Entropy-Based Memory Data Placement for Power and Thermal Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing memory technologies face inefficiencies in power consumption and thermal management due to the varying entropy levels of data being stored, particularly in DRAM devices, leading to increased heat generation and reduced performance.
Innovation Solution
The solution involves storing high and low entropy data portions in different memory devices based on temperature indicators, disabling unnecessary circuitry for low entropy data operations to reduce power consumption and thermal stress.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If data is stored in memory devices without entropy-based optimization, then memory devices must process all data uniformly, but this causes increased power consumption and thermal stress especially for high entropy data
Solution Approach 1:
The patent segments data into high entropy and low entropy portions based on compression ratios, and stores them in different memory devices. This segmentation allows the system to apply different processing strategies to different data types, reducing overall power consumption while maintaining necessary processing capabilities for high entropy data.
Solution Approach 2:
The patent implements local quality by enabling memory devices to operate in different modes based on the entropy characteristics of the data they store. Memory devices can switch between full processing mode for high entropy data and reduced processing mode for low entropy data, optimizing power consumption locally without compromising system functionality.
2Temperature
If memory devices process all data uniformly, then data access is simplified, but thermal conditions deteriorate due to continuous high power consumption
Solution Approach 1:
The patent introduces dynamic operation modes for memory devices that adjust their processing behavior based on data entropy characteristics. Memory devices can dynamically switch between full processing mode and power-saving mode, allowing the system to maintain simple data access for low entropy data while reducing thermal stress during operation.
Solution Approach 2:
The patent changes operational parameters of memory devices based on the entropy characteristics of stored data. By adjusting processing parameters such as data processing depth, error checking intensity, and access speed based on entropy levels, the system optimizes thermal conditions without significantly complicating data access operations.
3Quantity of substance
If high entropy data is stored in all memory devices, then data distribution is uniform, but power consumption and heat generation increase
Solution Approach 1:
The patent segments data into high entropy and low entropy portions and distributes them to different memory devices based on their operational characteristics. This segmentation enables the system to reduce energy loss by storing low entropy data in memory devices with reduced processing capabilities, while maintaining uniform data distribution across the memory subsystem.
Solution Approach 2:
The patent uses copying mechanisms to replicate data across multiple memory devices with different operational modes. By copying data to appropriate memory devices based on entropy characteristics, the system maintains data availability and distribution uniformity while reducing overall energy loss through selective processing.
Data Source
AI summary
A fixed size block of data may be compressed, but not reduced in size, resulting in a high entropy (i.e., information carrying) portion and a low entropy (i.e., little or no information carrying—e.g., all zeros) portion. The high entropy portion and the low entropy portion of the compressed block may be stored by a controller in different memory devices of a memory module. The selection of the memory devices on the memory module to store the high entropy portion versus the low entropy portion may be based on temperature indicators associated with the memory devices. When reading or writing data to the module, entropy indicators (e.g., high or low entropy) are communicated and stored on a per memory device basis.