On-Demand Variable Memory Page Size for Energy Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing memory systems face limitations due to fixed page sizes, leading to unnecessary energy consumption and inefficiencies when handling varying amounts of data, as they often require activating entire pages even when only a small amount of data is needed, which affects performance and power consumption.
Innovation Solution
Implementing a memory system that supports on-demand or variable memory page sizes, allowing memory devices to adjust page sizes based on the specific requirements of data operations, using non-volatile memory technologies like FeRAM that can operate with multiple page sizes, and utilizing a bus protocol to specify page sizes with memory commands, thereby optimizing energy usage and performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If a fixed memory page size is used, then memory system simplicity is maintained, but energy consumption increases when handling varying data amounts
Solution Approach 1:
The patent implements dynamic page size adjustment by allowing the memory system to change page sizes based on workload characteristics. The memory controller monitors access patterns and dynamically selects appropriate page sizes from multiple supported sizes, enabling the system to adapt to varying data amounts and access frequencies, thereby reducing energy consumption for both small and large data operations.
Solution Approach 2:
The patent changes the parameter of page size from fixed to variable by supporting multiple page sizes (e.g., 512 bytes, 1 KB, 2 KB, 4 KB). The system selects the appropriate page size parameter based on the specific workload requirements, allowing optimization of energy consumption by matching page size to the actual data access needs rather than using a single fixed size for all operations.
2Loss of energy
If a small fixed page size is used, then energy consumption for small data operations is reduced, but performance degrades when handling large data amounts
Solution Approach 1:
The memory system dynamically adjusts page size based on the size of data being accessed. For small data operations, smaller page sizes are used to reduce activation overhead and energy consumption. For large data operations, larger page sizes are selected to improve throughput and reduce the number of page activations required, thereby maintaining high productivity across varying workload sizes.
Solution Approach 2:
The system changes the page size parameter adaptively based on workload characteristics. When small data amounts are detected, the page size parameter is reduced to minimize energy consumption. When large data amounts are detected, the page size parameter is increased to maximize throughput and operational speed, thus resolving the contradiction between energy efficiency and productivity.
3Productivity
If a large fixed page size is used, then performance for large data operations is improved, but energy consumption increases for small data operations
Solution Approach 1:
The memory controller implements dynamic page size selection that responds to workload demands. For large data operations, larger page sizes are activated to improve throughput and operational speed. For small data operations, the system switches to smaller page sizes to reduce the energy overhead of activating and maintaining large pages, thus optimizing the balance between productivity and energy consumption based on real-time needs.
Solution Approach 2:
The page size parameter is changed dynamically based on the detected workload size. When large data operations are required, the system increases the page size parameter to enhance productivity. When small data operations are detected, the system decreases the page size parameter to reduce energy consumption, effectively resolving the contradiction between maintaining high performance and minimizing energy waste.
4Loss of energy
If variable page sizes are supported, then energy efficiency and flexibility are improved, but device complexity increases
Solution Approach 1:
The patent implements variable page size support through dynamic configuration in the memory controller. The controller includes logic to detect workload characteristics and select appropriate page sizes from a set of supported sizes. This dynamic capability improves energy efficiency by matching page size to actual needs, while the complexity is managed through automated detection and selection algorithms that transparently handle the variable configurations.
Solution Approach 2:
The memory system supports multiple page size parameters (e.g., 512 bytes, 1 KB, 2 KB, 4 KB) that can be changed based on workload requirements. The increased complexity of supporting multiple parameters is offset by the energy efficiency gains, as the system automatically selects the most appropriate parameter for each operation, reducing overall energy consumption despite the added configurational complexity.
Data Source
AI summary
Systems, devices, and methods related to on demand memory page size are described. A memory system may employ a protocol that supports on demand variable memory page sizes. A memory system may include one or more non-volatile memory devices that may each include a local memory controller configured to support variable memory page size operation. The memory system may include a system memory controller that interfaces between the non-volatile memory devices and a processor. The system memory controller may, for instance, use a protocol that facilitates on demand memory page size where a determination of a particular page size to use in an operation may be based on characteristics of memory commands and data involved in the memory command.