GPU Frequency Selection Using Throughput-Based Dynamic Scaling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Limiting user programs to a single design frequency of a GPU results in underutilization of computing resources, as traditional methods cap frequencies at a worst-case scenario or a few fixed values, failing to optimize performance for individual applications.
Innovation Solution
A system that dynamically adjusts the operating frequency of GPUs by comparing sensor data from benchmark and user programs, using critical path monitors to determine the maximum throughput and customize frequencies for specific applications, thereby optimizing performance and power efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single design frequency is used for all user programs, then device complexity is reduced and ease of operation is improved, but processing resources are underutilized and productivity decreases
Solution Approach 1:
The system dynamically adjusts the operating frequency of the GPU based on the specific requirements of each user program. Instead of using a fixed design frequency for all applications, the frequency determination module continuously monitors program characteristics and adjusts the frequency in real-time to match the computational demands of each program, thereby improving processing resource utilization without requiring complex manual configuration
Solution Approach 2:
The patent changes the frequency parameter adaptively for different user programs. The frequency determination module analyzes program characteristics and modifies the operating frequency parameter accordingly, allowing each program to run at its optimal frequency rather than being constrained to a single design frequency, thus resolving the contradiction between resource utilization and operational simplicity
2Productivity
If design frequency is increased for better performance, then productivity is improved, but power consumption increases
Solution Approach 1:
The system implements dynamic frequency scaling where the GPU operates at higher frequencies only when necessary for performance-critical tasks. The frequency determination module continuously adjusts the operating frequency based on program requirements, allowing the system to achieve high productivity when needed while consuming less power during routine operations, thus resolving the performance-power consumption trade-off
Solution Approach 2:
The patent dynamically changes the frequency parameter based on program characteristics and performance requirements. By adjusting the frequency parameter adaptively rather than maintaining a constantly high frequency, the system achieves high productivity when needed while minimizing power consumption during normal operations, effectively resolving the contradiction between performance and energy usage
3Adaptability or versatility
If fixed design frequencies are used, then device complexity is reduced, but adaptability to different applications decreases
Solution Approach 1:
The frequency determination module automatically determines the appropriate operating frequency for each user program without requiring manual intervention or complex user configuration. The system self-adjusts the frequency based on program characteristics, providing application-specific optimization while maintaining ease of operation. This self-service approach resolves the contradiction between adaptability and device complexity by automating the frequency selection process
Data Source
AI summary
Apparatuses, systems, and techniques to select frequency of a processing unit. In at least one embodiment, an operating frequency of one or more integrated circuits is dynamically adjusted based, at least in part, on a dynamically measured maximum throughput of the one or more integrated circuits.