Programmable Impingement Cooling for Multi-Chip Die Temperature Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Multi-chip packages (MCPs) face challenges in efficient cooling due to varying thermal properties of individual dies, leading to suboptimal operating temperatures and increased power consumption, as existing cooling systems treat all dies uniformly.
Innovation Solution
A dynamic cooling system that monitors and controls the junction temperature of each die in an MCP by adjusting micro nozzle valves to direct fluid flow, allowing each die to operate at its optimal temperature, reducing leakage power and improving performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If a single heat spreader/heat sink is used to cool all dies in an MCP, then the cooling system is simple and easy to manufacture, but each die cannot operate at its optimal temperature leading to increased power consumption and reduced performance
Solution Approach 1:
The cooling system is segmented into multiple independent cooling channels, each dedicated to a specific die. Each channel includes its own heat spreader portion and controllable fluid flow path, allowing individual temperature optimization for each die while maintaining overall system manufacturability through modular construction
Solution Approach 2:
The cooling system incorporates dynamically adjustable fluid flow control for each die, enabling real-time modification of cooling intensity. This allows each die to operate at its optimal temperature regardless of varying thermal loads and environmental conditions, reducing overall power consumption by preventing overheating and associated leakage current
2Device complexity
If a single heat spreader/heat sink is used to cool all dies in an MCP, then the cooling system structure is simple, but the thermal management is suboptimal leading to reduced productivity
Solution Approach 1:
The heat spreader is divided into multiple discrete cooling channels, each serving a specific die. This segmentation enables independent thermal management for each die, allowing higher operating frequencies and improved productivity without excessive thermal interference between adjacent dies
Solution Approach 2:
Each cooling channel is optimized with local quality variations including die-specific heat spreader geometries, tailored fluid flow rates, and customized heat sink configurations. This local optimization enables each die to operate at peak performance levels appropriate to its specific thermal characteristics and computational workload
3Ease of operation
If uniform cooling is applied to all dies, then the cooling system is easy to control, but dies with different thermal properties cannot achieve optimal operating temperatures
Solution Approach 1:
The system incorporates dynamic fluid flow control mechanisms for each cooling channel, allowing real-time adjustment of cooling intensity. This enables each die to maintain its optimal operating temperature despite varying workloads and environmental conditions, while the control system adapts to thermal changes automatically
Solution Approach 2:
Temperature sensors monitor the junction temperature of each die and provide feedback to the control system. This feedback enables automatic adjustment of fluid flow rates to maintain optimal operating temperatures, balancing ease of operation with precise thermal management through closed-loop control
4Productivity
If higher frequencies are maintained for longer periods, then processing productivity increases, but thermal limits are exceeded leading to reliability issues
Solution Approach 1:
The cooling system dynamically adjusts fluid flow rates in response to real-time thermal conditions, enabling sustained high-frequency operation when thermal margins permit. This dynamic response prevents thermal limit violations while maximizing productivity by maintaining high operating frequencies for extended periods
Solution Approach 2:
The system proactively manages thermal conditions before critical limits are reached by continuously monitoring die temperatures and preemptively adjusting cooling intensity. This preliminary thermal management prevents reliability issues by maintaining safe operating margins while enabling extended high-frequency processing
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables each die to operate at higher frequencies for longer periods, reduces pump power requirements, and enhances cooling system efficiency by allowing tuned liquid flow rates, thereby improving silicon yields and meeting power/performance targets.
Implementation Method 1
a first heat spreader portion in thermal communication with the first die and a second heat spreader portion in thermal communication with the second die
Implementation Method 2
controlling an impinging fluid flow directed at each die within the MCP
Data Source
AI summary
Technologies for dynamic cooling include a computing device having a multi-chip package including multiple dies and a cold plate coupled to the multi-chip package. Micro nozzle valves are coupled to fluid passage zones of the cold plate positioned adjacent to the dies, and are configured to control fluid flow into the fluid passage zones. The computing device reads a predetermined die junction temperature for each die, determines a current die junction temperature for each die, compares the predetermined die junction temperature to the current die junction temperature for each die, and determines a fluid flow rate for each die based on that comparison. The computing device controls the micro nozzle valves adjacent to each die based on the respective fluid flow rate. The dies may include processor cores, field-programmable gate arrays, memory devices, or other computer chips. Other embodiments are described and claimed.