Switchable Graphics Management Scheme for GPU Selection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing switchable graphics systems lack the ability to dynamically select between integrated graphics (iGPU) and discrete graphics (dGPU) based on real-time system conditions such as thermal and power budgets, leading to inefficient performance and energy consumption.

Innovation Solution

A new switchable graphics management scheme that utilizes performance/watt information of both iGPU and dGPU, along with real-time system resources like SoC thermal and power budgets, to dynamically decide on the right GPU for rendering tasks, thereby optimizing performance and energy consumption.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Extent of automation

If drivers/OS decide on iGPU or dGPU for task execution, then GPU selection is automated, but the decision lacks information about performance per watt capabilities and system resources leading to inefficient energy consumption

Engineering Contradiction:
ImproveGPU selection automationVSAvoidenergy consumption
Core Design Contradiction:
Extent of automationVSUse of energy by moving object

Solution Approach 1:

The system implements feedback by continuously monitoring real-time system conditions (thermal budget, power budget, memory configuration) and performance per watt capabilities, then using this information to dynamically adjust GPU selection decisions. The power management algorithm receives feedback from system sensors and adjusts its decisions accordingly, ensuring optimal energy consumption while maintaining automated operation.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The power management algorithm serves itself by autonomously making GPU selection decisions based on pre-characterized performance per watt data and real-time system conditions. The system self-adjusts without requiring external intervention, automatically switching between iGPU and dGPU based on current thermal and power budget availability, thus resolving the contradiction between automation and energy efficiency.

Inventive Principle:
Principle #25Self-service

2Productivity

If dGPU is used for rendering tasks, then performance is improved, but power consumption and thermal load increase

Engineering Contradiction:
Improverendering performanceVSAvoidpower consumption
Core Design Contradiction:
ProductivityVSPower

Solution Approach 1:

The system dynamically switches between iGPU and dGPU based on real-time thermal and power budget conditions. When thermal headroom is available and power budget permits, the system transitions to dGPU for high-performance rendering. When thermal or power budgets are constrained, it dynamically switches to iGPU, creating a flexible, adaptive system that resolves the contradiction between performance and power consumption.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The power management algorithm changes operational parameters (GPU selection, rendering task allocation) based on varying system conditions. By monitoring thermal budget, power budget, and performance per watt characteristics, the system adjusts which GPU handles rendering tasks, enabling it to optimize the balance between performance output and power consumption at different operating points.

Inventive Principle:
Principle #35Parameter changes

3Power

If iGPU is used for rendering tasks, then power consumption is reduced, but rendering performance decreases

Engineering Contradiction:
Improvepower consumptionVSAvoidrendering performance
Core Design Contradiction:
PowerVSProductivity

Solution Approach 1:

The system dynamically determines when to use iGPU versus dGPU based on real-time conditions. During low-demand periods or when thermal/power budgets are constrained, the system dynamically switches to iGPU for power-efficient operation. During high-demand periods with sufficient thermal headroom, it dynamically transitions to dGPU, thus resolving the contradiction between power savings and performance requirements through time-varying operation modes.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system performs preliminary characterization of performance per watt capabilities for both iGPU and dGPU across various workloads before runtime decision-making. This pre-characterization data is stored and used by the power management algorithm to make informed decisions about when to use each GPU, allowing the system to anticipate performance needs while managing power consumption effectively.

Inventive Principle:
Principle #10Preliminary action

4Speed

If GPU selection is based on pre-defined application categories, then decision speed is improved, but adaptability to different system configurations is reduced

Engineering Contradiction:
Improvedecision speedVSAvoidsystem configuration adaptability
Core Design Contradiction:
SpeedVSAdaptability or versatility

Solution Approach 1:

The system changes its decision parameters from static pre-defined categories to dynamic parameters based on actual system configuration (memory configuration, thermal budget, power budget). The power management algorithm adapts its decision-making criteria to match the specific system setup, resolving the contradiction between fast decision-making and configuration adaptability by using flexible, condition-based parameters.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The power management algorithm autonomously adapts to different system configurations by reading actual system parameters (thermal budget, power budget, memory configuration) and adjusting its GPU selection strategy accordingly. This self-adaptation capability allows the system to maintain fast automated decisions while being versatile across different hardware configurations without requiring manual reconfiguration.

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP4018286B1Apparatus and method to improve switchable graphics system performance and energy consumption based applications and real-time system power/thermal budgets
Publication Date: 2025.06.18 INTEL CORP
  • EP4018286B1 patent drawingFigure 1
  • EP4018286B1 patent drawingFigure 2
  • EP4018286B1 patent drawingFigure 3

AI summary

A switchable graphics management scheme, which uses performance/watt information of both the iGPU/dGPU along with system real-time resources like SoC (system-on-chip) thermal, system power budgets to decide on the right GPU for rendering tasks. The scheme uses this threshold power point information along with system resources to determine the optimized GPU for tasks rendering for all applications and use cases. As such, the scheme of adapts to each system design based on capabilities of that specific system.