By Robert Aslett
President and CEO
Silicon Integration Initiative
Thanks to the participation and leadership of our members, Si2 has made very good progress on several topics this quarter. I highlight a couple of them below. If you are interested in learning more about any of them or to get more information about Si2, please check out our website Si2 or contact us at Contact Si2.
Welcome Two New Members to the Si2 Board of Directors
I want to start off by welcoming two new members to the Si2 Board of Directors. Manoj Selva is vice president and general manager of the Intel Product Enablement Solutions Group. Manoj is responsible for leading the development and deployment of design tools, flows and methodologies to Intel’s product development teams. Manoj is replacing Rahul Goyal who served on the Si2 Board for over twenty years including ten as Board chair. Jim Culp is a Corporate Fellow leading GlobalFoundries CAD and Design Methodology teams. His team is currently deploying 3d design kits for RF/AMS applications for next-generation consumer and A&D applications. Jim is replacing Richard Trihy who served on the Si2 Board for eight years.
The Inaugural International Compact Modeling Conference is Coming in June ’25
Planning for the inaugural International Compact Modeling Conference (ICMC) is progressing nicely. This conference is co-sponsored by Si2, the IEEE, and the Electron Devices Society. It will be held in San Francisco on June 26th and 27th. The conference will uniquely focus on compact semiconductor device model applications, model Q&A, parameter extraction methodologies, model development methods, and model standardization. It will be a great opportunity to share the best practices related to semiconductor device model use and development and increase industry and academic collaboration. For more information, please go to ICMC 2025.
Si2 Becomes a Core Member of the National Semiconductor Technology Center (NSTC)
Si2 has joined the National Semiconductor Technology Center (NSTC). The NSTC is a key part of the CHIPS for America initiative, which aims to bolster the U.S. semiconductor industry and serves as a central hub for research and development in semiconductor technology. It recently published an updated strategy and set of goals (see figure below). Of primary interest to Si2 and our members is the second goal to, “reduce the time and cost to prototype” and the subgoal to “facilitate access to electronic design automation (EDA) and design resources, tools, and shared datasets.” A key part of that subgoal is the establishment of a design enablement gateway (DEG) that “provides members with a secure, cloud-based service with convenient access to design tools, reference flows and designs, and shared data sets.” Si2 will be engaging with other members of the NSTC to explore this topic in more depth in the coming months. You can read more about the revised NSTC strategic plan at NSTC-Strategic-Plan
Open Access Continues to Extend into New Design Domains
As most people know, OpenAccess is an API into standard for EDA tool databases. It includes some powerful capabilities. For example, it is:
- Managed: Supports simultaneous users with a rich set of supervisory processes.
- Extensible: Enables adding new data types, classes, methods, and supervisory processes.
- Interoperable: Applications interact at the database level to enable interoperability and speed.
- Synchronized: Updated versions are synchronized across the EDA industry by the Change Team.
OA can also store anything you define. Not just layout, schematic, symbol, abstract, netlist, and parasitics, but also text, images, stimulus, vectors, results, and software. And the module domain allows high-level design data to integrate with physical design data.
Central to all Open Access extensions is oaScript. The API is wrapped in Python, Ruby, and tcl to provide scripting APIs. Extensions developed in C++ and directly used with the Reference Implementation are wrapped in Python, Ruby, and tcl, enabling everyone to work with the oaScript APIs.
All of this means that OpenAccess is used for more than just physical design by several Si2 OAC members. This is because most EDA tools accommodate OA import/export, even if they use a proprietary database. For example, some OAC members utilize OA in the DFM and DFT domains. Another OAC member utilizes OA in the HLD domain. There are some technical challenges, but they can be overcome by extending the API under the umbrella of the Extension Steering Group (ESG). Each additional area of the design process(es) using OpenAccess creates more value through increased interoperability.
One interesting application of OpenAccess capabilities is in the design of 3D ICs. The ability to traverse hierarchies and adopt and release multiple technology files (technologies) means that existing chip-level capabilities can be applied across a stack of chips built using different process technologies. Recently, an Si2 intern wrote a simple Python tool and used the OA API to compose and view a design comprised of three chips using three different technologies and demonstrated how using OA allowed the user to operate on the 3D design as if they were working on a single chip.
Getting a Handle on AI/ML for EDA via the Si2 SIG and OpenStandards Working Group
Activity around the application of is AI/ML for EDA continues to gain momentum in the SI2 Special Interest Group and the new AI/ML Schema Open Standards Working Group (OSWG). A summary of each is given in the picture below as well as a summary of the possible bridge from OA that is also under discussion.
The SIG is largely focused on LLM investigation. The objective is to investigate concepts that complement existing EDA efforts around topics like framework-tuned chip design, retrieval-augmented generation (RAG), knowledge embedding, cogeneration, and other topics that come up. The SIG focuses more on exploration and discussion, and the topic can be quite wide-ranging.
The new AI/ML Schema OSWG is focused on prototyping possible solutions that will enable academia and industry collaboration in this space. The objective is to come up with a standard schema spec that supports AI/ML algorithms and data analytics and enables the sharing of algorithms and efforts between academia and industry without requiring companies to share their data. The working group oversees two efforts: one at Drexel University, where they have their own EDA schema representation, and an effort at ASU in conjunction with NVIDIA, utilizing the CircuitOps schema. The working group specifies the requirements and priorities for these efforts and feeds those to the programs.
Another interesting topic in this space relates to the value that may be achieved by providing a bridge from OpenAccess (OA). For example, the input into the EDA schema from Drexel is currently implemented using parsers from vendor reports. OA could provide greater efficiency as a connection to the commercial tools they are using now. Similarly, CircuitOps currently interfaces primarily with OpenROAD. An interface with OA would expand the application space for CircuitOps to connect with commercial tools. OA may provide a common method to bring data from commercial EDA tools into the standard schema. The first phase could be feeding it directly into the schema, and the second phase could be providing data within a reinforcement learning loop.
