2.5D and 3D: Life Preservers for Foundries?

Wow, times have changed. Who would have thought the foundries would ever willingly turn to the packaging industry for the solutions to future scaling? But I heard it myself yesterday at the ECTC luncheon keynote address, straight from the lips of Global Foundries’ CTO Greg Bartlett, who said that “silicon is not a great place to be right now,” referring to challenges with scaling beyond 28nm, and how scaling is running out of steam. “We’re looking across the abyss to packaging and hope 2.5D and 3D guys are going to throw us a life preserver.” That statement, I’m pretty sure, was music to many more ears than just mine – and particularly those who have been driving the move to 3D integration for the past several years.

Calling the semiconductor supply chain the “most complex in the history of mankind”. Bartlett talked about what got us to this bend in the road. Mostly, it’s our own fault. We, as consumers, have created marketplace demand because of our need for instantaneous access to information.  “How many people do you see looking at their smartphones rather than talking into them?” he asked. “This means something.” Smartphones put the power of a laptop in our hands, but the current tradeoff for all this insatiable need for video and data bandwidth is reduced performance, greater power consumption and therefore shorter battery life. This has caused an “explosion of wafer level products driven by I/O density, performance and form factor. 2.5D and 3D technologies are expected to address these issues further, and Bartlett sees an “absolute revolution occurring” where 3D allows a fully integrated solution that reaches up to the system level architecture.

He reiterated that on the silicon side, “all is not rosy” because not only is scaling further proving to be very costly, while at the same time realizing reduced benefits. He said 28nm is turning out to be “the last good node.” Scaling to 20nm or lower requires double and even triple patterning, which in turn needs advanced lithography tools, each bearing a $70M price tag. As a result, he said silicon designers are looking for new knobs – particularly in packaging – to fill the bill.  2.5D and 3D packaging allows us to scale each function, rather than the whole SoC, and the savings can then be derived at the system level, for example from reduced board real estate and power consumption.

The other crystal clear message Bartlett left attendees with was GlobalFoundries’ commitment to collaboration. The current model of isolating the development of chip and package technologies to the foundry and OSAT respectively won’t work, because as Bartlett explained, “solutions developed in isolation are no longer adequate to address the complexity required.”  Additionally, why won’t old model IDMs work? Because technologies that are optimized for a single product or application lack flexibility. “The best solutions rarely originate from an isolated team,” noted Bartlett.

Rather, he called for supply chain alliances, comprising R&D consortia, foundries and customers, tool and material suppliers, component manufacturers, etc. “Collaborative solutions utilize the best minds,” noted Bartlett. He also talked about creating “Centers of Excellence” where collaborative partners can be co-located to develop new tools, process and technologies for joining silicon, because bringing all the elements together allows for rapid cycles of learning and subsequently leads to value creation. In this perfect world of Open Architecture, customers have transparency and decide where and with whom they want to work, explained Bartlett.

This ‘virtual IDM’ could be a fabless customer’s dream, and light years apart from both Intel’s isolated approach and TSMC’s vertically integrated approach. It’s also in alignment to what the OSATS discussed earlier in the week. By committing to this approach, Global Foundries won’t need a life preserver, because as the title to Bartlett’s keynote address implies, they’ve decided to build a bridge across the abyss between silicon and packaging instead. ~ F.v.T.