Memory Device Sub-Bank Metadata Storage Architecture
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Memory devices face challenges in efficiently assigning resources for metadata storage, leading to increased area requirements and potential row hammer effects due to repeated access to specific wordlines.
Innovation Solution
The implementation of a memory device with at least one bank comprising a first sub-bank for normal data and a second sub-bank for metadata, where metadata corresponding to each first wordline is stored in each second wordline, allowing for independent operations and reduced resource allocation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If metadata is stored in the same bank as normal data, then area usage is reduced, but row hammer effect increases due to repeated access to specific wordlines
Solution Approach 1:
The memory device is segmented into multiple sub-banks (first sub-bank for normal data, second sub-bank for metadata) within the same bank. This segmentation allows metadata and normal data to be stored in separate physical locations while maintaining logical association, thereby reducing the row hammer effect caused by repeated metadata access while keeping the overall area compact.
Solution Approach 2:
A corresponding relationship is established between wordlines in the first sub-bank and wordlines in the second sub-bank, where metadata for normal data corresponding to each first wordline is stored in each second wordline respectively corresponding to the first wordlines. This intermediary mapping structure enables efficient metadata access without directly impacting the normal data wordlines, mitigating the row hammer effect.
2Object-affected harmful factors
If metadata is stored in separate sub-banks, then row hammer effect is reduced, but device complexity increases
Solution Approach 1:
The second sub-bank serves multiple functions: it stores metadata corresponding to normal data in the first sub-bank, and the wordlines in the second sub-bank are respectively corresponding to wordlines in the first sub-bank, enabling efficient address mapping and data retrieval. This multi-functionality reduces the need for additional complex structures.
Solution Approach 2:
The patent changes the organizational parameter of memory storage by dividing a bank into sub-banks with specific functions. The first sub-bank stores normal data while the second sub-bank stores metadata, with wordlines established in a respectively corresponding relationship. This parameter change optimizes the balance between complexity and performance.
3Productivity
If metadata access is performed frequently, then operational efficiency is improved, but timing losses increase due to row hammer effect
Solution Approach 1:
By segmenting the bank into sub-banks for normal data and metadata, the patent enables independent access to metadata without affecting normal data operations. This segmentation reduces timing losses associated with row hammer effects while maintaining high operational efficiency for frequent metadata access.
Solution Approach 2:
The respectively corresponding wordline structure acts as an intermediary that enables efficient metadata access through the second sub-bank without directly accessing the wordlines containing normal data in the first sub-bank. This intermediary mechanism reduces timing losses by avoiding the row hammer effect while maintaining operational efficiency.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A memory device (200) includes at least one bank including a first sub-bank (SB 1_1) and a second sub-bank (SB2_1) disposed in a wordline direction. The first sub-bank (SB 1_1) may store normal data and may be connected to a plurality of first wordlines (WL1_0 to WL1_n), the second sub-bank (SB2_1) may store metadata corresponding to the normal data and may be connected to a plurality of second wordlines (WL2_0 to WL2_n), and metadata for normal data corresponding to each of the first wordlines (WL1_0 to WL1_n) may be stored in each of plurality of second wordlines (WL2_0 to WL2_n), respectively corresponding to the plurality of first wordlines (WL1_0 to WL1_n).