Abstract: Methods, systems, and devices for valid data identification for garbage collection are described. In connection with writing data to a block of memory cells, a memory system may identify a portion of a logical address space that includes a logical address for the data. The memory system may set a bit of a bitmap, which may indicate that the block includes data having a logical address within a portion of the logical address space corresponding to the bit. The logical address space may be divided into any quantity of portions, each corresponding to a different subset of a logical-to-physical (L2P) table, and the bitmap may include any quantity of corresponding bits. To perform garbage collection on the block, the bitmap may be used to identify one or more subsets of the L2P table to evaluate to determine whether different sets of data within the block are valid or invalid.

Abstract: A memory device includes a memory array including memory cells, a communication interface to a host device, and a memory control unit operatively coupled to the memory array and the communication interface. The memory control unit is configured to encrypt write data received via the communication interface to produce encrypted data, program a portion of the memory cells of the memory array with the encrypted data, read the encrypted data from the portion of the memory cells in response to a memory read request, decrypt the read encrypted data to produce read decrypted data only for portions of the read encrypted data not stored in purged regions of the memory array.

Abstract: Systems, methods, and apparatus related to memory devices. In one approach, a cross-point memory array includes memory cells. A media controller reads one or more first memory cells and determines a read status. The read status indicates an error when reading the first memory cells. In response to this error, the controller refreshes the first memory cells. The controller uses the read status to determine zero-to-one failures associated with the first memory cells. If a number of these failures exceeds a threshold, then a refresh is applied to neighboring memory cells of the first memory cells. The physical addresses for the neighboring memory cells are determined by the controller from the physical addresses for the first memory cells.

Abstract: A method may comprise detecting an error associated with accessing a set of data items. The set of data items are programmed to a respective memory page associated with a stripe of a plurality of stripes. In response to determining that the set of data items comprises one or more codewords, a first data recovery process is performed to recover the one or more codewords based at least in part on RAIN redundancy metadata. In response to determining that the set of data items comprises additional parity metadata, a second data recovery process is performed to recover the additional parity metadata based at least in part on LUN redundancy metadata. In response to determining that the set of data items comprises RAIN redundancy metadata, a first data reconstruction process is performed to regenerate the RAIN redundancy metadata based at least in part on one or more sets of codewords.

Abstract: A memory device to program a group of memory cells to store multiple bits per memory cell. Each bit per memory cell in the group from a page. After determining a plurality of read voltages of the group of memory cells, the memory device can read the multiple pages of the group using the plurality of read voltages. For each respective page in the multiple pages, the memory device can determine a count of first memory cells in the respective page that have threshold voltages higher than a highest read voltage, among the plurality of read voltages, used to read the respective page. The count of the first memory cells can be compared with a predetermined range of a fraction of memory cells in the respective page to evaluate the plurality of read voltages (e.g., whether any of the read voltages is in a wrong voltage range).

Abstract: Disclosed in some examples are improved address prediction and memory preloading that leverages next-delta prediction and/or far-delta prediction for scheduling using a DNN. Previous memory access sequence data that identify one or more memory addresses previously accessed by one or more processors of a system may be processed and then converted into a sequence of delta values. The sequence of delta values are then mapped to one or more classes that are then input to a DNN. The DNN then outputs a predicted future class identifier sequence that represents addresses that the DNN predicts will be accessed by the processor in the future. The predicted future class identifier sequence is then converted back to a predicted delta value sequence and back into a set of one or more predicted addresses.

Abstract: Methods, systems, and devices related to auto-referenced memory cell read techniques are described. The auto-referenced read may encode user data to include a predetermined number of bits having a first logic state prior to storing the user data in memory cells. The auto-referenced read may store a total number of bits of the user data having a first logic state in a separate set of memory cells. Subsequently, reading the user data may be carried out by applying a read voltage to the memory cells storing the user data while monitoring a series of switching events by activating a subset of the memory cells having the first logic state. During the read operation, the auto-referenced read may compare the number of activated memory cells to either the predetermined number or the total number to determine whether all the bits having the first logic state has been detected.

Abstract: Disclosed in some examples are methods, systems, devices, memory devices, and machine-readable mediums for using a non-defective portion of a block of memory on which there is a defect on a different portion. Rather than disable the entire block, the system may disable only a portion of the block (e.g., a first deck of the block) and salvage a different portion of the block (e.g., a second deck of the block).

Abstract: Some embodiments include a memory cell having charge-trapping-material between a semiconductor channel material and a gating region. The charge-trapping-material includes silicon, nitrogen and trap-enhancing-additive. The trap-enhancing-additive includes one or more of carbon, phosphorus, boron and metal. Some embodiments include an integrated assembly having a stack of alternating first and second levels. The first levels include conductive structures and the second levels are insulative. Channel-material-pillars extend through the stack. Charge-trapping-regions are along the channel-material-pillars and are between the channel-material-pillars and the conductive structures. The charge-trapping-regions include a charge-trapping-material which contains silicon, nitrogen and trap-enhancing-additive. The trap-enhancing-additive includes one or more of carbon, phosphorus, boron and metal.

Abstract: Methods, systems, and devices for memory operations are described. A first code for detecting one or more errors in a first set of bits of data and a second code for detecting one or more errors in a second set of bits of data may be generated. The first set of bits and the second set of bits may be transmitted over a channel between a memory device and a host device in an interleaved pattern. The first code and the second code may also be transmitted over the channel. The first set of bits and the second set of bits may be deinterleaved by the receiving device. The first set of bits and the second set of bits may also be processed by the receiving device using the first code and the second code.

Abstract: Apparatuses and methods including multiple read modes for reading data from a memory are described. An example apparatus includes a memory including a first read mode and a second read mode. The memory has a read operation for the first read mode including a first pre-access phase, an access phase, and a first post-access phase. The read operation for the second read mode includes a second pre-access phase, the access phase, and a second post-access phase. The read operation for either the first read mode or the second read mode is performed responsive to the memory receiving a read command. The second pre-access phase is different from the first pre-access phase, with the second pre-access phase having a shorter time than the first pre-access phase measured from receipt of the read command.

Abstract: A controller of a memory device may determine that an endurance parameter associated with a wear leveling pool of a memory of the memory device satisfies a threshold. The wear leveling pool includes a plurality of memory blocks of the memory. The controller may divide, based on determining that the endurance parameter satisfies the threshold, the plurality of memory blocks of the wear leveling pool into a first wear leveling pool subset that includes a first subset of the plurality of memory blocks and a second wear leveling pool subset that includes a second subset of the plurality of memory blocks. A first subset of a plurality of data partitions is stored in the first subset of the plurality of memory blocks, and a second subset of the plurality of data partitions is stored in the second subset of the plurality of memory blocks.

Abstract: Methods, systems, and devices for selective access for grouped memory dies are described. A memory device may be configured with a select die access protocol for a group of memory dies that share a same channel. The protocol may be enabled by one or more commands from the host device, which may be communicated to each of the memory dies of the group via the channel. The command(s) may indicate a first set of one or more memory dies of the group for which a set of commands may be enabled and may also indicate a second set of one or more memory dies of the group for which at least a subset of the set of commands is disabled. When the select die access mode is enabled, the disabled memory dies may be restricted from performing the subset of commands received via the channel.

Abstract: A trigger rate associated with a scan operation of a set of memory pages of a data block is identified. The trigger rate is compared to a threshold rate to determine that a condition is satisfied. In response to satisfying the condition, a refresh operation is executed on the set of memory pages of the data block.

Abstract: Spiking events in a spiking neural network may be processed via a memory system. A memory system may store data corresponding to a group of destination neurons. The memory system may, at each time interval of a SNN, pass through data corresponding to a group of pre-synaptic spike events from respective source neurons. The data corresponding to the group of pre-synaptic spike events may be subsequently stored in the memory system.

Abstract: A memory sub-system configured to: measure a plurality of sets of signal and noise characteristics of a group of memory cells in a memory device; determine a plurality of optimized read voltages of the group of memory cells from the plurality of sets of signal and noise characteristics respectively; generate features from the plurality of sets of signal and noise characteristics, including at least one compound feature generated from the plurality of sets of signal and noise characteristics; generate, using the features, a classification of a bit error rate of data retrievable from the group of memory cells; and control an operation to read the group of memory cells based on the classification.

Abstract: An inspection system for determining wafer defects in semiconductor fabrication may include an image capturing device to capture a wafer image and a classification convolutional neural network (CNN) to determine a classification from a plurality of classes for the captured image. Each of the plurality of classes indicates a type of a defect in the wafer. The system may also include an encoder to encode to convert a training image into a feature vector; a cluster system to cluster the feature vector to generate soft labels for the training image; and a decoder to decode the feature vector into a re-generated image. The system may also include a classification system to determine a classification from the plurality of classes for the training image. The encoder and decoder may be formed from a CNN autoencoder. The classification CNN and the CNN autoencoder may each be a deep neural network.

Abstract: Some embodiments include an assembly having a stack of alternating dielectric levels and conductive levels. Channel material pillars extend through the stack. Some of the channel material pillars are associated with a first sub-block, and others of the channel material pillars are associated with a second sub-block. Memory cells are along the channel material pillars. An insulative level is over the stack. A select gate configuration is over the insulative level. The select gate configuration includes a first conductive gate structure associated with the first sub-block, and includes a second conductive gate structure associated with the second sub-block. The first and second conductive gate structures are laterally spaced from one another by an intervening insulative region. The first and second conductive gate structures have vertically-spaced conductive regions, and have vertically-extending conductive structures which electrically couple the vertically-spaced conductive regions to one another.

Abstract: A method of forming circuitry components includes forming a stack of horizontally extending and vertically overlapping features. The features extend horizontally though a primary portion of the stack with at least some of the features extending farther in the horizontal direction in an end portion. Operative structures are formed vertically through the features in the primary portion and dummy structures are formed vertically through the features in the end portion. Openings are formed through the features to form horizontally elongated and vertically overlapping lines from material of the features. The lines individually extend laterally about sides of vertically extending portions of both the operative structures and the dummy structures. Sacrificial material that is elevationally between the lines is at least partially removed in the primary and end portions laterally between the openings. Other aspects and implementations are disclosed.

Abstract: In a computer host system, a system and method to compress the transmission between the central processing unit (CPU) and the dynamic random access memory (DRAM) of either of an extended consecutive series of ‘0’ bits or an extended consecutive series of ‘1’ bits. The CPU or a Compute Express Link (CXL) Initiator associated with the CPU identifies the consecutive strings of ‘0’ bits or ‘1’ bits. The CPU or the CXL Initiator sets data flags in a FLIT data structure, using just two bits or four bits to indicate the strings. The data structure is sent to a CXL memory, which interprets the flags and constructs the extended series of ‘0’ bits or extended series of ‘1’ bits.