AI Super-Resolution for Gaming GPU Load Management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Modern mobile games require high frame rates and resolution, leading to increased GPU processing capability and power consumption, which can cause thermal issues and performance throttling, resulting in an unstable gaming experience due to excessive processor loading and heat generation.
Innovation Solution
A computing system that dynamically reduces GPU output resolution and selects an AI model based on graphics scenes and power consumption estimates to perform AI super-resolution operations, restoring the video resolution while managing power consumption and maintaining performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If GPU processing capability is increased to support high frame rates and resolution, then gaming performance is improved, but power consumption increases causing thermal issues and performance throttling
Solution Approach 1:
The system segments the processing workload by introducing a dedicated AI processing unit (APU) separate from the GPU. The GPU handles traditional graphics rendering while the APU specifically handles super-resolution tasks, dividing the overall processing function into specialized components that can operate more efficiently within power constraints.
Solution Approach 2:
An AI processing unit acts as an intermediary between the GPU and display output. The GPU renders at lower resolution, the APU performs super-resolution processing, and the final high-resolution output is displayed. This intermediary processing stage allows the system to achieve high visual quality without requiring the GPU to directly render at full resolution, reducing power consumption.
2Use of energy by moving object
If GPU resolution is reduced to lower power consumption, then power consumption decreases, but picture quality degrades
Solution Approach 1:
The APU serves as an intermediary that takes low-resolution output from the GPU and transforms it into high-resolution display output through AI-based super-resolution algorithms. This allows the system to maintain low power consumption during GPU rendering while achieving high picture quality at the display stage.
Solution Approach 2:
The system changes the resolution parameter at different stages of the processing pipeline. The GPU operates at a lower resolution parameter to reduce power consumption, while the APU transforms this to a higher resolution parameter for the final display, effectively decoupling the resolution requirement from the power consumption of the main graphics processor.
3Adaptability or versatility
If multiple AI models are maintained for different graphics scenes, then adaptability improves, but device complexity increases
Solution Approach 1:
The system dynamically selects and switches between different AI models based on the current graphics scene type. Rather than loading all models simultaneously, the system activates only the appropriate model for the current scene (e.g., different models for static images, video frames, or specific game types), reducing memory usage and system complexity while maintaining adaptability.
Solution Approach 2:
The system implements partial loading of AI models, maintaining only the most commonly used models in memory while allowing for on-demand loading of less frequently used models. This approach provides adaptability for different graphics scenes without requiring all models to be permanently resident, thereby reducing the active system complexity and memory requirements.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A computing system (100) performs artificial-intelligence, Al, super-resolution, SR. The computing system (100) includes multiple processors (110), which further includes a graphics processing unit, GPU, (112) and an AI processing unit, APU, (113). The computing system (100) also includes a memory (120) to store AI models (125). When detecting an indication that the loading of the GPU (112) exceeds a threshold, the processors (110) reduce the resolution of a video output from the GPU (112) in response to the indication. One of the AI models is selected based on graphics scenes in the video and the respective power consumption estimates of the AI models (125). The processors (110) then perform AI SR operations on the video using the selected AI model to restore the resolution of the video for display.