3D-IC: Great Opportunities, Great Challenges

The growing adoption of 2.5D and 3D integrated circuits (3D-ICs) marks a major inflection point in the world of semiconductor design. Electronic designers demand greater integration densities and faster data transfer rates to meet the growing performance requirements of AI/ML, 5G/6G networks and autonomous vehicles as these technologies have outpaced the capabilities of any single chip. 3D-IC technologies provide a solution embraced by leading-edge design companies and increasingly supported by semiconductor manufacturers.

3D-IC is an umbrella name for a range of manufacturing technologies that enable integrating multiple silicon die into a single package. The dice or chiplets can be placed closely on a connectivity substrate (2.5D) or stacked vertically on top of each other (3D). 3D-ICs offer many advantages over SoCs. They:

• Circumvent reticle limits that define the maximum size of a single die
• Provide faster signal speeds at lower power than traditional packages on printed circuit boards (PCB)
• Improve system performance by stacking memory chips on top of the logic that accesses it reducing delays
• Make it possible to use a mix of different silicon process technologies in one system

That’s the good news. The flip side is that designers must overcome serious technical and organizational challenges before they can capture these benefits and achieve a competitive advantage. The principal issue is that 3D-IC blurs the lines between three traditionally separate design disciplines: chip design, package design, and board design. The three draw on different design tools, data formats, terminology, and manufacturing knowledge bases. Now suddenly, 3D-IC is smashing them all together into a novel design challenge with significant contributions required from all three. Fundamental, disruptive structural change is coming to product engineering organizations with the emergence of a brew of competencies required to succeed in this new 3D world.

Blurring the worlds of chip, package, and board design

Today, the package design team is typically separate and distinct from the chip design team with only limited communication between the two. That all changes when design spans multiple dice and the packaging becomes an integral part of the system. Indeed, if logic blocks are dispersed across two or more die, they communicate through wiring contained in the package substrate or interposer layer, meaning the design of that package is no longer separate from the design and floor planning of the chip. Chiplets stacked vertically on top of other chiplets connected through direct microbump contacts further complicate the distinction between package and die.

The blurring is further exacerbated by the interconnect routing on a large 3D-IC substrate. The routing is a complex mixture of chip and board strategies like non-Manhattan routing, river routing, and electromagnetic modeling. Each 3D manufacturing technology is different, but they typically erase the notion of an easy separation between PCB design techniques and chip design techniques.

This blurring of the lines means semiconductor development teams will have to adapt to accommodate three major change areas:

• Cross-functional and interdisciplinary experts face the cultural/organizational challenge of working together as a close-knit team to bring 3D-IC designs to market. Designers must now collaborate more closely than ever, mastering the art of the board, the package, and the silicon simultaneously. The usual, siloed organization of design teams will have to be rethought. Silos must be eliminated so that collaboration can flourish.
• Traditional EDA tools do not support the advanced technical design requirements of 3D-IC design. A new generation of purpose-built tools are emerging that offer essential new capabilities like multi-die 3D floor planning. The expanded range of physical effects impacting the design is driving the need for open, flexible, extensible, and high-capacity design platforms. These new platforms must accommodate the broad array of analysis and design tools needed to solve the multitude of 3D-IC design challenges. No one tool addresses all the issues. And the huge capacity demands of these giant 3D systems make cloud computing absolutely essential. Closed single-company design systems won’t cut it!
• Multiphysics analysis is already a fact for advanced node design and a necessity for 3D-IC. Designers need to consider a range of novel physics challenges, many of which they never had to deal with before. Multiphysics is not the same as multiple physics – these physics interact in a non-linear fashion and need to be analyzed concurrently. 3D-IC teams must master new design constraints. Chief among these is thermal analysis at every step of the design flow.

Power dissipation is a primary constraint in 3D-IC design, and careful analysis is essential to ensure a reliable working device. Mechanical aspects related to stress and warpage of the package also must be considered right up front starting with floor planning, where the unfortunate placement of hot and cold components can doom the design from the outset. Other unfamiliar effects can come to the fore with distributed systems, like low-frequency power supply oscillations between components on the substrate. Multiphysics, not just multiple physics!

Managing organizational change for 3D-IC success

The pioneers in 3D-IC design have been mainly large, well-resourced corporations that benefit from a vertically integrated culture and organizational model that enables interdisciplinary collaboration, and stunningly advanced bespoke silicon. Other early adopters are start-up companies in the AI/ML field and similar businesses organized with 3D-IC in mind from the get-go. More mainstream, horizontally integrated companies can be challenged to achieve the level of teamwork, synergy, and expertise needed to optimize a 3D-IC design.

My experience working with Ansys customers leads me to believe the answer lies in creating a team environment with an open design platform that pulls together designers from different groups with the range of expertise required to address 3D-IC. Realistically, not every designer can be expert in the full range of physics that impact these systems, so it is essential to arm them with design tools and multiphysics analysis tools that automate the analysis as much as possible. A good analysis tool encapsulates the required expertise, making it available to designers through easy-to-understand outputs.

This kind of ecosystem provides a powerful advantage in focusing the entire, cross-functional team on a shared, real-time view of the product design – and driving the innovation that leads to success in today’s competitive worldwide semiconductor industry.

A good place to learn more about the state-of-the-art in multiphysics and 3D-IC design is the recent Ansys IDEAS Digital Forum, now available on-demand.