Historic FinFET/2.5D Firsts

Like the first trips to space, 2.5D design takes a lot of experimentation and practice.

popularity

Recently, I had an opportunity to watch the Academy Award nominated movie “Hidden Figures.” If you’re a geek at heart, you need to see this movie. It chronicles the strong contributions of three black women to the NASA space program during the 1960s. The civil rights backstory of the movie is powerful, but there is another aspect of the movie that stayed with me as well: all of the “firsts” for the United States that the movie documented. The first man in space. The first man to orbit the Earth. The first black female engineer. The list goes on.


Image courtesy of chrisdorney/Shutterstock.com

It got me thinking about “firsts” in our industry. eSilicon recently participated in one. Our first production 2.5D/finFET tapeout. This design was full of all kinds of interesting IP and sophisticated package technology. 2.5D and finFET technology have both been gaining momentum recently, and this chip is clearly not the first product to use both. From an ASIC perspective, it does represent an important milestone for us, however.

What you will learn from the movie is that getting to space and back required a lot of practice. Many simulations and experiments before NASA got it right. The same is true for our 2.5D design capability. We started doing test chips for this technology in 2011. There were a lot of tests, and a lot to learn. Tests to check package reliability. Others to test the robustness of assembly. Yet more to test thermal stability, and finally tests to verify the throughput of the completed device. All of these test vehicles formed the foundation for our opening night event of a full production tapeout of a complex 2.5D system.

Application-specific standard products (ASSPs) have been using 2.5D and finFET technology for a while now and there have been some dramatic firsts along the way. You can find out how AMD beat Nvidia to the market with its first 2.5D graphics card that utilized HBM memory stacks here. And then there is the Xilinx Virtex-7, the first commercial device to use this technology. All this can make someone a lot of money. Perhaps a $170B market by 2022.

FinFET technology has blazed some trails as well. Like the first 10nm finFET cell phone chip. Notwithstanding all these exciting firsts, there are “lasts” looming as well. Will Moore’s Law finally be repealed? Is there a point in time when transistor shrinking simply stops working? The folks at MIT think it’s coming soon, by 2021. So what happens next? The future actually looks quite bright. Lots more “firsts” to conquer. If things like quantum computing are interesting to you, take a look into the future here.

Meanwhile, back at eSilicon here on planet Earth, we will be taping out many more finFET/2.5D ASICs over the coming months. Each will do something that has never been done before. We’re excited and honored to be blazing these trails with our customers.