The interdependence of semiconductor devices and companies in manufacturing was a recurring theme at this year’s SEMICON West, both in presentations and one-on-one discussions. Challenges range from sharing data securely across a highly integrated supply chain, particularly in light of heterogeneous integration, security concerns, and the increased use of AI, as well as concerns about the robustness of the supply chain.
All of this becomes increasingly challenging as the IC industry shifts from cramming everything onto a single SoC to advanced packaging with chiplets.
“If you think about traditional advanced packaging, you had a single flip chip, you could have a single bond layer. You’d reflow that chip, and if there were any issues with the connection, you could pop that out and disassemble it,” said Brad Perkins, product line director at Nordson Test & Inspection. “When you get into heterogeneous packaging, and multiple stacks of them, the ability to disassemble, reflow, and do that in a cost-competitive way becomes a problem. Also, you may have a non-conductive film that isn’t workable in a high-bandwidth memory stack. It just gets more complex because you have multiple layers and more things that can’t be reworked.”
Chiplets and heterogeneous integration also require a strategy for sharing data, particularly as chiplets shift from internally developed to a commercial, off-the-shelf model. “It requires adding another layer on top of the data for access controls, very similar to managing export controls for IP,” said Simon Rance, vice president of marketing at Cliosoft. “And it needs to be completely programmable, so customers can program that exactly to their requirements. They can isolate or geofence a piece of the data or a piece of the design and control who can have access to it, who can see it or even know about it.”
For the chip industry to make progress on any major effort, such as smart manufacturing, cybersecurity, or sustainability, it requires collaboration up and down the supply/value chain. And that collaboration comes in many forms, from defining standards, partnering in the development of new technologies, to joining regional groups with common business objectives.
For instance, to address the carbon emissions from power utilities requires a consortium to negotiate a cost-effective rate. “We will be aggregating the demand and the consumption of the suppliers depending on where they are geographically,” said Dallai Slimani, vice president of the semiconductor segment at Schneider Electric. “Many of the suppliers are so small they don’t have access to some energy markets. This aggregation will actually provide the whole cohort access to those markets.”
Alan Weber of Cimetrix, a PDF Solutions company, presented results of SEMI member survey of their usage of the EDA standard. Defined Circa 2001 it supports a higher density of equipment data (volume and rate) as well as the flexibity to change data collection needs. Users stressed the importance of EDA in supporting real-time dispatch capabilities and other benefits to streaming data real-time. Speaking of real-time dispatch, Minds.ai and D-Simlab Technologies both presented on AI methods in to improve real-time scheduling in wafer fabs. SemiEngineering