Switchable Graphics Management Scheme for GPU Selection
Find Innovative SolutionsGenerate 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
Engineering 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
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.
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.
2Productivity
If dGPU is used for rendering tasks, then performance is improved, but power consumption and thermal load increase
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.
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.
3Power
If iGPU is used for rendering tasks, then power consumption is reduced, but rendering performance decreases
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.
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.
4Speed
If GPU selection is based on pre-defined application categories, then decision speed is improved, but adaptability to different system configurations is reduced
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.
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.
Data Source
Figure 1
Figure 2
Figure 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.