On-Die Data Transfer in Memory Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional memory devices require external data lines for data consolidation, reconsolidation, and manipulation, leading to power-intensive and bandwidth-consuming processes due to the need to read and rewrite data across the memory array, which can result in data fragmentation and loss.
Innovation Solution
The implementation of memory devices with internal data transfer capabilities, utilizing local and global caches, data read/write lines, and tracking mechanisms to move data within the device without relying on external IO or DQ data lines, allowing for flexible consolidation, reconsolidation, and rearrangement of data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If external data lines are used for data consolidation and manipulation, then data can be transferred between memory locations, but power consumption increases and IO bandwidth is consumed
Solution Approach 1:
The patent extracts the data transfer function from external IO paths and relocates it to internal device resources. By using internal caches and on-die transfer mechanisms, the system removes the need for external data line transactions for consolidation operations, thereby reducing power consumption and IO bandwidth usage while maintaining data transfer capability
Solution Approach 2:
The patent introduces internal caches as intermediary storage structures between the memory array and external IO interfaces. These caches serve as mediators that enable data consolidation and manipulation operations to occur internally without requiring external data line transactions, thus resolving the contradiction between transfer capability and power/bandwidth consumption
2Productivity
If data is read and rewritten across the memory array using external lines, then data consolidation can be performed, but IO bandwidth is consumed and data fragmentation may occur
Solution Approach 1:
The patent segments the memory system into distinct functional regions including memory arrays, internal caches, and control logic. This segmentation allows data consolidation operations to be performed within specific internal regions without affecting the entire memory array, thereby preventing data fragmentation while maintaining consolidation productivity
Solution Approach 2:
Internal caches act as intermediary structures that facilitate data consolidation without requiring external read-rewrite cycles. By transferring data between internal cache regions rather than through external IO lines, the system achieves efficient consolidation while preventing data fragmentation that would occur with external line transactions
3Loss of energy
If internal data transfer mechanisms are implemented, then power consumption and IO bandwidth usage are reduced, but device complexity increases
Solution Approach 1:
The patent implements multi-functional internal caches that serve multiple purposes: they act as temporary storage during read/write operations, enable data consolidation without external IO, provide buffering for timing synchronization, and support on-die data transfer. This universality reduces the need for separate dedicated structures, thereby limiting the increase in device complexity while achieving energy savings
Data Source
AI summary
Memory devices and systems with on-die data transfer capability, and associated methods, are disclosed herein. In one embodiment, a memory device includes an array of memory cells and a plurality of input/output lines operably connecting the array to data pads of the device. In some embodiments, the memory device can further include a global cache and/or a local cache. The memory device can be configured to internally transfer data stored at a first location in the array to a second location in the array without outputting the data from the memory device. To transfer the data, the memory device can copy data on one row of memory cells to another row of memory cells, directly write data to the second location from the first location using data read/write lines of the input/output lines, and/or read the data into and out of the global cache and/or the local cache.