²ÝÁñÉçÇø

Integration of Multiple Dies Creates Interdependencies

Workload-intensive applications such as high-performance computing (HPC) and AI drive demand for multi-die systems to deliver the performance, power, and area (PPA) these applications require. A single package integrating multiple individual dies containing different types of circuits, a multi-die system presents an efficient way to accelerate the scaling of system functionality. With key performance indicators (KPIs) in place, designers can plan their systems and components accordingly.

Multi-physics effects in monolithic SoCs are concerning but can typically be modeled and analyzed upfront and addressed with design adjustments. With multi-die systems, these effects, if unaddressed, become much more detrimental to the entire system.

For example:

• With so many dies talking to each other via die-to-die interfaces, the system becomes more susceptible to signal integrity issues such as electromagnetic interference (EMI) and crosstalk
• Thermal issues rise to the forefront, since all of the dies in the system, along with the interconnections between them, could trap heat and, thus, impact the system¡¯s performance and/or timing
• The robustness of the system¡¯s power distribution network in alleviating effects like EMI while providing the power needed for the entire die system is critical
• Mechanical stress stemming from the fabrication, assembly, and packaging of chips can affect the electrical performance of the system
• Process variation occurring in the system¡¯s individual dies can impact system performance

To address these effects, analyses of the system must be performed in the context of the multi-die system, examining all the physical aspects and interactions to validate and optimize the system. Multiple properties, from electrical and thermal effects to structural mechanics, must be simulated together to identify power and signal integrity impacts at the system level. What¡¯s needed to accomplish this is a unified STCO solution that can deliver full-spectrum design closure and convergence for optimal PPA per cubic mm. 

Understanding Design Tradeoffs Early On

STCO starts at the beginning when design teams are planning the architecture. As they determine their system¡¯s components and how it should all be partitioned, they need to be able to perform ¡°what if?¡± analysis to understand their tradeoffs. The more granularity they have as they move toward implementation, the better as they converge toward signoff. At this early design stage, limited information is available; however, system prototyping helps answer several questions: How many dies are needed? How should the die be stacked? What are the tradeoffs of mixing older and newer nodes in the system? How should the power distribution network be designed? What kinds of thermal issues might arise? As the team converges on an architecture to meet their requirements, they can take outputs from feasibility studies and move into system planning and building prototypes.

The ²ÝÁñÉçÇø Multi-Die System Solution provides a comprehensive platform for fast 2.5D and 3D heterogeneous integration to support an STCO approach that helps answer the ¡°what if?¡± questions. Two key components of the solution that address multi-physics effects are:

• ²ÝÁñÉçÇø 3DIC Compiler multi-die/package co-design and co-optimization platform, which provides modeling capabilities and the associated data points that can guide the development of a system floorplan that delivers the targeted PPA. Built on the standard, single-data-model Fusion infrastructure of the ²ÝÁñÉçÇø Digital Design Family, 3DIC Compiler provides a complete architecture-to-signoff platform.
• Design analysis and signoff technologies, which address static timing, signal integrity, power integrity, thermal, parasitics, and electromigration/IR-drop. 3DIC Compiler and ²ÝÁñÉçÇø signoff solutions are integrated with , which provides multi-physics power integrity, signal integrity, thermal integrity, and mechanical stress simulation and analysis for 2.5D/3D multi-die systems.

Since the Multi-Die System Solution¡¯s components are tightly integrated, design teams can find and resolve problems earlier on to achieve better yields. Disparate tools can be used for this analysis, but if they¡¯re not well connected, they could miss important ¡°red flags¡± along the way, potentially forcing design teams to address problems closer to tape-out when it is much more costly to do so. ²ÝÁñÉçÇø continually assesses emerging multi-die system effects, and enhances its solution accordingly.