Sirk TV Spotlight: Ben Gu For “Millenium M1 Supercomputer” [Cadence]
Ben Gu, Vice President of Research & Development at Cadence Design Systems talks about design, fluid dynamics, practical application, impact and evolution in regards to their new Millenium M1 Supercomputer optimizing new AI chips.
Category
🤖
TechTranscript
00:00 [MUSIC]
00:10 We live in a world that is increasingly technologically intelligent and hyper-connected.
00:29 The electronic systems of tomorrow have redefined our way of life.
00:34 From autonomous driving and robotics to hyper-scale computing and 5G infrastructure.
00:40 These breakthroughs will shape the landscape of tomorrow.
00:44 All of this is made possible by Cadence's intelligent system design.
00:50 We're talking about with something like the Millennium AI supercomputer.
00:54 We're talking so many calculations, but can you talk about the aspect of the leap in the chip set
01:02 and what it means within the structure of the fluid dynamics, the computational fluid dynamics.
01:10 Could you talk about that in layman's terms so people understand what it is able to do?
01:15 Today, first of all, thank you for the question.
01:18 I'm joining you live from Levi Stadium, Santa Clara, California.
01:22 We're really excited to get ready for our big launch, Millennium M1 supercomputing platform for AI and for CAD.
01:33 You know, this Millennium M1 is going to transform everything we design today,
01:39 including computer, automobiles, and aircrafts, and the data centers.
01:45 The Millennium M1 will help engineers complete weeks of work in a matter of hours,
01:51 which will result in much better cars with much better fuel efficiency,
01:57 much less carbon footprint, and overall much better performance.
02:03 Could you talk about – we've seen stuff with McLaren, obviously, with race cars,
02:08 but are we talking about elements of aircraft in terms of supersonic airflow, all that kind of stuff?
02:15 Is it about figuring out what the stress points are?
02:18 Is that taken from a certain data set in the way this works?
02:25 Yes, well, Millennium M1 has specialized acceleration architecture
02:32 that we can do a lot of CFD simulations in a matter of hours.
02:36 That's much more accurate than the normal technology can do,
02:40 and that will accurately simulate all the phenomena, all the physics phenomena you mentioned,
02:45 the pressure points, turbulence, aeroacoustics, and so on.
02:50 That will give our engineers the accurate description, the accurate prediction of their design.
02:57 That's why we feel the technology is so revolutionary.
03:00 Cadence is a leading provider of computational software, hardware, and IP
03:07 for designing next-generation intelligent electronic systems.
03:11 The company has expanded from its early days of electronic design automation,
03:15 specifically for semiconductor design, to full solutions that address bigger and faster chips
03:20 and high-performance systems design for hyperscale computing, autonomous vehicles, 5G, AI/ML,
03:27 industrial IoT, intelligent medical devices, and so much more.
03:31 With nearly 10,000 employees globally, Cadence is built with a foundation
03:36 of some of the world's most creative engineers.
03:39 We're proud to be a year-after-year member of the 100 Best Places to Work.
03:43 With an accelerated revenue growth, expanded operating margin, combined with a strong balance sheet,
03:49 we are poised to continue our global leadership in advanced computational software technology
03:54 for tomorrow's biggest technological challenges.
03:56 You talk about how, you know, just looking at the sort of progression of the technology
04:01 and where we're going, where we've been, but what this allows for the human race right now.
04:07 In terms of, we're talking about different applications for both automotive and obviously aircraft,
04:13 but there's so many other things for communication, so many things.
04:17 Can you sort of talk about that and the wide-ranging possibilities of what this represents?
04:24 I think that we at Cadence, we are a computational software company.
04:32 We strongly believe computation can change everything, can make our life better.
04:37 We are working hard to make computational software can simulate everything in our physics.
04:43 You know, M1, it's a millennium, M1 is just one good example in our offering.
04:49 We believe that doing simulation can help our engineers to predict what can be designed,
04:56 so that can help us to make nearly everything better.
04:59 We talk about the automobiles, we talk about the aircraft, aerospace.
05:04 Of course, you know that lots of companies that design rockets will go to Mars.
05:08 We are really excited to be part of that effort.
05:11 Hopefully, we can send human race to Mars earlier.
05:15 And also, on the environment side, carbon neutrality is a big deal.
05:19 And you look at today, the AI wave is pushing GPU to be vastly adopted.
05:24 There's lots of electricity got used by data centers.
05:28 Had all the technology been used, all the data centers, that the power can be fully optimized.
05:34 We would have made a huge impact, huge impact on the carbon footprint reduction.
05:41 In other words, we believe, we are confident our technology can help human race,
05:47 can help our society reach the net carbon neutrality sooner.
05:52 Nearly every major technology trend relies on Cadence for world-class design software, hardware, and system analysis products.
05:59 From cloud computing and AI to aerospace, data science, and robotics.
06:04 Our customers can design and verify their enormously complex chip and system designs accurately and with a shortened time to market.
06:12 Silicon design is becoming exponentially more expensive and complex,
06:17 and the explosion of data and artificial intelligence require systems optimized for compute and high-speed communications.
06:24 Building on our computational software know-how, we're looking to the future by developing cutting-edge system analysis
06:31 and computational fluid dynamic solutions and improving design quality, safety, and security,
06:37 using machine learning as the systems of the future become more complex.
06:41 As far as AI, because obviously learning, it's about learning about seeing what the different data centers say.
06:47 Obviously, with your background, physics, research, and development,
06:51 what were the important milestones from your point of view in creating this supercomputer?
06:59 Because it takes time, it takes steps and steps, and we're all taking small steps so we can get to the next level.
07:06 Can you talk about that?
07:08 Sure. It's an amazing question. Thank you.
07:11 You compare, you look at the AI wave, it has been phenomenal in the Internet, in image recognition, voice recognition, and so on.
07:22 But you look at the AI application, at the scientific research, at the physics, at the mathematics,
07:27 it seems to be less revolutionary.
07:30 Why? It's because of the lack of data.
07:33 It's very expensive for us, for our engineering community, to come up with a huge dataset,
07:41 to be trained, to build a model on, and to be reinforced in the real application.
07:49 That's why this Millennium M1 supercomputer is so important,
07:54 because that can help our engineers to accelerate simulation,
07:59 to be able to generate the dataset much faster,
08:04 so that we can use the dataset to build an AI model,
08:08 and to inference the model to replace the future simulations.
08:14 That will help us adopt AI faster in all our applications.
08:20 We are a leading provider of computational software,
08:23 offering a truly end-to-end, full flow for chip design and system analysis.
08:28 And now, we are leveraging that position and technology know-how to expand into adjacent markets.
08:34 Cadence has a history of balancing top-line growth with strong profitability,
08:39 and of delivering values for our customers, shareholders, and partners.
08:43 Importantly, our revenue is widely diversified across multiple vertical markets.
08:48 Contracts are generally 2 to 3 years, and 85 to 90% of our revenue is recurring in nature.
08:54 We will continue to deliver world-class computational software for intelligent system design.
09:00 Do you have an example that you can give that may be a, not a prototype,
09:05 but a system where you've tested it on? I think McLaren is maybe one.
09:09 But can you talk about using this in a practical application for a company?
09:15 Can you speak about anything like that at this point? About what you guys have done?
09:20 Sure. Yeah, we have exciting results from working with Honda engineers.
09:26 Honda is using our software, hardware, Millennium M1 to get beautiful results about acceleration of the simulation.
09:37 And using our technology to simulate the Honda Jets and the Honda cars,
09:42 the results are very, very accurate and really fast.
09:45 They just love our results. They're very excited to work with us even more down the road.
09:51 You were talking about data sets. Has it just gotten to the point in the past couple of years
09:55 where we have this immense amount of data that we can use such computer to sort of,
10:01 it's about the cart before the horse and the horse before the cart in certain ways.
10:06 Is it just the fact that the data has gotten to such a level that the machine learning
10:11 and our intelligence has sort of caught up to what needs to be done?
10:15 I think that is exactly the point you're making here. It's like a chicken and egg problem.
10:19 We need to have lots of data first to be able to train the model and use the model to predict the future,
10:26 to understand the physics even more. So how to generate the data is a huge challenge
10:31 because it takes so much computation to generate data. And you need a super large data set,
10:36 typically maybe 100,000 data points. Each of the data points will take you hours or maybe days to generate.
10:44 Think about that. Putting them together is a huge, huge competition.
10:48 But using this millennium supercomputer, we are able to generate a super data set,
10:53 super large data set, much, much faster. That will, putting together, will accelerate our AI adoption
11:00 in physics simulations and in the design optimization.
11:06 That's why we are so excited about this technology.
11:09 We feel that it's going to revolutionize everything we design today.
11:13 One key to keeping costs down and increasing performance is the use of 3D IC technology
11:20 that allows heterogeneous chiplet integration, power reduction, and performance improvements.
11:26 With a massive abundance of data, new technologies are increasingly reliant on artificial intelligence and machine learning.
11:34 I just want to ask Apollo for you, because you've been through all the steps through,
11:39 what are you excited about looking forward at the evolution of what this could become?
11:46 Oh, I said this, we look forward to working with lots of customers.
11:51 We've already been working with Honda, and we're working with the other leading automobile companies, aircraft companies.
11:58 We believe this millennium one will become a platform to help our engineers to, number one, accelerate simulations.
12:06 Number two, to build an AI machine learning model.
12:09 Number three, use the AI machine learning model we build to inference, to replace the future simulations.
12:16 So one device, one platform, one hardware platform can do three jobs for our customers.
12:22 Simulation, simulation acceleration, model training, and the model inference.
12:29 It's going to be super exciting. We look forward to the bright future.
12:33 The advent of pervasive intelligence marks the fourth industrial revolution,
12:38 and Cadence's foundational strengths in computational software deliver us to the forefront of this revolution.
12:46 From the data center and hyperscalers to the edge, deep learning training and inferencing
12:53 allows vehicles to drive themselves, automate tasks, improve predictability, and preserve resources.
13:01 [MUSIC]
13:12 [MUSIC]