Hi everyone, and welcome to Advanced Insights. I'm Mark Papermaster, CTO and EVP at AMD. We're so excited to continue this new series where we provide just what the show name suggests, advanced insights into some of the most exciting trends and topics in technology. I love talking about technology, and so do the industry experts that we interview on this show. With that, let's get right into our chat with AMD partner and my colleague, Dr. Y.J. Mii. On this episode, the semiconductor industry is concerned about what happens at 2 nanometer and beyond. Have we run into the fundamental limits of physics, or will innovation prevail once again? Y.J. Mii is executive vice president at TSMC, Taiwan Semiconductor Corporation, and is newly named as the co-COO of the company. Many people say Moore's Law has stopped, that we cannot continue this innovation. I have had such a pleasure working with you over the years, and I know that is not your belief. Thank you, Mark. It's really a very exciting time for us to work together. It's also our great pleasure to work with you. I'd actually like to start with a little bit of looking back. When I joined AMD, it was 12 years ago, and one of the things that impressed me is that AMD had had such a long history of partnership and co-innovation with TSMC. It was about 10 and 12 years ago that we really started seeing the limits of scaling. It became apparent to you and I both at that time that it might need a different approach, that a fabulous company working with a foundry, such an established foundry such as TSMC, would probably have to change the way that they work together going forward. If you could share with us a little bit of your view. You've been on this journey for so many years. How did you think about scaling? I joined TSMC for 30 years. For 30 years, we started from 0.5 micron, and now we are preparing 2 nanometers. Really, the scaling is over 4,000 times. I think performance and the power of our electronics product also improve accordingly. Scaling is getting more challenging, obvious, as we all know. However, we still see a lot of opportunity, not only from density scaling, in addition, or more importantly, from performance and power improvement. The industry is moving ahead with more and more innovation, and that's one of our key focus. There are thousands, tens of thousands of people our whole industry working together. We still see innovation come out here and there. It will surprise you and surprise us, too. Every time you come out with a new technology node, Y.J., you have all of these different vectors that have to come together to create that new technology node, and it all has to work perfectly. How do you juggle all of these pieces that have to come together? In the past, indeed, it probably took two, three years to get things done, but right now, because technology is more and more challenging, it takes more time, sometimes five years or even seven years. So we do have a step-by-step process, starting with research center, our past finding and research. Then we introduce something with supplier. Then, of course, our technology development team. We have two different technology development teams. They go one generation, the other group take another generation, so we have more time and resource to work on the technology. And also very important thing is we listen to our customers. So many of the technology, how we define, we have to work with our customers in order to make sure this technology really has its value on our customer's product. So you could sometimes drive some design rule too aggressively, which doesn't bring us enough benefit. So I think we did a good job balancing this for the past few generations. It's challenging, but we have a lot of experience, and we're confident on that. Before we get to two nanometer and new materials and things, just like three nanometer, did it start off just as you planned, or did you need to rally the team around solving challenges and developing that cutting edge foundry node. We have our three nanometer technology starting production end of 2022, and we continue to enhance it. And three nanometer technology, just like many technologies starting from seven and five nanometer, almost every generation of technology starting from seven nanometer, you need to have new innovations. And as I described earlier, we need innovation from material, from processes, and in addition, we also need innovation from design. So in our latest few generations of technology, we have a lot of collaboration with our customers and also very importantly with our EDA tool vendor. So in order to co-optimize the process and design, we call it DTCO. So that really helps us relieve the pressure of pushing the scaling. And with that kind of design and technological optimization, it will make it easier, make us be able to choose a less aggressive process, but still get the same performance and density. And as TSMC, because we have large volume manufacturing for many generations, we do have a lot of experience that help our team to pick the right technology, knobs for both density and performance and power. And we consider the potential risk so we can have a more balanced choice. I think that's extremely critical. I think that's exactly the nugget that people don't realize. It's that experience to know where you can push and where you actually need to back off. And that's not inherent. You don't read that in a textbook. That comes from deep experience and you all have the ability to leverage that experience and react quickly. I'm glad you mentioned DTCO, design technology co-optimization. I feel you and I personally have embraced that approach. We brought our teams together very closely. We almost operate as one company and we respect, of course, the proprietary information from each side. But then on the areas where we intersect, where design meets technology, it's an incredible collaboration. And I think we've leveraged that, I think, to great success, but never more than 2 nanometer. 2 nanometer brings on a whole other set of challenges. I would love to get your thoughts about 2 nanometers. Some of the challenges and why is it that DTCO, this technology and design co-optimization, is so important along with the process and tooling and materials development. I would love to get your perspective. Yeah, I think 2 nanometer is the whole industry moving from FinFET to what we call the Gate All-Around or what TSMC calls Nanosheet Transistor Technology. This technology, this new generation of transistors really have Gate All-Around the transistor channel. So it really has better control of the transistor performance and short channel effect. So it will provide you better performance and better power efficiency. However, to get that technology done, it changed the whole process. So it's indeed a lot more complicated. And so far we are doing quite well, both performance and yield. We are either on schedule or ahead of schedule. So we are confident we are preparing for volume production in 2025. And why is it getting more important to have design and technology co-optimization? I think today's technology, because of the physics challenge, so the process becomes more complicated. And the rule and the SPICE model also get a lot more complicated. So in order to fully get the benefit of a technology, it's extremely important to have design and technology team work together. That's what we call design technology co-optimization. The last bit of juice we can squeeze out from the technology, as we have all experienced for the past few years, our very successful DTCO drive. And we see the good result from it. And just because technology is getting more complicated, it's hard without the co-optimization. Yeah, that's so well stated. And I know, as your customers, we make it hard on you, because we have devices which are battery operated. We have devices that need to run at the highest performance and can tolerate even additional power or current that we may need to give to those devices. So think about the difference between a gaming desktop machine versus a notebook, or your customers are driving mobile phones. So you end up with such a range of design requirements upon you. And I've been impressed with DTCO that you've been able to segment these different requirements. And that had to be a change for you, to think about this different type of application and how do you support that from TSMC with new technologies like Gate All-Around and, as you say, the nanosheets with 2nm. Yeah, I think the strengths of TSMC and the Foundry model is we do have a large variety of customers starting from mobile, very low energy, to product, need performance, need power efficiency. So we do have a lot of experience on that. We work with multiple customers. That's the strengths that TSMC has. In the past, indeed, we started mostly from mobile. But mobile products, like a smartphone, is also complicated. It requires performance at the same time, power efficiency and battery lifetime. And today we see the emerging of high-performance computing, especially AI machine learning. But we find the AI machine learning product also getting more and more conscious and especially focused on power efficiency. And actually, that's not too far away from the mobile requirement. So I think that also plays TSMC's strengths. And because of our past, we're also working on high-performance computing products, which push for maximum performance. So we know how to push that. So we have a base built on power efficient compute for both mobile and AI machine learning. And then we extend our technology with additional features. A lot of time in the backend interconnect, as well as we have a differentiated device optimization. Such we can cover all that. So a good thing is really we have a lot of experience, so we know where to go. These new devices, which are staying on a Moore's law pace with these innovative chiplet approaches different nodes coming together, vertical stacked, horizontal stacked. That's challenged you in whole new ways. So in addition to cutting edge new semiconductor nodes, it's requiring new technologies to interconnect these devices. I think TSMC start to see the need of advanced packaging because we started more than 10 years ago and the study of advanced packaging technology because we saw the system performance start to get limited by packaging. And I think in TSMC, people have the vision in the past and pushing for that. So I think we are ahead of the industry in that development. So far right now, we already have a lot of solutions for our customers, including 2.5D info CoWoS for mobile, as well as high-performance computing, as well as the latest developed 3D die stacking SOIC. And AMD is one of the strongest partners for us, not just from the advanced logic technology, also for the advanced packaging technology. And together, that's why we see the MI300, the successful release of MI300, which not only use our leading edge technology with our 3DIC die stacking, as well as our latest and best CoWoS technology together. What is your thought going forward of beyond 2 nanometer? Can we continue to drive innovations to keep performance and density of performance going forward and advancing in our industry? We see still a lot of room for scale, for the most to extend, although indeed the scaling become more and more challenging. And we still see room for density and also especially on performance and power. So there are more process knob opportunity we can work on that. At the same time, because 2D scaling become more challenging and sometimes costly, so we are also working on 3D scaling, not just in the chip level, but also in the transistor level. As you probably already heard that we are researching on 3D transistor structure. It's called CFET. You're stacking M MOS and P MOS together. That itself can bring you additional density benefit as well as power efficiency because you're moving to a smaller dimension. And in chip stacking, in addition to have our SOIC and CoWoS in for similar technology, we are also looking into the silicon photonics and also strong collaboration with the DRAM supplier to optimize their high bandwidth DRAM. And I think together there's still a lot of room we can move ahead. And for the high-performance computer, in my view, the computer right now is past the threshold and AI machine learning become useful. And that's why we see the boom of AI machine learning. So we just started the threshold point. And this boom in this opportunity, I think, will be bigger than PC or smartphone or any mobile device. That's how we look at the future of the whole industry. Well, it's incredibly exciting, Y.J., because what you just described is a very strong view about where we can take the industry forward with semiconductor technology, new devices, how the whole solution is put together. And that is in concert with what we're seeing is an insatiable demand for more computing, as you said, given now all these AI applications coming out. And so, you know, that to me is what is so exciting, is innovation is not dead in the semiconductor and packaging industry. And the advancements that you described couldn't be coming at a better time. It's right when the industry needs high-performance computing to accelerate more than ever in the history, I think, of our careers. Yes, yes. We are also very excited about that. We see a lot more opportunities coming. Well, that's great. Y.J., I cannot thank you enough for joining on this podcast, this video show. It's been just great to hear your insights. And again, your view of the future. We need this continued innovation. You've been delivering that from TSMC. And we're banking on that collaboration, that innovation with you going forward. Thank you. Thank you. I've truly enjoyed chatting with Y.J. Mii about our journey to 2 nanometer and beyond our semiconductor challenge nodes to continuing to scale devices, even as many said that Moore's Law is dead. It gives me tremendous confidence for our industry going forward to meet the insatiable demand that our customers have to us to drive computing to ever higher heights. Thank you for joining us today on Advanced Insights. I look forward to bringing you more in-depth conversations on cutting-edge technology, industry insights, and visionary perspectives from some of the brightest minds in the field. If you enjoyed today's discussion, don't forget to subscribe to AMD on YouTube and follow me on LinkedIn so you never miss a future episode. Technology is always advancing, and I can't wait to share more insights with you the next time.
