GPU Power State Control via Segmented Bus Interface
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing computer systems face inefficiencies in power usage due to the need to keep unused graphics processing units (GPUs) powered on to respond to configuration cycles, leading to increased power consumption and decreased battery life in mobile systems, especially when determining when a primary GPU can be shut down during idle periods without affecting system performance.
Innovation Solution
Implementing a power manager module that transitions GPUs to a substantially disabled state when idle, using a bus interface module to respond to configuration cycles and maintain power to essential components, allowing for selective power-down of execution units and memory while keeping the GPU connected to the system.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If GPUs are kept powered on to respond to configuration cycles, then system responsiveness is maintained, but power consumption increases
Solution Approach 1:
The GPU system is segmented into two functional parts: the execution units (which can be powered down) and the bus interface module (which remains powered on to respond to configuration cycles). This segmentation allows the system to maintain responsiveness while reducing power consumption by isolating the essential communication function from the power-intensive processing function.
Solution Approach 2:
The bus interface module acts as an intermediary between the CPU and the GPU execution units. It remains powered on to handle configuration cycles and communication, while the actual processing units can be placed in a low-power state. This intermediary approach allows the system to maintain responsiveness without keeping the entire GPU powered on.
2Productivity
If multiple GPUs are used for parallel processing, then processing performance increases, but power consumption increases
Solution Approach 1:
The system dynamically adjusts the power state of individual GPUs based on workload requirements. When parallel processing is needed, multiple GPUs are activated; when workloads can be handled by fewer GPUs, the others are placed in low-power states. This dynamic approach allows the system to optimize the balance between processing performance and power consumption.
Solution Approach 2:
The power state parameter of GPUs is changed based on system needs. GPUs can transition between full-power operational states and low-power states, allowing the system to adjust total power consumption while maintaining the ability to provide parallel processing when required by the workload.
3Use of energy by moving object
If primary GPU is shutdown during idle period, then power consumption decreases, but system reactivation complexity increases
Solution Approach 1:
The bus interface module performs preliminary actions by remaining powered on and maintaining communication capabilities even when the GPU execution units are powered down. This preliminary maintenance of the interface layer simplifies reactivation, as the communication pathway is already established and ready to receive work requests without requiring full system reinitialization.
Data Source
AI summary
A method of operating a processing device is provided. The method includes, responsive to an idle state of the processing device, transitioning the processing device to a substantially disabled state. The processing device, for example, may be a graphics processing unit (GPU). Transitioning the processing device to a substantially disabled state upon detection of an idle state may result in power savings. Corresponding systems and computer program products are also provided.