Memory Operating Mode Register for AI Power Reduction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing memory devices experience high power consumption when accessed by artificial intelligence systems due to simultaneous access of multiple components, leading to increased energy usage.

Innovation Solution

Modifying the operating mode of memory devices to a reduced power consumption mode by writing a value indicative of the mode to a register, which reduces power consumption and allows only commands to switch to a normal operating mode, thereby conserving energy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Speed

If memory devices operate in normal operating mode to support AI system access, then data access speed and system functionality are improved, but power consumption increases

Engineering Contradiction:
Improvedata access speedVSAvoidpower consumption
Core Design Contradiction:
SpeedVSUse of energy by moving object

Solution Approach 1:

The memory device dynamically switches between normal operating mode and reduced power consumption mode based on operational requirements. The controller monitors access patterns and transitions the memory device to appropriate modes, making the power consumption adaptive rather than static. This resolves the contradiction by allowing high speed when needed while conserving power during idle or low-demand periods.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent changes operational parameters of the memory device by writing specific values to mode registers. When transitioning to reduced power mode, parameters such as refresh rates, voltage levels, and operational frequency are adjusted to lower consumption states. This parameter modification enables the memory device to operate efficiently at different power levels depending on system needs.

Inventive Principle:
Principle #35Parameter changes

2Reliability

If memory devices perform refresh operations continuously, then data integrity is maintained, but power consumption increases

Engineering Contradiction:
Improvedata integrityVSAvoidpower consumption
Core Design Contradiction:
ReliabilityVSUse of energy by moving object

Solution Approach 1:

Instead of continuous refresh operations, the patent implements periodic refresh cycles when the memory device is in reduced power mode. Refresh operations are performed at extended intervals rather than continuously, reducing power consumption while still maintaining data integrity over acceptable timeframes. This periodic approach resolves the contradiction between reliability and power usage.

Inventive Principle:
Principle #19Periodic action

Solution Approach 2:

In reduced power mode, the patent performs partial refresh operations on only certain memory portions that are actively used or recently accessed, rather than refreshing the entire memory array. This partial action maintains data integrity for critical data while significantly reducing the power consumption associated with full-refresh operations.

Inventive Principle:
Principle #16Partial or excessive action

3Productivity

If multiple memory components are accessed simultaneously by AI systems, then processing throughput is improved, but power consumption increases

Engineering Contradiction:
Improveprocessing throughputVSAvoidpower consumption
Core Design Contradiction:
ProductivityVSUse of energy by moving object

Solution Approach 1:

The patent segments the memory device into multiple banks or regions that can be independently controlled. When AI systems need simultaneous access, only the necessary segments are activated in normal mode while other segments remain in reduced power mode. This segmentation allows selective activation, maintaining processing throughput for active segments while conserving power across the entire memory device.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The controller implements multi-functionality by managing both high-performance access and power conservation within the same memory system. It can dynamically allocate memory regions between normal and reduced power modes based on AI workload requirements, making the system adaptable to varying throughput needs while optimizing power consumption across different operational scenarios.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentUS20250238066A1Operating mode register
Publication Date: 2025.07.24 LODESTAR LICENSING GROUP LLC
  • US20250238066A1 patent drawing
  • US20250238066A1 patent drawing
  • US20250238066A1 patent drawing

AI summary

The present disclosure includes apparatuses and methods related to modifying an operating mode in memory. An example apparatus can include a memory array and a controller coupled to the memory array, wherein the controller includes a register configured to receive a mode register write command and write a value indicative of an operating mode in which the apparatus has reduced power consumption relative to a normal operating mode.