DRAM Rank Refresh Distribution for Asymmetrical Memory Layouts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional DRAM memory ranks in portable computing devices are refreshed inefficiently, leading to power consumption issues and asymmetrical size distribution problems, which affect power conservation and refresh efficiency.
Innovation Solution
A method and system that interleaves Partial Array Self-refresh (PASR) requests among asymmetrical memory ranks based on their sizes, using a resource power manager to determine the most efficient distribution method, which can be linear, interleave, or a combination of both, to optimize power consumption and refresh efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If PASR requests are processed sequentially in linear manner across memory ranks, then implementation simplicity is maintained, but power consumption increases and refresh efficiency decreases
Solution Approach 1:
The patent segments the memory refresh operation by dividing PASR requests into multiple parallel sub-requests that can be distributed across different memory ranks simultaneously. The memory controller breaks down a single large PASR request into smaller chunks that are processed in parallel across available memory ranks, reducing the total time and power required for refresh operations.
Solution Approach 2:
The patent transitions from sequential (one-dimensional) processing to parallel (multi-dimensional) processing by utilizing multiple memory ranks simultaneously. Instead of processing PASR requests in a single linear sequence, the system distributes requests across multiple ranks in parallel, effectively adding a temporal dimension to the refresh operation and improving overall efficiency.
2Productivity
If PASR requests are distributed across asymmetrical memory ranks, then refresh efficiency improves, but distribution complexity increases
Solution Approach 1:
The patent implements dynamic adaptation to asymmetrical memory rank configurations. The memory controller automatically detects the size differences between memory ranks and dynamically adjusts the distribution strategy for PASR requests. This dynamic approach allows efficient utilization of available memory capacity while simplifying the distribution logic compared to static allocation methods.
Solution Approach 2:
The patent changes the distribution parameters based on the specific characteristics of asymmetrical memory ranks. By adjusting request size, distribution pattern, and allocation strategy according to the actual memory configuration, the system achieves high refresh efficiency without requiring complex custom distribution logic for each asymmetrical scenario.
3Loss of energy
If multiple PASR requests are processed in parallel, then power consumption is reduced through efficient refresh, but request coordination complexity increases
Solution Approach 1:
The patent enables memory ranks to serve themselves through self-refresh capabilities while the controller coordinates parallel operations. Each memory rank can autonomously manage its own refresh operations when placed in self-refresh mode, reducing the coordination burden on the controller. The system leverages the inherent self-service capability of DRAM ranks to simplify parallel request management.
Solution Approach 2:
The patent utilizes periodic self-refresh operations inherent to DRAM technology to maintain memory data without requiring continuous external intervention. By scheduling PASR requests to align with the natural refresh periodicity of memory ranks and utilizing self-refresh modes, the system reduces power consumption while minimizing the complexity of request coordination through regular, predictable operation patterns.
Data Source
Figure 1~2A
Figure 2B~3D
Figure 3E
AI summary
A kernel of an HLOS may originate one or more memory refresh requests. Each memory refresh request may have a first memory address range and a size value. A resource power manager may be coupled to the kernel and coupled to memory. The memory may have a plurality of memory ranks. The resource power manger may receive a memory refresh request from the kernel. The resource power manager may then determine if the plurality of memory ranks is either symmetrical or asymmetrical. If the memory ranks are symmetrical, then the resource power manager distributes the memory refresh request evenly and in a parallel manner across the symmetrical memory ranks. If the memory ranks are asymmetrical, then the resource power manager will then determine if the memory refresh request should be one of: a linear only memory refresh; an interleave with linear memory refresh; or an interleave only memory refresh.