Thermal Management System for Processor Units
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current thermal management techniques for multi-core processors, such as the Cell Broadband Engine, often impact real-time guarantees and are inefficient, as they typically throttle all processor units in response to thermal limits, leading to costly cooling solutions and coarse power management.
Innovation Solution
A computer-implemented method and data processing system that determines the functionality of digital thermal sensors, enabling selective monitoring and disabling of power savings modes to maintain real-time guarantees while managing thermal conditions, allowing for dynamic throttling and interrupt generation to prevent overheating.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Temperature
If thermal management techniques throttle all processor units in response to thermal limits, then thermal conditions are managed, but real-time guarantees are impacted and processor performance is reduced
Solution Approach 1:
The patent segments the processor into multiple independent units (e.g., cores, threads, or execution units) and applies thermal management at the unit level rather than throttling the entire processor. Each unit can be independently monitored and throttled based on its local thermal conditions, allowing real-time critical units to maintain performance while thermal units are restricted.
Solution Approach 2:
The patent implements local thermal management by assigning different thermal thresholds and management strategies to different processor units based on their individual thermal conditions and real-time requirements. Units with real-time guarantees receive different thermal management treatment compared to non-real-time units, optimizing both thermal control and performance.
2Temperature
If coarse power management throttles much of the processor in reaction to thermal limits, then thermal limits are enforced, but processor productivity is significantly reduced
Solution Approach 1:
The patent divides the processor into fine-grained units that can be independently controlled. Instead of applying coarse throttling to the entire processor, the system applies fine-grained throttling to only those specific units exceeding thermal limits, preserving productivity of thermal units while enforcing thermal management.
Solution Approach 2:
The patent applies partial throttling rather than excessive (full) throttling. Only the portion of the processor that exceeds thermal thresholds is throttled, while the rest continues to operate at full performance. This partial action approach maintains productivity while still enforcing thermal limits.
3Temperature
If all processor units are throttled in response to thermal conditions, then thermal management is achieved, but the cost of cooling solutions increases
Solution Approach 1:
The patent segments thermal management into unit-level control, allowing the system to target cooling and thermal management resources only where needed. This reduces the overall cooling infrastructure requirements compared to designing for full-processor throttling, as only localized thermal control mechanisms are required.
4Reliability
If thermal management techniques are applied to maintain real-time guarantees, then real-time operation is ensured, but the complexity of thermal management increases
Solution Approach 1:
The patent segments the thermal management function into independent unit-level controllers that operate autonomously. Each unit has its own thermal sensor and control logic, simplifying the overall system architecture by distributing complexity rather than centralizing it. This modular approach makes real-time guarantees achievable through local decision-making.
Solution Approach 2:
The patent implements self-service thermal management where each processor unit autonomously monitors its own thermal conditions and applies appropriate throttling without requiring complex centralized control. This self-service approach reduces the complexity of the overall thermal management system while ensuring real-time guarantees for critical units.
Data Source
AI summary
A computer implemented method, data processing system, and processor are provided for managing a thermal management system. A determination is made as to whether a plurality of digital thermal sensors is faulty or functional. A power savings mode of at least one unit within the integrated circuit associated with the functional digital thermal sensor is monitored in response to at least one of the plurality of digital thermal sensors being functional. A functional digital thermal sensor is disabled in response to the at least one unit being in a power savings mode.