Stacked-Die Thermal Gradient Compensation for DRAM Refresh Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Semiconductor devices experience operating errors due to induced thermal gradients, particularly in stacked configurations where thermal sensors on DRAM chips may not accurately detect hot spots caused by mechanical coupling with logic chips, leading to inappropriate refresh frequencies and potential data loss.
Innovation Solution
Standardizing the location of thermal sensors across all devices in a stack, allowing the SoC to calculate temperature differences and communicate these to DRAM chips to adjust refresh rates using a temperature-compensated self-refresh circuit, or calculating the maximum temperature gradient to ensure more frequent refreshes and prevent data loss.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If thermal sensors are placed on DRAM chips to monitor temperature, then temperature monitoring capability is provided, but the sensors may not accurately detect hotspots caused by thermal coupling with logic chips
Solution Approach 1:
The patent introduces an intermediary mechanism where the SoC calculates the temperature difference between its internal thermal sensor and the DRAM thermal sensor, then communicates this difference to the DRAM. This intermediary calculation and communication process allows the DRAM to compensate for thermal coupling effects and accurately determine the temperature at the hotspot location, thereby resolving the inaccuracy of direct thermal sensor placement.
2Reliability
If refresh frequency is increased to prevent data loss from thermal gradients, then data integrity is maintained, but power consumption increases
Solution Approach 1:
The patent implements dynamic adjustment of refresh frequency based on actual temperature conditions. The DRAM uses the temperature difference information from the SoC to dynamically determine the appropriate refresh rate, rather than using a fixed high refresh rate. This dynamic approach maintains data integrity when thermal gradients are present while reducing power consumption when temperatures are stable, resolving the contradiction between reliability and energy use.
3Measurement precision
If multiple thermal sensors are placed on DRAM chips to detect hotspots, then measurement accuracy improves, but device complexity increases
Solution Approach 1:
The patent extracts the complex hotspot detection function from the DRAM device itself and relocates it to the SoC. Instead of placing multiple thermal sensors on the DRAM chip, the solution uses a single thermal sensor on the SoC that calculates the temperature difference and communicates it to the DRAM. This extraction of the detection function reduces device complexity while maintaining measurement precision through the centralized calculation approach.
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 ensures accurate detection of thermal gradients, preventing data loss by adjusting refresh rates based on actual temperature differences, thereby maintaining data integrity and optimizing power consumption.
Implementation Method 1
induced thermal gradients between one and another of the semiconductor devices
Implementation Method 2
temperature-compensated self-refresh circuit, or calculating the maximum temperature gradient to ensure data integrity
Data Source
AI summary
A temperature difference between a first thermal sensor and a second thermal sensor on a first die is determined. The temperature difference is transmitted from the first die to a circuit on a second die. A temperature from a thermal sensor on the second die is determined. The temperature difference and the temperature from the thermal sensor are utilized on the second die to modify operational characteristics of one or more circuits on the second die.