Adaptive Power Throttling via Sensitivity-Based Unit Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In computing systems, efficiently allocating power among components to optimize performance is challenging, as excessive power allocation to one component can lead to performance stalls and thermal issues, while insufficient allocation can result in delayed memory servicing, making it difficult to determine which components to reduce power for without impacting system performance.
Innovation Solution
A system management unit detects power changes and uses sensitivity data to identify units with lower or higher sensitivity to power state changes, adjusting power allocation accordingly to minimize performance impact, and updates the data based on actual performance comparisons.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If power is allocated to processors to improve execution speed, then processing performance is improved, but power consumption increases and may cause thermal issues
Solution Approach 1:
The system dynamically adjusts power allocation to processors based on real-time performance monitoring. The power management unit continuously monitors performance metrics and modifies power states of processors and memory subsystem accordingly, enabling the system to adapt power consumption to actual processing needs rather than using static allocation.
Solution Approach 2:
The system implements a feedback mechanism where performance is monitored and used to adjust power allocation. The power management unit receives performance data, compares it against targets, and adjusts power states of processing units and memory subsystem to optimize both performance and power consumption, creating a closed-loop control system.
2Speed
If power is allocated to memory subsystem to improve data servicing speed, then memory access performance is improved, but power consumption increases and processors may not have adequate power budget
Solution Approach 1:
The system dynamically adjusts power allocation to the memory subsystem based on real-time monitoring of memory access patterns and system performance. The power management unit can increase memory power state during high-demand periods and reduce it during low-demand periods, optimizing the balance between memory access speed and overall power consumption.
Solution Approach 2:
The system changes operational parameters (power states) of the memory subsystem dynamically. Different power states are applied to the memory subsystem based on workload characteristics, allowing the system to optimize memory performance when needed while reducing power consumption during normal operation.
3Use of energy by moving object
If power is reduced from components to meet power limits, then power consumption is reduced, but system performance may be degraded
Solution Approach 1:
The system uses feedback from performance monitoring to guide power reduction decisions. When power limits are approached, the power management unit monitors performance metrics and selectively reduces power to components whose degradation would have minimal impact on overall system performance, maintaining productivity while meeting power constraints.
Solution Approach 2:
The system dynamically changes power parameters of different components based on their current workload and importance to system performance. Rather than uniform power reduction, the system adjusts power states of specific components (processors, memory, I/O) based on real-time conditions, minimizing performance impact while achieving power savings.
Data Source
AI summary
Systems, apparatuses, and methods for managing power allocation in a computing system. A system management unit detects a condition indicating a change in power is indicated. Such a change may be detecting an indication that a power change is either required, possible, or requested. In response to detecting a reduction in power is indicated, the system management unit identifies currently executing tasks of the computing system and accesses sensitivity data to determine which of a number of computing units (or power domains) to select for power reduction. Based at least in part on the data, a unit is identified that is determined to have a relatively low sensitivity to power state changes under the current operating conditions. A relatively low sensitivity indicates that a change in power to the corresponding unit will not have as significant an impact on overall performance of the computing system than if another unit was selected. Power allocated for the selected unit is then decreased.