The ‘Intel’ on Advanced Packaging Options

Estimated reading time: 2 minutes

Dr. Tom Rucker is vice president in technology development at Intel and was a keynote speaker at the IPC Advanced Packaging Symposium, which helped set the table for the rest of the agenda. Tom understands this “radical and seismic” shift in terms of technology and breaks down what it means for the semiconductor and PCB fab industries. There’s absolutely a place at the table for PCB fabricators, but what are the first steps? 

Nolan Johnson: Tom, thanks for taking a moment to talk with me here.

Tom Rucker: My pleasure.

Johnson: You just finished your keynote. What was the message?

Rucker: My key message is that packaging is undergoing a very radical and seismic shift in terms of technology and driving the requirements to ensure that our customers, the whole industry, and the ecosystem can really support the computing changes happening across the industry. If you look at it from a compute perspective, there’s more data getting collected that needs to get stored and moved. It needs to get analyzed and decisions made. That drives a very large change in the capabilities that products need to deliver—more functions and performance. The key to enabling that technology is advanced packaging where you can take multiple different die and components, put them together in a very compact form factor, get that performance, and then give the consumer, the end user, what they need.

Johnson: In your keynote, you referred to a disaggregated implementation vs. the old-style monolithic approach. Your presentation included the discussion of several techniques for putting advanced packaging together. Is there room for these methods or will we see some of these shake out and consolidate?

Rucker: That’s a great question. Historically over many years, usually you had one silicon die—a monolithic piece of silicon—in one package. Now, there are many use cases for products which drive multiple die in one package. Some use cases may need more compute capability than graphics. Maybe some need two compute environments—one for massive calculations and one for simple data transformations or data aggregation. Machine learning, advanced artificial intelligence, with different architectures for different functions, need to be combined with more traditional general purpose computing architectures in one package.

With so many different use cases, putting all those on one die means the die would be enormous, and you wouldn’t have the optimized solution because you, the architect, cannot use different technologies and designs optimized for specific workloads and applications. With advanced packaging or disaggregated technology, it allows the architect to say, “Let me take this function and make it on this technology node, and this other function on a different technology node.” Then I can put those all into one small unit. For the end user it looks as one unit, but as we manufacture it, it’s internally, four or five different blocks. That’s a benefit of disaggregation. Higher performance in a small form factor.

To read this entire article, which appeared in the November 2022 issue of SMT007 Magazine, click here.