Ladies and gentlemen, please welcome President and CEO, Rory Read.
Good morning. Welcome to AMD's Core Innovation Update Day. It's pretty exciting today. It's part of a long-term strategy we've embarked on. We're excited to see you all here. First of all, I'd like to welcome everyone on the web. Thank you for joining us here in San Francisco. There's no doubt AMD is transforming. We're building a different, a differentiated AMD. It's clearly based on a sound and strong strategy. We're focused on our core. Creating new opportunities in high-growth markets where we can bring our leadership IP and differentiate our solutions with the market makers who define those high-growth segments. Today is an important day. You'll begin to see us unfold a strategy we began 3 years ago. What a difference a year makes.
Look at how we've consistently executed, first with the game ramp, the next generation's graphics cores, our first-generation Arm-based server, which you'll see demoed today. Today is about innovation. It's about how we can shape a new future for AMD. When I joined AMD three years ago, 95% of our business was centered over PCs, a potentially declining market in an unhealthy duopoly. We've embarked on a strategy three years ago to create an ambidextrous architecture to unlock our leadership IP, our IP that can truly make a difference in this next generation of the cloud era. Taking our world-class graphics and our compute design capability to create solutions that allow the market makers to win. We embarked on this strategy. Now by the end of 2015, you'll see over 50% of our revenue come from these new high-growth segments.
We've delivered this year in 2014 leadership in the discrete graphics space. Look at the gaming ramp. Everyone had questions about that, but when you marry our IP and our design expertise from AMD with a market maker's understanding of the market and their ideas, we can truly create a differentiated solution that wins in the marketplace. Today, we're going to talk about that next step as we embark on our ambidextrous architecture, something we've been working on for more than two years. An architecture based on a SkyBridge technology that connects our technology on a single SoC. We've got the best talent and IP in the industry. We're beginning to unlock the true value of this company by creating those strategic partnerships that truly make a difference in the marketplace. Today is a big day.
It's an important day, it's only a step in the journey we began three years ago. You see the tangible progress, the consistent execution, and better performance, and you'll see even more over the next two years. This is an important day. Look at what we've done here today, and as we show you what's going to happen over the next two years, it's truly an exciting time for AMD. With that, what I'd like to do is I'd like to invite up our Senior Vice President, head of all our business units, Dr. Lisa Su. Dr. Su?
Thanks, Rory. All right. Good morning, everyone.
Good morning.
It's really a pleasure to be here today, as Rory said, it's an important day for us. What I'd like to do is really lay out a bit more about our strategy. Ambidextrous for us, it's about markets, it's about vision, and today we're going to show you some proof points about execution. Let's start first with what's happening in the market. Many of you know this, we've had many conversations about what's happening in the market. It starts with the consumer. It starts with the incredible explosion of devices, whether you're talking about smartphones or tablets or two-in-ones or PCs or your home set-top boxes. Tremendous amount of technology adoption, all of it with more and more data expanding and diversifying the ecosystem. With that, it also changes fundamentally the data center.
When you look at the data center and when you look at embedded applications, if you look at all of the data that we have to move and be available anytime, anywhere, any place, that's yet another challenge that we have in the technology industry. The third point is, as Rory said, it's all about our customers. It's about how do we help our customers take all of this consumer and data center and environments and integrate them to form differentiated solutions. Those are the things that are shaping the marketplace today. Now, let's talk a little bit about silicon architectures. This is an important chart because this is really one of the underpinnings of why ambidextrous is so important to us. When you look at this is looking at revenue share of silicon architectures across, let's call it, a 15-year period.
If you look backward-looking, you can see that x86 is a very substantial piece of the market. About 50% revenue share. This is considering combined revenue across mobile, embedded, PCs, consoles, and servers. Clearly that's there. If you look in the purple, that's the Arm architecture. When you see it starting really as a mobile architecture, you see the tremendous growth over the last five to seven years and the tremendous projected growth going forward. Then in the gray, you see all other architectures. These are Power, MIPS, proprietary architectures that were very popular in the past when you had a lot of diverse, disconnected applications. You see that the number of proprietary architectures and other architectures is going down over time. What's the punchline here?
The punchline here is if you add up the x86 and the Arm TAM over the next number of years, this is where all of the silicon is going to be built. These are the dominant architectures. 80% of the revenue share, almost $80 billion in TAM in 2014, growing throughout the foreseeable future. This is the place where we bring ambidextrous together for the market because AMD is the only company that can bridge Arm and the x86 ecosystems. Now when you look at what does that mean for the compute market and what are we trying to accomplish, it's really about three things. It's about growth and compute, it's going to continue to grow.
x86 and Arm are going to continue to service those markets, we're going to continue to provide more and more compute for all applications, whether you talk about client applications, whether you talk about server applications, whether you talk about network and storage applications, or you're talking about the embedded home, automotive, all of those drive more compute. What's also key, and Rory talked about it, is it's the graphics and the acceleration capability. When you think about what's driving all that content, it's not just words, right? It's really visualization. It's graphics. It's all of the video content. It's being able to have your data on the go, whatever you want to look at. That is key as well in terms of the overall total compute message. Then the third point is the merging of ecosystems. It's not just about silicon. It's about software. It's about applications.
It's about how do we put those ecosystems together? Ambidextrous for us is really bridging the x86 and the Arm ecosystem. As we look at this, we really want to lay out for you our vision around ambidextrous computing. There are really three key components. The first centers around our customers and really helping our customers innovate and provide new areas and new ways of delivering their services and capabilities. The second is about choice. Okay? Our customers, everybody wants choice. You want to marry the best silicon with the best ecosystem. It's not about who's going to dominate. It's about we're all going to innovate, and it's about harnessing the best of the x86 and the Arm global ecosystems. The third point is about technology. We're a technology company. We're an innovation company.
Thank you for all of our customers and enthusiasts who have really adopted this in both the desktop channel and soon in other devices. Last week, we just announced Beema and Mullins. This is the follow-on to our system-on-chip architecture that integrates on a single chip very power-efficient x86 and graphics compute together for our mainstream and low-power applications. The first four or five months have been a good year for us at AMD. What we have today is really another major milestone for us, and Rory mentioned it earlier. We set out a couple of years ago to ensure that as we continue to evolve our core markets, we also build the bridge to the future, and the bridge to the future is offering both Arm and x86 in our portfolio. we started sampling our first 64-bit Arm chip, codenamed Seattle, earlier in the year.
This is special for a couple of reasons. We chose this because we believe that there's a fundamental disruption in the server market happening over the next few years. When you look at all of the data that needs to be moved through both web applications and in the network, it's a tremendous opportunity for us to offer choice in different sets of applications. We chose server as the place to really demonstrate our 64-bit Arm capability, given our server heritage, and it's been a very exciting road for us. Let me tell you a little bit about Seattle. It's unique in the sense that it's the first 28-nm server part in the industry that's sampling. that's a great thing. it's also using a57 cores from our partners at Arm.
This is special for a couple of reasons. We chose this because we believe that there's a fundamental disruption in the server market happening over the next few years. When you look at all of the data that needs to be moved through both web applications and in the network, it's a tremendous opportunity for us to offer choice in different sets of applications. We chose server as the place to really demonstrate our 64-bit Arm capability, given our server heritage, and it's been a very exciting road for us. Let me tell you a little bit about Seattle. It's unique in the sense that it's the first 28-nm server part in the industry that's sampling. That's a great thing. It's also using A57 cores from our partners at Arm.
We chose to license A57 cores for this first implementation of 64-bit Arm because we thought that was the best way to get time to market. The key for us with Arm is to help accelerate that ecosystem and get 64-bit silicon into the market as soon as possible, and we do believe that we've accomplished that goal. We've also done something very important, and it's about differentiation, but it's also about standards. Seattle is fully Arm SBSA compliant. Why is that important? The reason it's important is because when you think about what's so attractive about Arm for servers, it's that we will have multiple partners who are able to offer different differentiation, different acceleration, different packaging for that market. What's also important is we need the software ecosystem to be there.
The software ecosystem is that each of us independent Arm vendors have to adhere to some standards so that we get very good collaboration in the ecosystem. We're really proud of Seattle, and we're going to demonstrate here for the first time today, the public demonstration. Let me invite Suresh Gopalakrishnan, our VP and GM of servers, to come show it. Suresh, come on up.
Thank you, Lisa.
How are you doing today, Suresh?
Great. This is a super exciting day for the whole team at AMD who've worked on the processor, the development board, and the software that makes all of these things work.
Tell them what you got here.
We are super excited. One more second. We have a couple of members of the team here.
Absolutely
Who are going to help me with the demo. Let's see. We have a Seattle server here. We are going to use it for the demo. This is the development board that we are sampling for customers at this point. Before I kick off the demo, let me set a little bit of context on what we are going to show. On the software front, we will demonstrate the LAMP stack running on this server. For the folks who are not familiar with the LAMP stack, it is the software backbone for the internet. We have the Linux operating system, we have Apache web serving software, we have MySQL, which is a database for storing information on the server. Finally, the programming language PHP, that is used to create web pages dynamically.
If you go to Facebook, the page you see is going to be delivered to you using a LAMP stack. Seattle, because of the type of integration we have on it, is capable of doing web hosting, big data applications, a variety of storage applications, and so on. For this particular demo, I'm just going to focus on web hosting. Let's get that started. Right. We have the entire LAMP stack running on this server. We are going to use that LAMP stack to gather information of the processor, the various processes running on this server and present it to you using the LAMP stack as web pages. This page, as well as pages that are going to follow, are dynamically generated as the demo progresses. This is what you would typically see when you go to a website.
Content will be generated for you. What's the processor architecture that we are using today? We're going to execute this particular command on the server, grab the information, and then use the LAMP stack to give you that answer back. We are using AArch64, which is the 64-bit Arm architecture. What about the operating system? Again, execute the command, grab the information. We are using Red Hat Linux for this demo. The web server, we are using Apache for 64-bit Arm server. The database is going to come up next. It is MySQL, again for 64-bit Arm architecture. Finally, the programming language I talked about is going to be PHP for 64-bit Arm again. So far we have seen the LAMP stack being used to grab information from the server and create web pages on the fly for you to see.
Let's take a look at a couple more popular web applications, blogging and video serving. Let's go to the blogging page there. What we have here is the application WordPress running on Seattle. Why is this important? Why is this exciting? Why is this interesting? WordPress is the world's most popular blogging software. It powers over more than 60 million websites and blogs around the world. Now it is running on Seattle. It's a very important step in putting together a web hosting solution. What you have on the screen is a welcome blog that the demo team has published already. Hey, how about we do another blog?
Hey, don't say anything bad though, guys.
No.
Oh, it says we will be available to answer questions anytime after the demo.
Yeah, absolutely.
All right. That's good.
At the end of it, we'll stay around. You can come check out the hardware and software. All these posts are being uploaded onto the server live, and then it's being broadcast out for the world to see. Now to the final part of the demo, video. Video is becoming extremely important for web hosting. You have YouTube, you have Vimeo, you have a variety of paid services, all of which means that video is very important for all of us and for the web hosting community. Let's take a look at the video being served out of the Seattle server. Let's go.
We're dreamers, innovators, pioneers. Together, we are AMD. What we do is much bigger than technology. We're driven to make our industry a better place. We envision the future users want and turn those dreams into reality. We're bold, imaginative, unconstrained by convention. It's in our DNA. Together, we pioneer technology that frees people to push the limits of what is possible, because we believe that when you give people the freedom to innovate today, amazing things will happen tomorrow. At AMD, we design, we integrate, and we innovate. We do this every day, and you make it happen. We are AMD. Enabling today, inspiring tomorrow.
All right. What you saw was the entire LAMP stack running on Seattle being used to serve dynamic web pages, to host a WordPress blog, and finally to serve video. A complete web hosting solution running on 64-bit Arm, a complete web hosting solution running on Seattle. Thank you.
Suresh, fantastic.
All right. Thank you.
Fantastic partner. We also want to thank Jon Masters here in the audience from Red Hat, who is the Chief Arm Architect, who really helped us a lot, both with the demo as well as the ecosystem around Arm. How many people still doubt that Arm's going to be in servers? I think this shows you what we're capable of. It's not just about technology, it's about what we do for our customers. Next, I'd like to invite one of our key customer partners, Paul Santeler from HP, who's the Vice President and General Manager of the Hyperscale Business Unit, to talk about his perspective of Arm and servers. Paul?
Hi, Lisa. Good to see you.
Good to see you.
Great video. I'd love to jump off a building like that.
Paul, thank you so much for joining us. We really appreciate it, and thank you for HP's partnership with AMD. I know that there's so much going on in the server market today. Tell us a little bit about what you're hearing from customers and what they want you to help them with.
Sure. Well, customers certainly have challenges, but I think customers also have opportunities. If you look at the market now, with mobility, with big data, with the social media, and everything that's going on, our lives are changing. Most customers see this as an opportunity to grow their businesses, to expand their business, to go after new customers, to provide new services. It's an exciting opportunity. It's also a competitive leverage. Customers who are able to use advanced technology can really get a jump on their competitors and do things better, faster. I'll give you an example of a market expansion. I have a customer that last year purchased a multimillion-dollar Hadoop cluster for doing data mining of people that came to their website.
I'm sure you all know, you go to a website and it says, "Oh, people who have bought this have also purchased this, and this." If you're buying a vacuum cleaner, you probably want some extra bags. If you're buying a grill, you need something to clean the grill and all that, right? They were actually able to make that multimillion-dollar investment, and they paid that investment back in new revenue growth in five months.
Wow.
It's pretty amazing. From a challenges standpoint, I think customers still have the challenges that they've always had, and that is how do you keep up with IT? How do you keep up with the pace of change? Then also, for this great demand for computing, how do we really provide the compute that takes it, that drives that? If you think of data centers, I was just in London last week or a week before, and I was talking to a customer that was providing a software as a service. He was rolling out this new application. His biggest challenge was where do I find the data center space and where do I get the power from? Because in these large cities and large areas, they don't have enough power to be able to drive these infrastructures.
Power and space savings are key, which is part of the reason that we're partnering together on the whole HP Moonshot front.
Yeah, no, absolutely.
Exciting opportunities.
Well, traditionally, the server market has been viewed as somewhat slow to change, but it sounds like that's not what you're seeing at all.
Not at all.
That's cool. That's cool. We just ran Seattle. We're very proud of it, and we really look at it as an opportunity for new workloads. What are you seeing as the workloads that will drive 64-bit Arm?
I think it's going to be interesting. I think that you're going to see Arm really focus in in these new data center spaces. I think there's core things that have to happen, and I think you talked about them earlier, really to be able to make good on that promise, right? It's all about volume economics, it's all about the hardware ecosystem, it's all about the software ecosystem. The volume economics, most people think that Arm is pretty new to the game here, right? It's a brand-new architecture. If you look at it, Arm's been around overall in devices for a long time. There's not too many devices that you can touch that don't have Arm in them. Appliances, TVs, smart thermostats, printers. Kindle, you name it, and it's probably got an Arm device in it.
That volume economic provides that ability to grow the technology and continue to innovate, and at the same time, provide that core capability. From hardware ecosystems, I think what you're doing is really exciting here. You can take the Arm architecture and then design around it, but as you said earlier, it still gives everyone that basic core architecture that's compatible across the board, and that's really key because you have to be able to drive and create that LAMP stacks, as you talked about.
Yes.
That's the third one, which is a software ecosystem. I think you nailed them earlier. It's hardware, software, and volume economics.
Yeah. Do you see other dynamics that will either accelerate or help the 64-bit Arm in dense server space?
I think the dense server space, especially so because if you look at this dense server and this scale-out space, it's very different than the enterprise space. Typical enterprise is run in hundreds, if not thousands of applications, multiple versions of multiple operating systems across multiple architectures, and they've got to make that all run together. If you look at these next generation of scale-out data centers that are running all of the next generation architectures that we have, they're very homogeneous in how they build it. They have multiple application tiers. They run it, and they get that structured, then they scale it out as they need more capacity and more users. That allows them to have a whole infrastructure and have that core expertise. They have software developers, they have people that know all the different technologies, and they adopt new technologies pretty quickly, too.
If you look at all the next generation technologies are actually pushed faster in these scale-out data centers, things like flash memory, DC power, scale-out databases, open source. They're the ones that are driving that technology first because they need that capability, and they can see an immediate investment. I think you'll find that these scale-out data centers will probably be the on-ramp for the adoption of Arm 64 in the marketplace.
Absolutely. We're going to talk quite a bit about what we're doing in ambidextrous. What does ambidextrous mean for you? When AMD talks about ambidextrous, how does HP view that?
Well, I think we're really excited. I think you really showed it with a chart there that showed there are going to be two main operating platforms. There's going to be two instruction sets. There's going to be x86, and there's going to be Arm. That's pretty much going to be cast in, and it's a pretty fair fight. I think by AMD straddling both and being able to design the best overall systems and the best overall devices, being able to use both of those interchangeably, I think gives you the opportunity to really have the right solution next. I'm pretty excited about it, and I look forward to seeing what the development brings.
Fantastic. Paul, thank you so much. Thank you for your partnership.
You're welcome.
Thank you for HP's partnership, and thanks for being here today.
You're welcome. Thanks.
Technology is great, but it's always important to keep the customer in front of us, and I think Paul gave you a glimpse of why dense servers are so important and where Arm can really play in the market. Not saying that x86 is extraordinarily important, it is, but there's a new opportunity for disruption here with Arm. With that, though, I want to take this one step further. When we had talked about ambidextrous, it was really a very broad statement for AMD. It wasn't just a server statement, it wasn't just a Seattle statement. It's really, what do we see going forward in the future? For 2015, and Rory gave you a little bit of inkling for this, we're really trying to bring in new markets that can benefit from ambidextrous computing.
Today, for the first time, we are announcing a new family of products that we call Project SkyBridge. Project SkyBridge is really the very first time that you will be able to see x86 and Arm SoCs that are put together in such a way that they can be pin-compatible. You can have a motherboard that runs Arm, you can have a motherboard that runs x86. This is something that we believe adds tremendous flexibility in the market for our customers. What is Project SkyBridge? We're very excited about it. It really is a design framework. It's a family of products that we will be putting out, starting first in 20 nm technology. Again, for the first time, we're adding APUs that are designed with our Graphics Core Next technology with full Heterogeneous System Architecture support.
On the x86 side, we're using next generation Puma+ cores. We just announced Puma in the Beema and Mullins family last week. Next year, you'll see Puma+ in our Project SkyBridge family of APUs on the x86 side. Then on the Arm side, we're going to optimize 64-bit A57, again, in the same footprint that we have our x86 capability. This will also be the first Android platform that we have at AMD with our Heterogeneous System Architecture. The power of this is it's about a family. Where are you going to see this? You're going to see this in embedded markets. You're going to see this in some client markets. You're going to see it across a variety of marketplaces. What it is it's an opportunity for us to help customers innovate, differentiate, and also reduce their time to market.
Let me just give you a couple of customer examples so you can see what we're trying to accomplish here. I chose to do a couple of embedded examples because embedded is a market that perhaps is not as familiar to everybody in the audience. Let's start first with the industrial customer example. What are we trying to do here? This is a real case study. If you look at some of the large industrial customers, they really have a variety of products. At the high end, they typically use x86 for motion control systems, for HMI panels at the high end because they need the performance that x86 offers. At the low end, they have Arm SoCs because they're the right price point, the right power point, and there's a lot of diversity there.
What we can do with an ambidextrous strategy is we can work with that customer across that entire performance range with either Arm or x86, depending on what legacy software they're trying to bring along and what applications they're trying to service in the future. This is something very unique to what we can do at AMD. It's really about addressing multiple applications across the family, optimizing hardware and software investment. You may or may not use the same motherboard, depending on where your price points are, but you have the capability to reuse components both at the hardware and the software level. Very interesting case. You look at networking. Networking is a little bit different, but also somewhat similar. If you look at some of the big networking customers today, they have a variety of architectures.
The way those architectures got in there was it was what was best at a certain point in time. You may have some MIPS, you may have some PowerPC, you may have some x86 for the high-end control plane. You'll have some Arm that's coming in at the low end. If you look at trying to manage that code base and that capability, it's just a lot to try to manage. When we look at what we can do over the next few years with our ambidextrous capability, it's really about simplifying that customer use case by allowing them to really use the best architecture, the best silicon for a given software ecosystem and helping unify some of those other architectures. When you remember that market chart, why don't we see x86 and Arm growing?
It's just way too expensive to support all of those disparate architectures across a single ecosystem or several ecosystems. That gives you a little bit of the customer context of what we're trying to achieve over the next couple of years with Project SkyBridge and our ambidextrous strategy. Exciting stuff. There's one final piece of the puzzle which you should expect from us because we are a computing company, and it's really about our methodical way of putting together our ambidextrous strategy. As we go into 2016 and beyond, integrating everything that we've took in terms of the learnings of Seattle, in terms of the learning with our pin-compatible family of parts, the very key piece of differentiation for us is really around developing our own Arm cores.
Today, we are also announcing that we are an Arm architectural licensee, and we are well underway with developing our own Arm cores. With that, let me invite Simon Segars, CEO of Arm, to come up and talk to us a little bit about that development. Simon?
Lisa.
Hi, Simon. How are you?
I'm good, thanks.
Thank you so much for joining us.
You're welcome.
You're on many coasts these days, I hear.
It does feel like I live on an airplane right now.
Yes. Look, Arm has been incredibly successful in many markets, in mobile and others. Tell us, where do you see the largest growth for Arm over the next few years?
Well, we see it in many different dimensions. Mobile is going to continue to grow. Mobile is a really interesting space right now. People think the smartphone market is saturated, but there's only about 2 billion of them in use right now. There's about 7 billion people on the planet. That looks to me like a pretty big opportunity, and it is happening with the advent of low-cost devices. What they are doing is generating a whole ton of data that needs moving around and processing in the cloud. To do that efficiently is going to require a lot of innovation in the devices that go into the network infrastructure and into the servers. We are really excited about what you're doing here to bring innovation into those spaces.
Are you seeing a lot of uptake in both embedded and data center? What is your customer conversation telling you?
Oh, yeah. Across the board, in embedded, in networking, we're seeing a lot of adoption of Arm technology. In the server space, people are really waiting for high-performance, 64-bit implementations of Arm CPUs, to really make it a reality. We've been working on things like SBSA, we've been working on the software infrastructure, and now as the silicon is starting to come along, we'll start to see these servers deployed and become a reality. It's exciting times.
Fantastic. We've been working for some time on this Arm architectural license. I know we're happy to be able to talk about it finally.
So are we.
What do you think is the significance of that for the Arm ecosystem?
Well, Arm is all about creating building blocks. It's about licensing the technology out so that people who are experts in particular vertical markets can leverage all the infrastructure that goes into the ecosystem and bring innovation. We aren't about dictating what the answer is. We're about enabling choice. We're about enabling innovation through the entire supply chain. What you've already done with the Arm architecture and what it looks like you're going to do with it plays exactly to our strengths, and brings continued innovation in the Arm world. Really looking forward to seeing this hardware deployed.
You think Seattle will go a bit?
Yeah, looks pretty interesting.
When we talk about collaboration with Arm and there's also the competitive nature. People ask us all the time, "How do you manage this ecosystem that you have?" There's so many different competitors as well as collaborators.
What are you guys doing to manage the ecosystem?
Well, we're trying to create an environment where everyone can innovate using their own skills, their own experience. AMD has a huge heritage in servers, which makes your company such an interesting partner for us to work with to bring all of our expertise to bear. We aren't trying to pick winners. We're trying to partner with everyone, respect everyone's confidentiality, and create an environment where i nefficiencies can be kind of wrung out of the system. All the work that's gone on in the software space means that people can come together and focus on solving a problem once, solving it well, and then everybody leveraging it so they can spend their R&D innovating on top. I think that's one of the most exciting things.
I think that's the key. That's the key for us also, is really leveraging all of the strength of the ecosystem.
Also enabling us to differentiate on sort of the compute capability and the integration-
Yeah.
...that we have with our other IP. When you think about other things that are important for Arm to be successful in servers, what is the focus from your standpoint?
Well, it's much more than just the CPU. It may surprise you to hear me say that. The CPU is really important. Connecting the CPU up, all the IO, all of the fabric that connects all of these processes together, the accelerators, the GPUs on the chip, all of that has to come together. It has to be delivered in a silicon package at low power. There is much more to, as you know because you do it, to putting one of these chips together than just licensing some IP and hooking it up. I think that is often underestimated by a lot of people. It is the expertise of semiconductor companies and the strength that you have in this space, I think is really going to pay dividends.
Yeah. No, I think what we would like to do is really bring out some very unique both core technology as well as integration with system on chips to do that.
In terms of what today's news means, we've talked about the architectural license, we've talked about some of the things that we're doing with Project SkyBridge.
What does it mean from your standpoint?
We're all about working with leaders, and we're all about sharing knowledge, sharing innovation, and trying to drive all technology forward. Our partnership's really important because through working together, we've been able to share experiences, share knowledge, and some of that comes back into the Arm architecture that makes all technology better. It's really important for us to maintain this open relationship that we've had, and it's been great working together over the last few years, and we've known each other a long time. It really does help keep Arm moving forwards and keeping the partnership strong.
I think what's really been meaningful to us is, I know when Rory, Mark, and I started the discussion with Simon, you and your team, it was about how we could really marry the best of both worlds, x86 and Arm.
We are an x86 company. We're now adding Arm and our ambidextrous capability. Arm has just been extraordinarily supportive and flexible for what we're trying to do. Really, thank you for that.
Well, thank you. I think this pin-compatible x86 and Arm devices, that's a first anywhere, obviously. It's going to be really interesting to see how that does change the market in the way you've described.
Yeah. No, we're really excited about what it can do for customers. Very good. Simon, thank you so much.
Thank you.
Really appreciate your partnership, thanks for joining us today.
Thank you. My pleasure.
All right, you've got the picture now. If I just go over what I said to you when we started, is we started with a vision, but I wanted to make sure that we left with an execution roadmap of proof points of how we get there. In 2014, it's Seattle, and I got to say, I'm extraordinarily proud of what the team has been able to do with 64-bit Arm in Seattle. In 2015, it's Project SkyBridge, and this is truly just the beginning. It's the beginning of a family of products that will offer customers new flexibility, new capabilities. In 2016 and beyond, it's about how do we really innovate in our compute cores by bringing Arm and x86 to the market in our technology. The other thing is we look at ambidextrous computing across a variety of markets.
Rory talked about the five markets that we think are growth markets. Dense server, embedded, semi-custom, ultra-low power client, professional graphics, they're all incredibly important markets to us. Four of those will be extraordinarily well-served by our ambidextrous x86 and 64-bit Arm story. With that, I'll leave you with this. We are executing our vision. It is a happy day for us. It has been several years in the making. It is about a comprehensive roadmap. This is really just the beginning of what we are going to offer in this space. It is about leveraging our strengths, our strengths in both computing cores as well as graphics and heterogeneous compute. We believe in the total compute story, and it's just going to expand with our ambidextrous vision.
We believe very strongly that it's about the customer, and it's what we can do to enable our customers to do something very unique and differentiated in the market. With that, I'd like to bring up my friend and colleague, Mark Papermaster, who's our Chief Technology Officer, to talk to you about how we actually bring this vision into technical execution. Mark?
Thank you, Lisa. It really is an exciting day. You heard Lisa and Simon talking. We've been waiting to unveil this and talk about how we've implemented this ambidextrous strategy. Our teams at AMD are pumped up. They've been working on this. They've been working on it a while. I want to talk a little bit about how we've done this. We talked, it was an analyst briefing that we had over two years ago, we talked about creating this kind of capability that would bring flexibility to AMD, flexibility to how we could bring our IP to market, flexibility, we said at the time that we were going to think through how to in fact add an ISA, to change our implementation so it wasn't tied solely to the x86 that we've got such a strong and rich history.
That's what we've been doing. We've been hard at work. We've been executing. Just like you see the cadence of getting our products out on time, out at the quality that we promised, out with the capability we promised, we've done the same thing.
With the underpinnings. The underpinnings of how we get our IP and our SoCs to market. What we've done is really re-engineered. Re-engineered from top to bottom about how we go about this. It starts with really the design point at which you create each of your IP blocks. How do you create it to be a true IP so it can be reusable, so it can be put together with more flexibility? How do you connect it with the fabric, a network on chip that doesn't care if you're x86 or Arm? An ambidextrous network on chip to be able to put all of these pieces together, our IP of our partners, IP of a third party. It's been a lot of hard work, and we're really excited.
It starts with Project SkyBridge that you just heard about as we roll it out, again, it lays the underpinnings for all of our ambidextrous designs going forward. We're very excited, again, we took advantage of this to simplify. You look at how we put this together, it's a hard-focused engineering discipline that simplifies our interfaces and allows us to reduce complexity in how we go about bringing these products to market. This modularity, this sharing has been key, and we built it in all of our IP blocks. It starts at the center. It starts with those, we call it those crown jewel cores we have. Thrilled to be adding Arm to this complement. I told you about how we went about this, how we architected in this capability.
You'll hear from Jim Keller in just a few moments about how they've gone about it from the core team. How they've been very smart in creating that capability. We've done it in such a way that it marries. It marries with all the GPU, the graphics and compute capability in our GPU, with all of our multimedia, I/O memory, and all of the software that we do because as actually as Simon mentioned a moment ago, we're focused at AMD about delivering solutions. It's about that differentiating IP. It's about putting it together in a very flexible way, and it's about delivering solutions at the end of the day that matter to our customers. That's what we've been doing. We want to go about this in a very thoughtful way. In fact, it's a stepwise progression that we've went on.
Today, you saw the first implementation, again, right on schedule from our first announcement, we said that we would be demoing an Arm capability targeted to servers, and we'd be shipping that in 2014. Demoing in the first half, shipping in the second half. What did that do for us from an engineering standpoint? It was the first focus on developing that ecosystem. That's what you heard about from Suresh. Developing that software ecosystem for servers around Seattle processor. Next, for 2015, we're right on track in terms of creating a methodology that allows us to do a deeper tuning, create more of a power efficiency into that Cortex-A57 core. Then marrying that with the design, with the heterogeneity, bringing that graphics to bear, and in the cores methodology, bringing up that Arm ecosystem in tandem with our x86 design.
Right on track for delivery in 2015, again, you're seeing that in its implementation of that flexibility it gives our customers with Project SkyBridge. That sets the stage for, in fact, our third element. The third element is what we're really enthused about because it is a from-scratch, grounds-up, optimized design of the Arm 64-bit architecture, marrying all of that deep expertise we have at AMD. We developed the first 64-bit x86 processor. We delivered that with the Hammer technology, the Opteron. It is that team that we put together, again, you'll get a chance to talk to Jim in a moment with a little bit more detail. It's that team that's pulled that expertise together, and we believe that we can extend the Arm value proposition, marrying that AMD expertise with a very efficient Arm architecture. It's a beautiful combination.
It's a beautiful combination. Of course, we're not backing off x86. That's the beauty of this ambidextrous approach. We have a long legacy of innovation. We power forward in x86, but we bring that expertise with this efficient Arm architecture. That's what today is about, sharing with you some of that detail. People ask, well, why do we do this? Of course, you heard from Lisa, there's a very strong business rationale. The TAM, the market opportunity is huge. Of course, that's the compelling factor that's driven us to this move. I will tell you, there's actually a much broader Arm value proposition that we're leveraging. First, the ecosystem. When you bring together many companies, we'll all compete on the merits of our IP. We'll all compete on how we piece this together.
What an Arm ecosystem does is it adds economy of scale of creating an ecosystem, an infrastructure to deliver solutions. It lowers the cost for everyone to deliver your differentiation to those customers. That's key piece number one. Secondly, it's the Arm architecture itself. There is efficiencies. It's well-known. It's been utilized for years in the 32-bit space, bringing incredibly efficient solutions to a multitude of devices, and we're leveraging that efficiency as we transcend to 64-bit computation capability. We're leveraging aspects of the architecture, register files, and decode aspects that you'll see inherent in Arm. We leverage this to its advantage in our 64-bit optimized solutions. Again, I mentioned it earlier, but very interesting to us, very exciting for us is leveraging our experience.
Hundreds of engineers who have worked for years on 64-bit designs, leveraging our patent portfolio, our deep know-how of 64-bit computing, bringing that, extending that into the Arm ecosystem. We think it's a tremendous value proposition. We think we are uniquely positioned to be able to leverage this and to bring solutions to our customers that no one else in the industry can do. With that, I would like to introduce Jim Keller. Jim is our Corporate Vice President and head of our core development at AMD. I've known Jim for many years, and it's just been a pleasure the last two years to be working with you and making a difference here. Jim, why don't you grab a chair? I'll sit next to you.
All right
We'll chat a bit with the group here. You and I have been at this for many years in the industry, and we've seen change after change, and you've been such a part of it. You look at it, from your days back at DEC Alpha to even at AMD on the Hammer designs and at startups and Apple, you've been making a lot of change in the industry. Jim, tell us a little bit about AMD. You've been here over 2 years. What drove you to come back to AMD, and what were your thoughts?
That's a funny way of putting it, what drove me to come here. I joined AMD because I love processor design, and I love complicated system design. AMD was looking at taking a big swing on the next generation. There's definitely been some problems and issues over the years. We have some great products. I love the fact that we're in the consoles. We're in really power-efficient processors all over the place. We were looking at taking a big step forward. Working with some great people, leading a big team, and doing something new was kind of the challenge for me. It's been really fun. It's been almost 2 years, and we've delivered a number of products already. We're looking in the next 2 years of doing some really new, big things.
You mentioned running a big team. I know we talked about that. When we recruited Jim to run the cores team, we said, "Look, we're not going to create two different teams, one doing Arm and one x86." I said, "Jim, can you run the team that does both, right? That bring this together." I get asked the question a lot, Jim.
How can you do this? You're adding Arm into the mix. Doesn't this blow out your development expense? How have you gone about this?
Yeah. One obvious way to do it is have two different teams double your effort or cut the energy you put on one thing in half, and that obviously isn't the right way to do it.
It turns out high-performance computing is mostly about high-performance features, verification, methodology, CAD, all kinds of things. Those things are all common. The instruction set actually is important. The Arm ecosystem, the Arm architecture is very new. It has some inherent inefficiencies in it, which we think is really cool. In terms of the overall effort, we get so much leverage out of stuff that it's not just a double effort. Interestingly enough, going through the process of designing Arm has given us a whole bunch of new ideas, which I think actually drives better core design going forward. The other thing is we've talked about today, a lot of people have tried to do a completely new thing, boil the ocean, and they screw it all up. We have a step-by-step plan. Seattle is a software infrastructure play to get that working.
In 2015, we build SoCs that are ambidextrous, so we've got the plumbing right, fabric, memory controller, IO system. We also brought up a whole verification infrastructure for Arm on Arm's processor. My guys, I got to tell you, they were all excited when they thought they found a bug in the Arm processor with our tools. Turns out it was a bug in our tool. I'm still waiting for that email, Simon. I'll shoot it to you. We'll get that done. We brought up an Arm verification infrastructure that was really good to apply to our next-generation core. As we do a new processor, it's not our first time doing high performance or 64 bits. Actually, we've already been working on Arm tools for quite a while, so we're making good progress on that.
Excellent. You mentioned those Arm tools, and you heard me talk earlier about that Arm ecosystem. What are your thoughts? You've got deep experience with both. Do you see some advantages that you're leveraging with this Arm ecosystem?
Well, the big fundamental thing is the Armv8 architecture is actually quite good.
Has more registers. It has a proper three operand instruction set. We think of it as we spend less work and transistors on decoding instructions and dealing with the complexity of x86 and more transistors on performance.
Yeah.
That's a pretty straightforward proposition. That gives us more performance, and we can also turn more performance into better efficiency.
That's pretty cool.
What about power? If you look, you started with tuning Cortex-A57, then you've got the from-scratch design that we call K12. How has that worked in terms of leveraging Arm and your own designers' expertise to optimize power?
It's interesting. AMD has two families of processors today. The Bulldozer family focused on really high frequency, Jaguar family, super small cores. In our new generation, what I told my team is we got to take the DNA of both, the best of both, and put it there. This is really nice. We know how to do high-frequency design.
Yeah.
We know how to do dense design. Arm gives us some inherent architectural efficiency, the combination is pretty good.
It's probably a perfect lead-in for my next question, that is, given that, what's your sense of our differentiation? We all know there's lots of players out there. There are other licensees of the Arm architecture. What makes you confident?
We have the world's best graphics. That's great, right? We know how to do high-frequency designs. We know how to do power-efficient designs. Where we see adding on to the Arm infrastructure is, or ecosystem is, we've done servers, we've done scalability, we've done high frequency. We can extend the range that Arm's in. That's a nice play for us.
Excellent. I got to ask one more thing at you because it's something I'm enthused. I get a chance to walk the halls and talk to your team working on K12. I mentioned it earlier. What's the sense of your team working on K12? We've got deep expertise in x86. We continue to not let the foot off the gas in x86, and yet there's this whole new Arm in our offering here.
What's the thought of your team? What's the excitement?
Well, I have to say, my team was a little daunted. We're going to do a new from scratch x86 core.
We're going to do an Arm core. We're going to build a new methodology, right? There's a lot of clean pieces of paper in that plan. My ask to the team was, let's go design the best possible thing. That's the important thing. Great product. We did the high-level design as a clean piece of paper. It turns out we have all this technology already. We know how to use dense metal stacks. We know how to build lots of components. We brought that in. We filled in the plan pretty fast. The engineers are excited to be doing something new. You want to do processor design today, you should work at AMD because we're doing real new things. We're not starting from scratch. We have a lot of components that are. We make progress pretty fast.
It's like a supercharged new design plan, because since we're taking off-the-shelf stuff and ideas we worked out, that's really fun. On Arm in particular, we had to go from scratch because we didn't have any traces, performance models, ISA model. We worked with Arm. They have a really good Instruction Set Simulator. We quickly built Linux and ran traces. We started building workloads. On x86, we have tens of thousands of traces on all kinds of benchmarks. On Arm, we start from scratch. You know what? It's kind of fun to have that kind of play to go on. People got pretty excited about it. Then it turns out, learning about Arm, we leverage it in x86, and we're making a lot of progress.
That's excellent. Look, Jim, thanks very much. Why don't you join me back on stage in just a few minutes, we'll have some Q&A later.
Okay
we'll conclude this piece.
You bet.
Thanks very much.
Great.
It really has been exciting. You look at Jim, just one of the handful of industry leaders, not just core design, but Jim is a system thinker. Every aspect as he drives into a core design is, how do you put a system? How do you integrate it and design it such it can create differentiated systems? It's really been an exciting opportunity into driving our compute capability on x86, now on Arm, and marrying that with the rest of our IP, with our graphics, our multimedia, the rest of our IP portfolio. A very exciting time. Thanks for this opportunity. Rory, I'm going to turn it back to you. Thank you.
Thank you, Mark. You get to stand. Q&A. It's pretty exciting time, there's no doubt. Thank you, Mark. You think about it. It is, as I invite my colleagues to come up here, I think, Simon, you're going to join in, Lisa, Jim, Mark. Why don't you work your way up? It is an exciting time as they bring up the chairs. We're creating a different AMD. You can see it, and you can feel it. Our culture is one where we do what we say, and we own what we do. We said we would begin to turn around this company two and a half years ago. We said we would build a more consistent execution engine and a more consistent business model. That's exactly what's happening. You've seen it in the past three quarters as we begin to execute.
With this announcement, we talked about entering into high-growth markets that can create an opportunity where our IP can truly lead. We're showing that growth in the semi-custom space, professional graphics, embedded, new form factor client, and of course, in dense server. We said we were going to be ambidextrous. We were going to do something that no one else on the planet could do. We were going to take x86 and leadership, world-class leadership graphics, and we're going to combine that in a pin-compatible solution called Project SkyBridge and introduce Arm in an ambidextrous solution that opened up the world to AMD. Not in a single segment where one player dominates, but into a set of opportunities where we can partner with true market makers to create a new destiny for this company. That's what you're seeing here.
Over the past year and a half, our execution consistently better and better. Today with this announcement, we laid out the transformation over the next two years. You heard Jim and Mark talk about where that next-generation cores are going to go. It's world-class next-generation graphics and a sister set of cores, both clean sheet, bottoms up, both Arm and x86, introduced to create true leadership in a market opportunity that's much broader than we've ever had before. We've gotten the world's best people to join this team, and that's just the beginning. This is the beginning of a story we're writing. What I'd like to do now is to join my colleagues here and open it up for questions. I'm sure you have some. We have everyone here who talked today, including Jim Keller, who's a very good friend of mine.
We have a good debate on a regular basis. I'm surprised he didn't ask for more budget during the discussion. We'll work that out later.
No, not that? Okay, great. Let's have it. What would be the first question? There's microphones in the room. Nate, right here, you want to jump up there and grab one? We'll take some questions.
Sure. Thank you. I love the SkyBridge concept. I think it's the first time you're really demonstrating some of the synergies with the ambidextrous strategy. I have a couple of questions on it.
Please.
First, it sounds like you're targeting clients with this because you said APUs. Is there any chance that we would see a SkyBridge concept coming into servers as well?
Lisa, you want to take that first?
Sure, Nathan. Thanks for the question. Really, we look at SkyBridge as a design framework, so it can go across multiple markets. For the first family of products that we talked about in 20 nm technology in 2015, we will be targeting both clients and embedded with our APUs.
not server?
Not in 2015.
Okay. Is it going to be a socketed approach, or do you anticipate most of these will be soldered down?
They could be either. They're fully pin-compatible. Depending on the market, probably BGA, but we can do socketed as well.
What does that mean for the longevity of the design? Could I expect to see a second generation that would be pin-compatible with the first, or are you just making the cross x86 Arm promise for this one generation?
No, we actually see SkyBridge really as a family of products. I think you will see it at different performance classes and for different markets. You'll see it in clients, you'll see it in embedded. I think longevity is very important for embedded, so I think you'll see more from us there.
Okay, great. I have some more questions, but I'm going to sit down and let somebody else ask.
One of the most important things about what we're doing with the architecture is that we're applying it to each of the growth segments. The power of the strategy is this expansion of TAM and opportunity. We had great IP forever in AMD, but our execution was a little bit choppy, and we weren't able to bring it to bear on the marketplace. With SkyBridge and this ambidextrous architecture, we're positioned to take it into all segments, even into semi-custom, where I think we'll see very interesting applications as we move forward. Clearly embedded, long-term solutions that create sustainable and consistent revenue performance. That's key to the overall strategy. Next question.
Hi, Kevin Krewell from Tirias Research.
Hi, Kevin.
Follow on to Nathan's question on SkyBridge. Would you say SkyBridge is going to be a family of products that have pin compatibility x86 with Arm, or how would you call the SkyBridge family, Lisa?
Maybe let me start, and then I'm sure Mark can add to it. We look at SkyBridge really as a design framework for interchanging our Arm and x86 complexes and SoCs. In the implementation that we talked about, we drove pin compatibility because that forced a certain way of thinking for our engineering teams and what we were trying to do. It really is a broad framework for how we are able to reuse our IP and really help the customer reuse their IP across Arm and x86. Maybe, Mark, you want to comment?
Sure, I'll add on. Kevin, what Lisa described is, I'll say, absolutely a critical stage 1 of what we're doing. It is a framework, it sets the stage for even further optimized ambidextrous design. That's where it comes into play some of my description of a network on chip and the kind of flexibility that we put in across our design system to allow us to be very flexible on Arm and x86 and how we marry that with the rest of our IP. In fact, how we configure tailored SoCs for the segments that we play in and with semi-custom.
Yeah.
Kevin, one of the things that was really important about many of the questions we got as we took out that 30% of the cost structure about a year and a half ago during the restructure, we talked about how could we do that and introduce the solutions. I think, Jim, maybe you could add a comment or two about how your design teams have really worked across these two architectures. You touched on it in the Q&A with Mark, this was a big deal.
Yeah. Just to be clear, in 2015, we have a pin-compatible part. Inside, my team did the processor development for both hardening the Cortex-A57 and the Puma+ core. They didn't look anything like each other when we started, we said, "We want to make one SoC. There's one place to put the core. There's one box. It's got to be the same size and shape, plus or minus." We did a bunch of work to go reconfigure the core to be able to put either core in the same SoC. Now we have an existing fabric that supports the APU family that we currently ship. It turns out SkyBridge and that fabric are cousins in terms of some semantics on how it works.
We upgraded the fabric to be able to support either Arm or x86 in the same fabric. Did the work in the I/O system, the IOMMU, whole bunch of dirty details to get that right. That's in 2015. The next generation, we have a next-generation fabric that spans APUs to servers, and we put a lot more features in there to make that scalability work across Arm and x86.
Our new cores fundamentally started out with that target in mind. That makes it easier to make them look the same way. The term of the plan is we got the software right, we got the SoC right, we got the fabric right, and now we're doing new cores that kind of plug. It's compatible at the outside at the pin level. It's very compatible, actually, on the inside at the SoC interconnect level.
Going forward, do you think you're still mixing standard Cortex-A57 cores and the K12 core after 2016, or they're all?
Sure. They're differentiated in size, performance, and power.
Yeah.
They're both beautiful parts, and one will target the higher end, one will target the other end of the market for us.
Okay, thanks.
Next question, please.
Gil Russell, Bright Side Analytics. I guess it's a two-part question. The first part is, I see this part going to the infrastructure server market, which represents probably the biggest TAM. Also it means that choppy margins are going to be smoothed out somewhat through the sales of this, because the volumes are really extremely large, especially in Asia. That leads to the first question, is what are you doing to beef up your customer design centers in Asia? The other part of the question is, where's SeaMicro?
Sure. First, on the design centers, we have a huge presence already in Asia, both in India and China. I think, Mark, you can probably give an overview of just the huge amount of engineering talent we have there, and then we'll move it over to Lisa in terms of how we're targeting that market opportunity, and we'll touch on SeaMicro.
Great. Well, with regard to China, not only is it a huge market for us, it's a key element of our engineering development process. We have a very deep presence in China. We have over 1,000 engineers, and those engineers are developing our base IP and our SoCs, but they're also there working with our sales team and enhancing our relationship within customers in China. It's a very clear investment we have in the region, and we look to grow that going forward. Lisa?
Yeah, I'll just add to that, Gil. If you think about the total set of resources we need to help our customers be successful, we have our design teams, but we also have quite a bit of, I would call, field application support that's coming up, particularly around the concepts like SkyBridge, because there's a lot of detail in that. We're absolutely beefing that up. That's part of our strong investment in our new markets, so embedded and semi-custom and server, and those are skill sets that we're adding to the picture. Your question about SeaMicro, we're very pleased with SeaMicro. We look at that as a way to demonstrate the technology. Today, we're demonstrating Seattle in a reference board, in a customer environment where they can use software.
You will see Seattle in our SeaMicro systems, that's just another way for us to demonstrate sort of the disruption that will happen and really give customers different sets of system choices.
Thank you.
Gil, we really wanted to knit together that set of architectures in that SeaMicro fabric to really create the opportunity to create this inflection point in the marketplace. I think, Simon, one of the reasons you wanted to work with AMD from an Arm perspective was you saw a real opportunity in the server space. Any thoughts for the team in terms of where you see that going over the next four or five years?
I think with the growth of small embedded devices, such as those that are coming about through the Internet of Things, with the growth of mobility, you're just going to see all of this data. To deal with that requires a new approach. We're seeing network infrastructure being re-architected to deal with increased volumes. You can only have all the potential of all of this technology if we continually focus on reducing power and making the whole system more efficient. That's where I think our partnership's going to pay dividends.
Yeah, we were pretty excited about the partnership from the get-go. This was a part of that turnaround strategy from the get-go, and we looked at it from the standpoint that we knew we could win with a combination of graphics and x86, but there was a limited TAM in terms of that, and some of the markets are somewhat blocked for us. By combination of this ambidextrous architecture, a partnership with Arm, and really creating a much broader set of markets where we could bring this IP, we could unlock a whole set of opportunities for AMD to get a much more consistent level of performance, consistent growth, and consistent profitability. That's why we went after that architecture. We announced it today, but this architecture was something that we laid out-
Years ago
years ago. We've been talking about this from the get-go. Thanks, Gil.
Thank you.
Next question, please.
Hi. Bob O'Donnell, TECHnalysis Research. Jim, I want to follow up on a comment you made when you were talking about designing both for Arm and x86 at the same time, that it gave you some new ideas. Now, it may be too early, but can you give us a sense of what kind of new ideas could you potentially bring to x86 from having done Arm?
Yeah. Let me say, I have a really good team, the architects, the implementers, verification team, they're all really strong, they're really experienced. Whenever we go say, "Let's go make a faster design," we know how to do that, partly it's planning. You have to have the right time horizon, so you have time to actually think hard, get the work done. As you do the design, I say architecture is a journey, right? We need to know where we're going, faster and faster computers, more efficient computers. That's all important. As we build things, we take our best ideas, we put them down. We learn a lot. Sometimes the problem is bottlenecks, sometimes the capacity of some structure. As we go through all that, you get more ideas for the next generation.
The way we built the Arm processor was a little bit different from x86. We built a bigger engine. It has some interesting features, thinking about that has given us some ideas for the next generation. I don't want to be too specific yet. In the next year or so, we'll be talking about product details, it's partly about this is an intellectual journey. It clearly is, right? Doing the design is hard. Putting your best ideas down is what you have to do. You'll learn a lot from that. Every architect in the world kicks themselves for the last thing they did because all that stuff that looked so good on paper just, it was good, it's bad. You figure out a lot of stuff, you go do the next thing.
Making different choices gives you ideas, get different experiments, we go forward from there.
Just as a follow-up then, on the graphics side, are there any issues with using the same graphics core for both Arm and x86 in SkyBridge, for example, or any challenges associated with that?
Graphics really talks to memory, right? What graphics need is a really high-bandwidth memory system. The old kind of graphics was graphics had its high-bandwidth memory system, and it had its own memory system, right? For the CPU and the GPU to talk to each other, use PCI Express. You had a driver model that wrote registers or wrote into the other guy's memory. With HSA, we made a memory architecture where they share the memory. This is a great thing. Graphics sees memory, CPU sees memory. We can pass pointers between the two of them. We have a common address space. We developed that independent of Arm and x86. We say, well, we have this graphic system that understands common unified memory. That works really good. The processor architecture, Arm or x86, they talk to memory.
We know exactly how to do that. The interesting thing about the software, it's where the changes are in the software, right? We don't want to build two software stacks, one for Arm and one for x86. The software architects are figuring out all the common driver elements that are sort of ISA independent, and then only optimize the stuff that's very specific to Arm or x86. The plan sort of fit in with HSA. We had that plan before the Arm plan. When we got to Arm, one of the goals of HSA is to be a little more independent of some of the hardware details. That's working out just great.
Great. Thank you.
Next question, please.
Hi. This is Gabriel Ho from BMO Capital Markets. I have questions on your process technology roadmap for 2015. Mentioned you will be releasing 20-nm-based product. Do you think, as the transistor cost is going up quickly on a wafer pricing, do you think 20 nm is going to cut in 2015? Second part is where does FinFET plays in your roadmap?
Lisa, you can take first.
Sure. A little bit about our process technology roadmap. In 2014, we're 28 nm up and down our stack. We mentioned the first implementation of SkyBridge will be in 20 nm. We made that choice quite a while ago, and really the thought process was 20 nm would be the right time to market to implement that framework for these design systems. We are very actively designing in FinFET as well, and that'll be an important technology for us as we talk about some of the new cores and technology that Jim and Mark talked about.
Great. Thank you.
I think it's really important when you begin to see this architecture take hold. We've talked about it many times over the past couple of years, Jim kind of highlighted in the sense how he was able to reuse the IP and to create an architecture that allowed us to drive efficiency in our model. This was really important. Mark, one of the things we've talked about many times with this population and the media was about SkyBridge and SoC '15. Jim touched on this. Your view in terms of this, in terms of what it will mean to our ability to address these 5 to 7 key growth markets that we see over the next 3, 4 years, and to do it efficiently.
Sure. Well, efficiency is key. The bottom line is you can't have great designs if you don't have a vision up front as to how you're going to have that great design. That's what this was about, is starting early and saying, "What do we envision?" So it was this two-step process. Jim talked to you about with step one in terms of having this capability that we talked about with SkyBridge in 2015, of being able to have this pin for pin compatibility, and it had ramifications right back into his core design, to be able to anticipate this. So that was step one, is 2015 with SkyBridge. Rory, going forward, it is truly the underpinnings of our ambidextrous approach because the flexibility that we built into this network on chip, into this onboard fabric, is quite flexible.
When I look at historically, when we would have gone about targeting these markets very specifically, because it takes unique combinations of IP, as you want to have that winning server part, you want to have that winning part in desktop as opposed to what you might do in mobility, as opposed to what you might do in some of the embedded clients. Not to mention semi-custom, which is, of course, a highly tailored design. Frankly, historically, this drove incredibly high resource allocation to each of these instantiations. Really with this kind of thinking, the strategy up front that we built in, it is actually quite significant the efficiency that builds into, based on that design approach, to be able to put these solutions, these more tailored solutions together to the market.
In first quarter, we saw each part of our business begin to make important contributions to our results, and that was really interesting. Lisa, you're doing a lot of work with customers across the planet in terms of how they're relating to AMD. We talked about in that earnings call about how critical it was and really the power of the strategy was this marrying of our world-class IP and design capability with the market makers, knowledge of the market, and their great ideas to create great solutions like Xbox One, PlayStation 4. These were really differentiated solutions. You're seeing it across the board. How is the customer set reacting to this, and why do you see it as such a big deal?
Yeah, I think it's really exciting to be able to offer something different to the customer set. If I just look over the past two years, the conversations that we've had with customers, they went from, "Hey, can you do the basics of execution?" Which we had to demonstrate. They saw us ramp some very aggressive semi-custom designs, as well as our graphics and leadership products. Now what we're talking about with customers is how do you take all of the building blocks that we just talked about? How do you take our leadership IP? How do you take our SoC methodology? How do you take our semi-custom and embedded business models and really do something market-changing? That is a dynamic that I think is very exciting, and it's a different dynamic than when we first started these conversations.
No doubt. Let's take another question. We'll go first one in the back, and then we'll come to the front. Ian.
Hi, Rory. I think you guys have made a pretty strong case for why Arm and servers make sense, and clearly, you're going pretty hard at that. A lot of questions you've had have been about resource allocation, and I guess the biggest question that springs to mind when everybody looks at the financials is why continue to invest in x86 to servers? Why continue to spend that money? Why not just give Jim everything you can to go full steam ahead in Arm-based servers and break away completely from what you described as an unhealthy duopoly?
Well, it's kind of interesting. When we had the opportunity to look at the portfolio and how we were spending the money, we really wanted to look at the IP differently. Every part of our IP, they're not all created equally. Cores and graphics, Arm and x86, we're going to over-invest in those, what Mark calls truly leadership, the crown jewels of AMD. We want to create leadership there that's undisputable, and we'll make the investment in there. We think, Ian, there's a unique opportunity not only in server, but in embedded, in semi-custom, that this kind of architecture is going to unlock for us. Because of the architecture that Mark and Jim created, working with the architects at Arm and Simon's team, we've been able to do it in a very efficient way.
You look at the work that we've done in terms of taking our cost down over the past two and a half years, yet we're delivering every part of a roadmap much more effectively than we've done in the past. We're showing you what we're going to do the next two years. I think we're at the beginning of a really important inflection point. Maybe, Lisa, you want to add a little bit of more color there?
Yeah. Ian, if I just add a little bit to that, I think the question shouldn't be about Arm or x86. What we're trying to get really as a key message is it's going to be Arm and x86, okay? These are the two most important architectures. They will continue to grow. When you look at the server market, the server market is also bifurcated. I think you see traditional enterprise will be x86. It'll be x86 for a long time because of the legacy software environment there. What we've chosen to do today, and really in our short-term and medium-term strategy, is focus on the fastest-growing segments in server, which happen to be in the dense server market, which Arm serves very well. We're also doing APUs, so our Kyoto product is another x86-based APU that we see in the market.
We will continue to grow that over a period of time. I think it's really about the merging of those two ecosystems and using the best silicon for the right software environment.
Excellent. Question here in the front.
Sure. Congratulations on the announcement today.
Thank you.
It sounds really interesting. You definitely have a unique position there. I guess I'm not sure how much users will want in one board to do that capability of having either architecture, but I'm sure that some will, and that's certainly something that's unique that you can do. The real question, though, is how will you differentiate your Arm SoCs and your Arm cores, given that there's 10 licensees out there, and most of them are a year or two ahead at least in doing this for some of the same markets that you're looking at, either the Applied and the Broadcoms looking at communications and servers or the Qualcomms and so forth looking at mobile. How do you differentiate?
From the first perspective, the key differentiation is the high-frequency design, the low-power capability, and the world-class graphics. We're going to take this architecture and apply it into these five growth segments, as well as our core business. This is going to give us another set of arrows in the quiver in order to go attack those opportunities. It creates a whole set of opportunity, both in embedded, in semi-custom, in server, and even in client, where I think it's going to be very powerful. Maybe, Lisa, you want to add a little bit on that?
Yeah. Rick, I think that's a very good question, important question. What we like about the Arm ecosystem is how broad it is in terms of the volume dynamics it drives. We're also very specific about where we are going to differentiate. If you look at the entire Arm sort of market segments, you're probably not going to see us in low-cost smartphones. Actually, I think you could say you're not going to see us in low-cost smartphones because that is in our DNA. You're not going to see us in $2 microcontrollers because that's not in our DNA. We chose an inflection point to really jump into Arm. There's a reason we started with 64-bit. There's a reason we started with Cortex-A57 and then going to K12.
We are catching the point where we believe the intersection of the Arm ecosystem with higher performance applications, with graphics, with total compute, with our system-on-chip methodology, with everything that Jim and Mark are bringing to the table. That's really where we're going to differentiate, in those higher performance, power efficient, more complex SoCs that have typically been of the x86 variety, and we really intend to bring some of those to the Arm ecosystem.
This cloud era is going to definitely be about the experience and the graphics. Simon, you've been very deeply involved in this strategy over the past couple of years, as we've met many times to talk about it. What was the attraction to work with a company like AMD, and why do you see it as a unique opportunity? You've got a great set of architecture and IP, but we bring something to the table, and your team has seen it from the get-go. Any thoughts on why we might be a really interesting competitor over the next three or four years? Not for you, but for the others. We're going to be your partner. Yeah.
It is about expertise in particular markets. AMD's heritage in the server space, in high-performance computing, brings additional breadth to the overall Arm offering. We're trying to get Arm into a position where it can target as much of the semiconductor space as possible. Things that the market is evolving, new applications are emerging, and we're looking to try and create an environment where those markets can be addressed in as efficient a way as possible. The work we're doing to drive things like HSA, the base server architecture, are enabling these innovations to come about more quickly. We think that creates a really exciting space, and we want to work with anyone who wants to run really fast, and that's what you guys want to do.
Yeah, there is no doubt. I got very excited when I heard Jim say the reason he came to AMD was because we wanted to take a big swing. That's exactly what we want to do. We want to do it in a very methodical, orderly way, built on a strong foundation of execution, but to go after it and change the market. I love that. That was the best part about it.
Rory, we're going to make this the last question.
Okay. Thank you.
Well, in that case, it's going to be a two-part question.
Oh, yeah. It's going to be the last two.
With regard to K12, can you give us a little bit of a sense of the business strategy there? Which segments are you targeting, or at least which segments will K12 show up first?
I'll start with that, Nathan. I think you will see K12 in servers. We've talked a lot about servers. You'll see it in embedded and other applications as well.
What about semi-custom?
semi-custom.
Oh, good.
K12 is going to be a huge addition to our semi-custom IP portfolio.
You didn't say client.
Well, we've talked about Arm-based clients, so with our Cortex-A57, and I think as you see the Arm roadmap unfold, you'll see more from us.
Okay. This last one might be a little geeky, but I just want to follow up on something Jim was saying. When you introduced Kaveri, you talked a lot about full HSA support, which meant heterogeneous uniform memory architectures, heterogeneous queuing between the GPU and the CPU, and so forth. Then you said you're going to have full HSA support, easy for you to say, in the 20 nm products next year. Does that mean that all of the HSA features that we see in Kaveri today are going to be in the Arm-based SoCs that are going to come out next year?
I can take that. That's correct. Nathan, as you recall, when we rolled out our implementation of the Heterogeneous System Architecture, what we were doing with the foundation, with Arm and others, it was in fact to make sure that that capability, so the overarching architecture, would be instruction set architecture independent. So in our roadmap, that's absolutely correct. We'll have all of that capability would extend to both facets of that SkyBridge implementation, both the Arm and the x86 implementations.
Very cool. Thank you.
Thank you. Well, I'd like to thank my colleagues here for being such good sports and answering the questions. I wanted to thank you all for taking the time to join us today. This is a big event, and you clearly can see something changing at AMD, and it's changing for the better. This is a series of announcements you'll see over the next two years as we finish this transformation and continue to deliver that consistent performance and growth and profitability that we've talked about. With that, we'll finish up the meeting. I'll say goodbye to our friends across the world on the internet, and we'll finish today. Thank you all for coming. Appreciate it very much.