Stacked Cache Layout With Central TSVs for Balanced Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional planar cache configurations experience increased latency and power consumption due to pipeline staging as cache size increases, leading to inefficient data access and return times.
Innovation Solution
A stacked cache system with vertically oriented cache dies and centrally located cache control circuitry, utilizing connection vias in the center to minimize pipeline stages and achieve balanced latencies, thereby reducing access request cycles and power consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If cache size is increased in conventional planar configurations, then cache capacity is improved, but latency and power consumption increase
Solution Approach 1:
The patent transitions from a planar two-dimensional cache layout to a three-dimensional stacked architecture. Multiple cache dies are stacked vertically and interconnected through through-silicon vias (TSVs), enabling data access across vertical dimensions. This dimensional change allows larger cache capacity while maintaining shorter access paths, as TSVs provide direct vertical connectivity rather than requiring long horizontal wire routing across the die.
Solution Approach 2:
The cache is divided into multiple separate dies stacked vertically, with each die containing a portion of the cache capacity. The cache control circuitry is segmented and distributed across different levels of the stack. This segmentation allows each die to be independently optimized and connected through TSVs, reducing the latency penalty associated with larger cache sizes by breaking the monolithic structure into manageable segments.
2Quantity of substance
If cache size is increased in conventional planar configurations, then cache capacity is improved, but power consumption increases
Solution Approach 1:
By stacking cache dies vertically and using TSVs for interconnection, the patent reduces the horizontal wire length required to access cache data. Shorter wire lengths mean lower capacitive loading and reduced dynamic power consumption. The vertical stacking allows larger cache capacity to be achieved without proportionally increasing power consumption, as the vertical TSV connections are more efficient than extended horizontal routing.
3Quantity of substance
If pipeline stages are added to increase cache size, then cache capacity is improved, but access cycles increase
Solution Approach 1:
The stacked architecture with TSV interconnections provides direct vertical pathways for data access, eliminating the need for additional pipeline stages that would be required in planar configurations to route data across larger distances. The three-dimensional interconnect structure allows data to reach the cache control circuitry in fewer stages, maintaining high access speed even as cache capacity increases.
4Quantity of substance
If wire routing is extended to access distant cache portions, then cache capacity is improved, but signal loading and heat generation increase
Solution Approach 1:
By organizing cache data in a vertical stack and using TSVs for interconnection, the patent minimizes the physical distance signals must travel. The vertical TSV connections provide direct, short pathways between cache dies and control circuitry, reducing capacitive loading and signal degradation. This also reduces heat generation compared to extended horizontal wire routing, as the shorter interconnect paths dissipate less power and generate less heat.
Data Source
AI summary
In accordance with described techniques for balanced latency stacked cache, a stacked cache system includes a first cache die and at least a second cache die in a stacked orientation with the first cache die. The stacked cache system includes cache control circuitry that is centrally located in the stacked cache system. The stacked cache system also includes connection vias configured vertically in a center of the stacked cache system as interconnected inputs and outputs of the first cache die and the second cache die.