 Welcome, Mr. Nandan and Pelle to the stage. Good morning, I'm very happy to be here. Hopefully I said that right. So I'm really excited to be here. And I promise I'll lean the rest of my translation to the right people over there. So it's great to be back to come to Texas. And as you're aware, usually this day, day four, we launch new products. Especially in the premium IP suite. Now, we changed the script a little bit. So this year, about three days ago, we actually did the launch with a panel in San Francisco. However, this is still fresh. And there's a lot more to talk about the premium IP suite in context today. So as we talk about it, we'll also bring in some of information from the panelists that talked about it, about what you could do with the premium IP suite going forward. But before we start, let's look back at what Arm's been doing over the last year. So if you remember last year, this time, we were about nine months in to be a new soft bank company. Now it's about 20 months plus on. And what we focused on doing is changing our reach, changing our approach, and looking at the big picture from the ground up. So how do we make it easier for our partners? How do we make it easier for our ecosystem? And how do we make it easier for the developers? Because that's truly what we need to do to enable the next level of innovation and the next level of products and solutions coming out of Arm. We also started with a big picture with the little things like the logo. It's now lowercase arm to show that we consistently think about big and little at the same time. So along with this, we also delivered a lot of products that have relevance to what we're going to talk about today. As you're aware, last year we talked about Total Compute. And the concept of Total Compute is it's not just a CPU that computes, not the GPU. You need all kinds to solve the bigger picture problems. So the first thing we launched right now, I mean, a year ago on this stage was the next generation of arms premium processor IP, the Cortex-A75, Cortex-A55 with the new 8.2 architecture, the 8.2 architecture. But most importantly, with the new dynamic technology which gave it the next generation of big little capability, much more scalable. But also, it was the hub that would connect to all the other key components of compute for more comprehensive Total Compute, AI, et cetera. We married that with the Mali G72, which was a pretty big step up in performance and performance density for our graphics products. The Mali G72 was also pitched at delivering the kind of performance you expect for high fidelity gaming and also the R experiences. And you would see products based on these two processors and GPU combinations in the market by the end of this year. The third thing we announced, early part of this year, was our Project Trillium experience, which is to pull together all types of compute that are needed for machine learning and AI. Now, as we've said before, this completes the offering for compute that you need. You have, obviously, a lot of machine learning fat workloads still running on CPU, more happening on the GPU. But you also need specialized acceleration, object detection capability for vision. And that's what Project Trillium combines with products that are coming out in short order. So with this Total Compute platform set for premium, we also went the next step and said, how can we drive a lot of these experiences into mainstream and entry-level mobile? And to that effect, we delivered the Mali Suite for graphics G52, G31, alongside video and display products that help you get very similar high-end experiences on mainstream and entry-level devices. Now, that talks about compute. That talks about the capability to deliver new visual experiences. But one of the key things that has been tied together all the constraints for new next generation devices is security. So around tech con last year, which would be October, November, we released our security manifesto. This is not a new document. It's a working document that has been worked on for a long time, as you're aware. ARM talked about trust zone and launched trust zone technology nearly 15 years ago. Since then, we have delivered the next generation for IoT products or microcontroller products with trust zone for Cortex-M. We have continued our investments into secure core that help with tamper resistance and secure elements. So the security manifesto is pulling it all together into the big picture of how do you security not just per device, but for the new paradigm of IoT. Along with that, not just the manifesto, we released our platform security architecture plans. It is to deliver security that is fit for purpose and fit for solution. How do you analyze? How do you architect and how do you implement a platform security architecture that is scalable and then develops an ecosystem around it? And we talked about the principles for it and how we were going to do the architecture and specification. First for IoT, but we also had our platform security architecture for mobile, which will actually be pushing further. And then finally, we announced our products that are focused at preventing physical attacks, that is the Cortex-M35P, alongside a security enclave. So what you see are three events out of a lot of work that has been going on on security from ARM and ARM partners. What you don't see on this slide is I'm sure a set of questions that I will have coming to you later in the year, which is Spectre and Meltdown. Now, as you're aware, Spectre and Meltdown are a new class of side-channel attacks that attack high-end out-of-order processors that have speculative execution. As you're also probably aware, more than 90% of devices that have ARM processors in them are not affected by these side-to-class of attacks. The classes of processors that are just generally the high-end Cortex processors. And as you're also aware, we put out software updates for them, and also we're looking at how to put out the RTL or hardware changes for the affected processors. So I'm sure we'll discuss that later during the Q&A. But I just wanted to round out what we've been doing and dealing with on security. And then, of course, finally, as of three days ago, there was an announcement on the new premium mobile suite of processors, which we'll get to in a lot more detail later on. Now, obviously, the premium mobile suite and numerous of the other products that we've talked about are driven by the innovation that is emerging. So it's about, obviously, new experiences, augmented reality, virtual reality experiences. How do you actually visit a place without having been there and without leaving your couch? How can you visualize your room with new furniture that you want to bring? How can you visualize a new house that you're going to build? Alongside that, how do you actually train your device to know more about you? What can you include about your body, about your environment that can be actually using your devices to make better? So innovation is driving growth. It's driving growth at the consumer level in terms of devices. It is driving growth in terms of the infrastructure that needs to support it, whether it be networking, whether it be servers. How do you actually have automotive components to it? So there's a lot happening that is driving this growth. And that can be seen in the kind of success ARM has had over the last 22-plus years. If you look, end of 2017, we had about 120 million processor units shipped. End of 2017, in one year, we shipped about 21 billion units. And if you can see this, graph is still growing rapidly. And we talked about 120 billion units at the end of 2017. By the current rate, we're already over 130 billion units now. And so the vision of how do we get to a trillion units is not that far off. It's also that this growth is not limited to the IoT side. It is also seen around our mobile and consumer business as well. Our estimates, the silicon tan for mobile and consumer is about $55 billion right now. We expect, based on studies, that it will be about $77 billion in 2026. So within the next eight to nine years, we'll have gone up another 40%. It's about smartphones, it's out of tablets, and obviously large form factor devices. But it's also about everything that goes around it, not just the app civilization processing, but the connectivity, touch screen, sensing, et cetera. And for ARM and its partnership, it also means that there is more high-value processing capability, high-value solutions that go into it. Obviously, total compute, as I said, big little capability, graphics, and other machine learning type processing as well. So a lot of growth that we still expect from ARM and its partnership and overall for its ecosystem. And if you look at where that's going and why that's happening, the 5G transformation that we expect to hit towards the 2020 Olympics timeframe adds to it. There's a substantial jump in the connectivity levels in terms of the bandwidth that is possible, the latency that is possible, that makes different kinds of services, different kind of usage cases possible. As the devices become that much more capable, you can actually have even more immersive experiences. And we'll get into that in the later part of this discussion. Naturally, machine learning and artificial intelligence bring in another paradigm of capability, of software innovation, of new types of use cases, new types of expectations from consumers as well. One of the examples that we will talk about is also high-end gaming, which combines the level of computing, the level of connectivity, the level of intelligence, and how to make it more immersive. And then finally, security is important. It doesn't come for free. It is going to add to the computation and the growth required in how we deliver these solutions. So let's just touch on gaming as an example. If you go back about five years, the global gaming market was about $70 million. It had less than 20% was delivered on mobile. It was primarily a console gaming market and then, of course, PC. Fast forward to end of last year. The market grew nearly double, about $120 billion. That's roughly about a 70% growth. But what you see change is a huge amount of that is mobile. So mobile gets to about half of the gaming market end of last year. If you go five years ahead, we have $180 billion projection according to New Zoom. And nearly 60% of that, or about $100 billion, is in mobile. And obviously, a large part of the driver from mobile gaming, if you split it out, is Asia. 65% potentially growing. And you can see this is driving a lot of changes. One of those, of course, is that you might actually have specialized devices that focus on gaming and gaming type use cases. We already see ZTE, Xiaomi, Razer, introducing phones that are focused on a gaming experience. So as you can see, there is a lot of growth that we still see coming, because people are experimenting with new ways of delivering, consuming, and interacting with the user. So what do we need to make this next level of innovation possible? And what is interesting and what has been consistent is if you can make the experience unencumbered, if you can make it untethered, not connected, not needing a constant connection to a cooling system or power, it's useful. The more connected we get, the more data that goes in and out, that actually drives the experience further. And then, of course, how do you make that experience more immersive? So with that kind of set, let's go into the announcement that we made last week. So to deliver that kind of experience, you need a substantially greater level of compute performance. The CPU, we're delivering nearly laptop class performance, effectively laptop class performance, with the Cortex-A76. Cross-platform gaming-type experiences are needed, so the visuals and the graphical capability has to be higher, which is delivered by the Mali-G76. You need to actually have a much more visual experience. So Mali-V76 delivers the first real IP available that can support 8K ultra-high-definition content. We're working, as we talked about Project Trillium, to bring in the object detection as well as machine learning extensions. But most importantly, underlying all that, all of this suite is built with security in mind, and we consistently have to keep thinking about that. So as we step into the compute levels, let's first understand what we've been consistently trying to do. Moore's Law is something that has been doing well in the past decades about getting more compute in silicon faster, but it hasn't been scaling in deep sub-micron to deliver the extra performance in the same power budgets as before. And what ARM has consistently done is deliver better architectural performance generation to generation that sits over and above what process technology can offer. And as you can see consistently, especially over the last few years, we've been delivering nearly 20% year over year in terms of architectural performance. So when you marry that with what process technology gives us, that actually adds up to a lot more. And Cortex-A76 today will be delivering about 35% more performance than the last generation Cortex-A75 when compared. So let's just jump into that. The Cortex-A76, we believe in large screen form factors, will deliver about two times the performance for laptop class devices. We have a lot of demos here, and one of them shows a Windows and ARM-based device today. As you can see, it's already made a lot of differences because the thin form factors, 20-hour battery life, I think that device started yesterday. It's been working for 9 and 1 half hours, and the battery is kind of still half full. But compared to that device, A76-type devices will deliver twice the level of performance. It is certainly a largely micro-architecture-driven that focus on efficiency as well as game performance. It is also backward compatible in that it fits right into our dynamic technology and fits right in with the Cortex-A55 processors. So actually the systems can progress much quicker. And obviously the focus still is on efficiency, so battery life all day. So if we get into the numbers, 35% more performance compared to the Cortex-A75-based devices that you will see at the end of this year. 40% better power efficiency, a large part of that, again, coming from the architecture itself. Now, last year we talked about our version 8.2, the architecture, adding in machine learning instructions. What A76 does is with those instructions, it actually delivers about four times the level of machine learning workload performance as A75. So substantial growth in performance, general purpose, substantial improvement in efficiency, and especially targeted improvement in machine learning level of performance. Now you combine that into the CPU cluster. Obviously in seven nanometer, A76-class devices will get to three gigahertz and above. Especially married with the ARM, POP, Process Technology, and physical IP. We have mapped the Cortex-A75 to match, so the private level two cache is to match so that we can actually have a much more seamless, big little experience. The overall system gets better both in terms of the capacity of the level three cache, which actually helps with the overall performance and optimize memory systems to DRAM. This is not just us, but us working closely with our partners to deliver the best memory systems which will be needed for large screen compute devices, but also premium level smartphones. If you combine all that, as I mentioned, compared to A73-class devices that you see today in large screen form factors, you'll have over two times the performance in the same power budget delivered with the A76. We'll also see a very similar 2X type performance growth when you put it in context for big little. So combining all that, right, we have a very, very compelling jump in processor performance with this new IP suite. Now it's not just about CPU performance, it's about how do you deliver sustained compute as well as visual performance and sustained gaming kind of experience. And that's where the Mali G76 GPU comes in. So what you've been noticing, I hope, is now there are a lot more titles that are coming to the mobile world that were popular in the PC business before. So for example, Fortnite is now as popular and on mobile devices and growing. For that kind of experience, you do need a GPU that can match up and the Mali G76 is targeted to deliver exactly that. It also then adds to the total compute side by adding improvement in machine learning performance. So going to the numbers and this particular comparison is like for like on the same process node, same kind of performance levels, 30% more performance density with the Mali G76 compared to the G72, 30% better energy efficiency, also compared on the same process technology node. We're also improving the machine learning class of workload performance by nearly three acts, 2.7 to be precise. So if you look at where this is going, consistent improvement in performance while maintaining and improving efficiency and in particular focus on improving machine learning class workloads. Getting into a little bit more on what the Mali G76 is and what it does. It is still based on the Bifrost architecture as the G72 was. It still has three execution units per shader but we've doubled the number of lanes so it's much wider in terms of capability and hence that's actually where a lot of the density benefit comes from. Same amount of control but twice the data path. We've added dual texture mapping capability which was really needed for the next generation of premium devices. This is also where we think we can improve the high end capability by making it configurable all the way up to 20 shader cores which really improves the capacity for the high end. And then it is never enough to just work on execution unless you can't feed the data right. So we've actually made sure that we can feed this very capable, very wide machine with the right amount of data. So you have wider data paths into the machine. You have a lot of configurability into slices so that you can actually move the data in and out much more efficiently. But that's me just talking about our graphics capability. Let me let one of our key partners and lead partner in fact on Mali GPUs, Seokjoon Lee from Samsung. Talk about it a little more. Let's play the video. Hello, my name is Seokjoon Lee. I'm in charge of the SOC design team at Samsung's system LSA business. Smartphones transform our daily lives and drive current user experience innovations. Increasing performance by maintaining efficiency are key factors for this. For years, Samsung's system LSA business have worked closely with our to deliver the best graphics and gaming experiences. And the next generation GPU with improved GPU graphics and machine learning performance will enable new and exciting mobile gaming experience such as virtual, video, visual, and bring intelligent computing to consumer devices. We've got to the CPU part. We've talked about graphics. It wouldn't be complete if we didn't actually deliver it to screen. So at Mali V76, we introduced the first real IP product or IP processor that can support 8K content. It is about sharing not just the code part but also the encode part. So we can view and share ultra high definition content. And it's capable of doing encode and decode at 4K side by side. So you can do actually bi-directional streaming of high quality video. And more importantly, if you do two 4K decodes effectively 4K per eye, that is when you truly get motion sickness independent or independent for high resolution AR and VR capabilities. So numbers again, twice the decode performance of the prior generation V61 products. Now, consistent thing, you can't keep doing that by blowing up silicon area. So for a 4K 120 frame per second decode capability that the V61 had, if you do the same on the V76, it'll take about 40% lower silicon area. So we have to consistently do more with less silicon area. We're also focusing not just on the decode side but the encode side. And we have improved our encode capability about 25%. So effectively what we'd be saying is you can keep the same bit rate and get 25% better quality or keep the same quality and get 25% lower bit rates. So this is exactly the focus for how do we do better video and better visual processing. Now, having said that, again, by 2020, we are talking about 8K class screens being possible, especially around the Olympics. But really our focus is on how do you do decode well. So 8K 60 frame per second decode on Mali V76. Dual stream 4K, which really gives us high res AR VR kind of experiences. But you could actually then split it out and go into a much more expansive, you know, video wall type experience. If you did 1080p, you could get 16 simultaneous screens or if you did 2K, you'd get four simultaneous screens. So all of the above kind of are possible now with the Mali V76 processor. If you put it all together, the next generation platforms that you probably see coming out in the 2019 could combine the Cortex A76 alongside the A55, the very high end compute, Mali G76 and Mali V76 for better gaming, better visual, but don't forget, it's an overall solution. So if you look at Cortex RA, which we launched about a year or so ago, that's really the engine behind next generation 5G modems. What you will soon see is from our partners as well as from our own products around there, machine learning, processing coming in. The system IP that connects it all together is still around. And then finally importantly, there's software layers that are needed to make this happen. So the ARM NN is our neural net platform that plugs into the frameworks that are available today like Android NN, Cafe. And if you look over there at the demos around, you can see how actually it works together on devices today already, both on the Cortex A-class and the application smartphone side, but also as you can see on the sensor fusion or MCU type platforms with CMSs and NN and compute libraries. Look at the demos that show you what you can do with the always-on processors as well as broader application processors. We have the streamlined capabilities around our DS5 offering that brings in better debug and optimization for software. And then let's not forget the trusted boot and trusted firmware that is needed to secure all these devices. So combining all together, we have with our ecosystem a solution that's really needed for next generation devices that is capable of performance, very efficient, also very secure. So summarizing, we think for the next generation of innovation in devices that stretches from kind of the entry-level phone all the way up to large screen DTV type devices and user experience you need, we have three premium pieces of IP that can be connected together in the solution with laptop class performance, high-end gaming type experiences, and of course the 8K viewing experience. An arm along with its ecosystem is aligning to deliver that with the high level of connectivity that 5G brings as well as how we bring security to market in that space. So what we see today is improving and what we see tomorrow will be better. If you look at console-level graphics, that's what we're aiming for. You can see console-like graphics today at the demo stall that talks about what you can do on a Mali G72 class device. We're talking about personalization as a combination of all the compute that you have plus the specialized security that you can bring into that world while maintaining multi-day battery life and adding in the capabilities that allow for your device to be managed and secured whether it is in the physical world or the virtual world. So if we look at what was announced last week, you've kind of gone through it already. Last week we had Rene Haas, the president of IP groups actually doing this high-level presentation. But more importantly, we had a panel of experts that talked about what they were seeing for their customers and what kinds of device experiences they needed. To walk through them, Brian Meek, the CTO of Striver, which is a virtual reality training platform solution. We had Rahul Prasad, product manager from Google on the Daydream VR-type product lines. We had Matt Barlow who does Windows and devices and looking at what kinds of productivity devices and what kind of productive experiences you need. And we had Ralph Hover, a director of development for Unity who actually walked about what kinds of gaming experiences and what kind of immersive experiences are needed and what do you need for those. So I've spoken a lot already but let me let some of the panelists last week who unfortunately couldn't be here tell you more in their own works. Let's play the videos. Hi Taiwan friends, I'm Matt Barlow with Microsoft at Run Windows Marketing and I wish I could be there at Computex with you. You know, we really value our partnership with ARM and I think the benefits we're seeing from the ecosystem really fall into two really distinct areas. The first is connectivity. Being able to get LTE or Wi-Fi effortlessly and especially with 5G approaching rapidly, customers are gonna love the connectivity benefits they get from this device. And when you partner up those connectivity benefits with the long battery life, I mean up to 20 hours, up to 24 days of standby, you don't even have to have your power cord with you when you travel on a business trip. Those two things, connectivity and battery life with the partnership we have with Windows, with ARM really makes this device something everyone's gonna want to own. So hi to our friends in Taiwan. I'm Ralph Howard, head of platforms at Unity Technologies and I'm sorry we couldn't be there today. Unity represents the majority of content on mobile, AR and VR. And we find that most of our software then runs on ARM processors. So our partnership with ARM, what we can do there and what we can do together with ARM in terms of running it performately but also power efficiently is invaluable for both users as well as creators. Well as you can see, we are along with our partners, our developer community and the broader ecosystem and I hopefully you as well, quite excited by what this premium IP suite can offer. So that's all I had is a formal speech. I think we can open up for Q&A now. Thank you, Nandan. And we'll now open up the floor for the Q&A session. So if you have any questions, please put your hand up and our staff will hand you a microphone. We'll come back to the opening of the live chat and we'll give you the microphone that our media friends will give you. Hey Brian. Hey Nandan, thanks for this. So two questions of your mind. First of which on Trillium, it's just a fact check. Can you confirm whether any of your licenses are currently shipping and if not when do you expect that to happen? And secondly, I want to double click a bit on your point about laptop class performance. I think Renee was citing in the story that I saw, citing that being competitive with Core i7. I'm not sure if that was misquoted. Can you confirm or qualify that as well as talking about the corresponding impact on the power draw basically battery life. Okay. So first things first, in terms of the project Trillium, if you talk him on the acceleration and the object detection part. So some part of the object detection from the prior life are still in the market, right? But specifically project Trillium based products we haven't formally announced yet and our partners haven't formally announced yet. So I'm not going to presuppose any of that. Second piece was around Core i7, not Core i7. So I believe that is misquoted. So our focus is on the always connected type laptops but really we're talking about Core i3, Core i5 class performance points. I think what statement we had made was it is a scalable architecture and if you notice a large part of the higher numbers come from the amount of caching, the amount of memory system, et cetera, that is assigned with it. So there's the scalability aspect that is possible with it. However, our goal is the M3i3 i5 class which are the broader connected laptop key. I mean said that I also say we probably have our base that first truly connected laptops that have hit the market. I've impact on battery life. So all the numbers that we talked about still hold, right? So if we're still looking at multi-day battery life in that context. Thank you. Are there any other questions? Do you have any other questions? Please raise your hand. Our staff member is on the microphone. Hello, I would like to ask that the product was more on the active device. I know that there is a small box in the driver's car. Can you share with us the future plans of the driver's car? So the question is for the new products, it seems like they mostly focus on mobile devices. But what are ARM's strategies or plans for driverless cars in the future? Thank you. Okay, I'll take the jump. So if you look at driverless cars, there's various aspects to it, right? So you have the compute cluster around it. You have the sensing and how do you actually do real-time response to it. But underlying all that is functional safety. And if you look at what ARM has been doing over the last few years, we've upgraded our development environment on functional safety because it is all about from specification to delivery. How do you actually test what the specification of the product was and is it according to it? How do you handle full tolerance? So we've been focusing on that and we didn't talk much about it today because it's primarily a mobile and consumer device discussion rather than an automotive. But for driverless cars, again, it means all kinds of compute that we are actually focusing on. Every product that we're building has functional safety associated with it, whether it be high-end Cortex-A-Class devices that are going to work on the actual computation for ADAS. We have the real-time processors that are dealing with ADS, drivetrain, powertrain, and body. We're also looking at the microcontrollers that will be doing a lot of the sensing and response to that. So we believe that we have a very comprehensive set of solutions that will be supporting automotive and the driverless car's initiative. Thank you. So Ananda, you talked about Cortex-A76. Seven nanometers, and that was compared to about one hundred and thirty five nanometers. 10 nanometers, but can we make a comparison between a 75 7 nanometers and a 76 7 nanometers in terms of performance? Good question. So the other way I'd answer that question is if we took them on the same process technology, right, and kept the memory systems and everything in the same, we'd still see at least 20 to 25 percent increase varies on benchmarks in some benchmarks you can see high as high as 30 percent or higher on average you'll see about 25 percent also the efficiency on those parts is also substantial it won't go to 40 percent as we talked about going from 10 nanometers 7 but it will be better than parity. Thank you. Are there any other questions? I'd like to ask you a question about using mobile devices for machine learning. So what are your thoughts? Do you think it's really feasible? So let's also understand what machine learning is talking about, right? So we actually do believe that for truly artificial intelligence type experiences to happen you need to do more at the edge now you will still need to do quite a bit in the cloud but the more you can do on your device it's going to be important because one it is personalized, two it's secure because it is restricted to your device and certainly things like inferencing can actually be done on endpoint devices. There's obviously a lot more training class things which are much higher data sets much higher compute that will be done in the cloud but on a small scale you will actually see that happening closer in device and in fact what the devices here in the demo will show you is there's a lot you can do that is machine learning related on not just the Cortex-M smartphone class products but on Cortex-M and memory controller class products and it's going to be a combination of levels. So for example when you hear a visual or a verbal command and and it gets smarter and smarter those are done on always on type processors like the Cortex-M as you do bigger compute problems that are visual those will be done on the Cortex-A class devices but machine learning is a pretty large spectrum and you will see a lot more happening on the M device. Do you have any other questions from the floor? All right if not I'd like to thank Nandan so please give a round of applause to Mr. Nandan Ayampali. Thank you all thank you very much Thank you, thank you very much, thank you very much, thank you to our media friends Lili, Lili, Eba, Arm Computex for the exhibition.