Memory Read-Threshold Prediction for Lower Read-Retry Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Solid state memory storage with advanced multi-level cell techniques experiences significant performance degradation due to a large number of read-retry operations, which increase read latency and fail to meet stringent quality-of-service (QoS) requirements.
Innovation Solution
Utilizing a deep neural network (DNN) to track read voltage thresholds without additional reads, leveraging a large number of host reads to generate an updated read threshold set, thereby improving memory device performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If read-retry operations are performed to ensure reliability, then data accuracy is improved, but read latency increases significantly
Solution Approach 1:
The system performs preliminary tracking of read voltage thresholds using DNN during normal host reads, so that when read-retry operations are needed, the voltage thresholds are already updated and ready, eliminating the need for additional calibration reads and reducing latency
Solution Approach 2:
The system uses the existing host reads to simultaneously serve dual purposes: data retrieval and voltage threshold tracking, eliminating the need for separate calibration operations and reducing overall read latency while maintaining reliability
2Measurement precision
If additional reads are performed to track read voltage thresholds, then threshold accuracy is improved, but read latency increases
Solution Approach 1:
The host reads are made multi-functional by using them for both data retrieval and voltage threshold tracking simultaneously, eliminating the need for additional dedicated calibration reads and thus reducing latency while maintaining threshold accuracy
Solution Approach 2:
The existing host reads are leveraged to self-update the voltage thresholds through DNN processing, eliminating the need for separate calibration operations and reducing overall read latency while maintaining threshold accuracy
3Measurement precision
If a large number of host reads are used to train DNN, then threshold prediction accuracy is improved, but data processing complexity increases
Solution Approach 1:
The system extracts only the necessary voltage threshold parameters from the large set of host reads, filtering out redundant information and focusing DNN training on the most relevant data points, thereby maintaining high prediction accuracy while reducing processing complexity
Data Source
AI summary
Devices, systems, and methods for improving performance of a memory device are described. An example method includes extracting parameters, which include a read threshold set, from each of a first set of host reads, replacing, based on the parameters, at least one host read from a second set of host reads by at least one host read from the first set of host reads, using a deep neural network (DNN) to generate an updated read threshold set, wherein an input to the DNN comprises the parameters from each of the second set of host reads subsequent to the replacing, and applying the updated read threshold set to the memory device to retrieve information from the memory device. In an example, the number of the first set of host reads is at least two orders of magnitude greater than the number of the second set of host reads.