The Need for System-Technology Co-Optimization (STCO)

Modern semiconductor components are becoming more and more complex and cost sensitive. To master technological and economic challenges, new chiplet approaches and heterogeneous integration technologies are becoming increasingly relevant. This, in turn, calls for new heterogeneous design approaches. They make it possible to combine different design domains across technological options while satisfying economic and performance demands on the application. It is crucial to ensure that the heterogeneous system designed meets the requirements with all relevant effects considered while fulfilling the targeted cost figures. Thus, a broad spectrum of information must be considered and optimized altogether.

Handling this range of technological and design possibilities while targeting cost figures requires a new and extended design approach: the System-Technology Co-Optimization design methodology – or STCO for short. This approach addresses overall system optimization across different fields of expertise, across abstraction levels or hierarchies, and across different domains of KPI measures. For example, design experts discuss with process technology experts covering everything from the system view down to the technology detail while at the same time tweaking performance and cost targets. This sheer complexity is such that “gut feeling” is no longer enough to handle it entirely.

That is what makes support from cross-domain tools a critical contribution to decision-making in the design process of heterogeneous systems. At the same time, these tools must largely cover the design flow from requirements to implementation while also providing backward paths for optimization. Such tooling must help capture, aggregate, and connect requirements, data, rules, etc. to enable the monitoring and control of economic, technological, and design implications stemming from the components, packages, and chiplets that altogether form complex heterogeneous systems for various target applications. This calls for a representation of the many details, which must be made available to various experts across the development teams and in the value chain.

Rather than being a single tool, this is a philosophy. Still, tools that support this philosophy in a way that connects the dots will bring enormous value: Problems at handover points will be identified earlier, iteration across teams will be accelerated, and the alignments across expertise fields and domains will be improved. Overall, the many variables from system design to process technology will have to be clearly visible, and their effect on system performance and costs should be visible quickly and in a context.

We need System-Technology Co-Optimization to design cutting-edge heterogeneous hardware, and we need the right tools and establish the right philosophy to make STCO happen. This will push the semiconductor ecosystem ever further forward into the new and exciting chiplet era.