Coherent Cache Fabric Scaling for Active-Core Power Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The power consumption of processor system integrated circuit packages is a major limiting factor for performance due to heat dissipation challenges in slim form factors, leading to performance bottlenecks as components operate within a limited power budget.
Innovation Solution
Dynamically controlling coherent cache fabric (CCF) utilization by reducing or bypassing it to match active core scenarios, switching between high-performance and low-power modes based on core activity levels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If the coherent cache fabric is fully operational to maintain high performance, then processing speed is improved, but power consumption increases
Solution Approach 1:
The coherent cache fabric dynamically adjusts its operational state based on system workload conditions. When few cores are active, the fabric transitions to a reduced power mode with deactivated cache agent instances, minimizing power consumption while maintaining necessary functionality. When more cores become active, the fabric transitions back to full performance mode, ensuring processing speed is maintained when needed.
Solution Approach 2:
The system changes operational parameters of the coherent cache fabric by deactivating specific cache agent instances based on the number of active cores. This parameter change allows the fabric to operate in different states (full performance vs. reduced power) by modifying which components are active, thereby resolving the contradiction between speed and power consumption.
2Use of energy by moving object
If cache agent instances are deactivated to reduce power consumption, then power usage is improved, but access latency increases
Solution Approach 1:
The coherent cache fabric is segmented into multiple independent cache agent instances, each associated with specific shared cache circuit blocks. This segmentation allows selective deactivation of only those cache agents needed for inactive cores, while keeping cache agents for active cores operational. This resolves the contradiction by minimizing power consumption through selective deactivation while maintaining low latency for active operations.
3Use of energy by moving object
If the coherent cache fabric is reduced to lower power consumption, then energy efficiency is improved, but cache capacity decreases
Solution Approach 1:
The reduction in cache capacity is localized to only those cache agent instances that are deactivated based on core activity. The cache capacity associated with inactive cores is deactivated, while cache capacity for active cores remains fully available. This local quality approach ensures energy efficiency is improved without unnecessarily reducing the cache capacity needed for active processing operations.
Data Source
Figure 1
Figure 2A~2B
Figure 3
AI summary
Power may be reduced by dynamically controlling coherent cache fabric (CCF) utilization to efficiently support the number of active cores. In some embodiments, this may be achieved by dynamically reducing or even bypassing the CCF.