Unified Processor and Heterogeneous Memory Stack for Lower RC Delay

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional semiconductor devices face challenges with cross-talk and high loading on processors due to the use of homogeneous memory types and long-distance metal wires, leading to RC delay and increased PCB area, which affects data storage performance and efficiency.

Innovation Solution

A unified semiconductor device with a processor core and heterogeneous memories (DRAM and NAND) integrated in a multi-chip package or on the same bonded chip, utilizing short-distance vertical metal interconnects and reducing or eliminating bus interface units to enhance data storage performance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If homogeneous memory types and long-distance metal wires are used, then device complexity is reduced, but RC delay increases and data storage performance deteriorates

Engineering Contradiction:
Improvememory structureVSAvoidRC delay
Core Design Contradiction:
Device complexityVSLoss of time

Solution Approach 1:

The patent transitions from planar (2D) memory architecture to three-dimensional (3D) stacked architecture, where memory layers are stacked vertically above the processor. This dimensional change enables short-distance vertical interconnects through the substrate, dramatically reducing RC delay while maintaining heterogeneous memory integration.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Solution Approach 2:

The patent segments different types of memory (DRAM, NAND, SRAM) into separate stacked layers, each optimized for specific functions. This segmentation allows heterogeneous memory types to be integrated without increasing overall device complexity, as each layer operates independently with dedicated interconnects.

Inventive Principle:
Principle #1Segmentation

2Quantity of substance

If heterogeneous memories are integrated in multi-chip package, then data storage capacity increases, but device complexity increases

Engineering Contradiction:
Improvedata storage capacityVSAvoidmulti-chip package
Core Design Contradiction:
Quantity of substanceVSDevice complexity

Solution Approach 1:

The patent merges processor and heterogeneous memory into a single integrated device structure, eliminating the need for separate multi-chip packages. The processor and memory layers are bonded together in a unified stack, increasing data storage capacity while reducing package-level complexity through monolithic integration.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The integrated device structure serves multiple functions simultaneously: the processor handles computation while stacked memory layers provide various types of storage (volatile DRAM, non-volatile NAND, cache SRAM). This multi-functionality eliminates the need for separate chips and packages, reducing overall device complexity.

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Manufacturing precision

If long-distance metal wires are used, then manufacturing precision requirements are reduced, but RC delay increases and processor loading increases

Engineering Contradiction:
Improvewire placementVSAvoidRC delay
Core Design Contradiction:
Manufacturing precisionVSLoss of time

Solution Approach 1:

The patent replaces long-distance lateral wire connections with short-distance vertical interconnects through the substrate. This dimensional change in signal routing dramatically reduces RC delay while the standardized vertical interconnect structure simplifies manufacturing precision requirements compared to complex lateral routing.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

4Productivity

If bus interface units are eliminated, then processor loading is reduced, but device complexity increases due to direct integration

Engineering Contradiction:
Improveprocessor efficiencyVSAvoidintegration structure
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent merges the functions previously performed by separate bus interface units directly into the processor and memory layers. Data transfer occurs through direct vertical interconnects between stacked layers, eliminating the need for intermediate bus interface units and reducing processor loading while simplifying the overall integration structure.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentUS11749641B2Unified semiconductor devices having processor and heterogeneous memories and methods for forming the same
Publication Date: 2023.09.05 YANGTZE MEMORY TECH CO LTD
  • US11749641B2 patent drawing
  • US11749641B2 patent drawing
  • US11749641B2 patent drawing

AI summary

Embodiments of semiconductor devices and fabrication methods thereof are disclosed. In an example, a semiconductor device includes NAND memory cells and a first bonding layer including first bonding contacts. The semiconductor device also includes a second semiconductor structure including DRAM cells and a second bonding layer including second bonding contacts. The semiconductor device also includes a third semiconductor structure including a processor, SRAM cells, and a third bonding layer including third bonding contacts. The semiconductor device further includes a first bonding interface between the first and third bonding layers, and a second bonding interface between the second and third bonding layers. The first bonding contacts are in contact with a first set of the third bonding contacts at the first bonding interface. The second bonding contacts are in contact with a second set of the third bonding contacts at the second bonding interface. The first and second bonding interfaces are in a same plane.