bank basis, per-channel basis or per-memory media device basis. The memory media device may be DRAM.

Aliased row hammer detector

Described apparatuses and methods relate to erroneous select die access (SDA) detection for a memory system. A memory system may include a memory controller and a memory device that are capable of implementing an SDA protocol that enables selective memory die access to multiple memory devices that couple to a command bus. A memory device can include logic that determines if signaling that conflicts with the SDA protocol is detected. If it is determined that conflicting signaling is detected, the logic may provide an indication of the conflicted signaling. In doing so, the erroneous SDA detection described herein may reduce the likelihood of a memory device erroneously masking memory dice, thereby limiting the memory device from exhibiting unexpected, and in some cases, dangerous behavior.

Erroneous select die access (SDA) detection

Described apparatuses and methods relate to adaptive memory registers for a memory system that may support a nondeterministic protocol. To help manage power-delivery networks in a memory system, a device includes logic that can write values to memory registers associated with memory blocks of a memory array. The values indicate whether an associated memory block has been refreshed within a refresh interval. Other logic can read the registers to determine whether a block has been refreshed. The device also includes logic that can access data indicating a row address that was most recently, or is next to be, refreshed and write values representing the address to another register. The register can be read by other logic to determine whether a wordline potentially affected by an activation-based disturb event is near to being refreshed. These techniques can reduce the number of refresh operations performed, saving power and reducing costs.

Adaptive memory registers

Systems and methods for finite time counting period counting of infinite data streams is presented. In particular example systems and methods enable counting row accesses to a memory media device over predetermined time intervals in order to deterministically detect row hammer attacks on the memory media device. Example embodiments use two identical tables that are reset at times offset in relation to each other in a ping-pong manner in order to ensure that there exists no false negative detections. The counting techniques described in this disclosure can be used in various types of row hammer mitigation techniques and can be implemented in content addressable memory or another type of memory. The mitigation may be implemented on a per-bank basis, per-channel basis or per-memory media device basis. The memory media device may be a dynamic random access memory type device.

Growing community of inventors

Boise, ID, United States of America

Yang Lu