GPU Power Management via Bus Interface Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

In multiprocessing systems, particularly in graphics processing units (GPUs), there is a need to reduce power consumption without interfering with system behavior or bus configuration cycles, as keeping unused GPUs powered can lead to inefficient power usage and decreased battery life in mobile systems.

Innovation Solution

A method and system where a bus interface module remains powered to respond to configuration cycles, allowing GPUs to be transitioned to a substantially disabled state while still simulating a powered state, thereby avoiding complex operating system processes that can negatively impact user experience.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If GPUs are kept powered to respond to configuration cycles, then system reliability is improved, but power consumption increases

Engineering Contradiction:
Improvesystem behaviorVSAvoidpower consumption
Core Design Contradiction:
ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent segments the GPU into two distinct parts: a bus interface module that remains powered and responsive to configuration cycles, and a processing core that can be completely powered down. This segmentation allows the system to maintain reliability for bus communication while eliminating unnecessary power consumption from the processing core when not in use.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extracts the bus interface module from the main GPU processing core, allowing it to operate independently as a powered state module. This extracted interface module can respond to configuration cycles without requiring the main processing core to remain powered, thereby resolving the contradiction between system reliability and power consumption.

Inventive Principle:
Principle #2Taking out (Extraction)

2Use of energy by moving object

If GPUs are powered down to save power, then power consumption is reduced, but system reliability deteriorates due to complex OS processes

Engineering Contradiction:
Improvepower consumptionVSAvoidsystem behavior
Core Design Contradiction:
Use of energy by moving objectVSReliability

Solution Approach 1:

The bus interface module performs preliminary action by remaining in a powered state and pre-positioned to respond to configuration cycles before the main GPU processing core is fully powered down. This ensures that when the OS needs to access the GPU, the interface is already ready to handle the communication, avoiding complex recovery processes.

Inventive Principle:
Principle #10Preliminary action

3Use of energy by moving object

If all GPU components are powered down, then power consumption is minimized, but device complexity increases due to state management requirements

Engineering Contradiction:
Improvepower consumptionVSAvoidpower state management
Core Design Contradiction:
Use of energy by moving objectVSDevice complexity

Solution Approach 1:

The patent implements dynamic power state management where the bus interface module can independently transition between powered and powered-down states based on system needs. This dynamic approach allows the system to optimize power consumption while maintaining manageable complexity through standardized state transitions and control mechanisms.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP2476038B1Method and apparatus for disabling a device
Publication Date: 2018.07.04 ADVANCED MICRO DEVICES INC
  • EP2476038B1 patent drawingFigure 1
  • EP2476038B1 patent drawingFigure 2
  • EP2476038B1 patent drawingFigure 3

AI summary

A method of operating a device is provided. The method includes transitioning the GPU to a substantially disabled state in response to a first received signal, and generating, while the GPU is in the substantially disabled state, a response signal in response to a second received signal. The response signal is substantially similar to a second response signal that would be generated by the GPU in a powered state in response to the second received signal.