Good afternoon. Thank you for being here. I'm Colette LaForce, Chief Marketing Officer of AMD, and we want to welcome you on this very historic day. A special thanks to our partners, customers, press, analysts, and all of you watching on the web. We appreciate you joining us. Before we get started, I want to turn your attention to a slide here that describes the fact that we may be mentioning forward-looking statements today. With that, we'll get started, and I want to introduce Rory Read, our President and CEO. Rory?
Hello, everyone, and thank you for joining us today on this historic day. I think it is a key moment in the industry. I don't think there's any doubt about it. Today is about what AMD really is at our core, and it's in our DNA to innovate and to think about what's next. We look to disrupt the status quo. We look forward to drive the industry to where it needs to go to inspire competition and to enable our customers to do more. AMD will transform the data center computing environment today. As we announced, we're the first company to offer both 64-bit Arm and x86 server processors. Yeah, I like that too. There's no doubt that the cloud changes everything.
In today's modern data center, it needs more change than it's probably anywhere in the industry, and the cloud truly is the killer app that's unlocking the future. It's driving the fastest level of growth across the industry. Over the last 10 years, the last decade, we've seen an annual increase of about 33% in CapEx spending in the data center on the large mega data center cloud services. That's only going to continue to evolve and expand. At AMD, we have the IP, the experience, and the engineering talent to disrupt, to create the differentiated solution and differentiated choice for our customers. Dense serving is the key to our long-term strategy and growth. We talked about this, and we're going to continue to pave the path to success by providing our customers with that differentiated solution that changes the game. We've talked about an ambidextrous architecture.
What we're doing here, just as we did with AMD 64 in the last decade, is we're going to open up a next level of computing with ARM64, transforming the server area into a whole new opportunity, creating differentiated IP across an agile, ambidextrous architecture with reusable IP blocks that allow us to deliver lower-cost solutions on a faster timeline, unlocking the computing power of the server environment with our HSA heterogeneous systems architecture that allows you to run the workloads on the CPU and the GPU where they run best. We are uniquely positioned to lead in this era of cloud servers. We have a full suite of x86 processors, IP graphics, and now Arm processor cores. We'll couple that with our SeaMicro Freedom Fabric. That SeaMicro Freedom Fabric is ultimately very, very important in this announcement.
It is the only production-ready, production-tested super compute fabric on the planet, and it differentiates our offerings from others as it combines both Arm-based servers and x86 solutions. We have years of experience in the server market. This is not new to AMD. It is in our DNA to innovate. We have the expertise, the engineering capability, and of course, the customer and partner relationships. It's really important if you think about why did we choose to go in this space. What I think is important is about the partnership that we create with Arm. Why together? Arm shares that same deep passion for innovation, that drive to look for what's next. They don't accept the status quo. They want to change the future.
They want to create a future that's differentiated, where the customer has choice, where the customer can ultimately create a high-speed, fast solution at the right price point that unlocks the future. Arm and AMD together can change the industry landscape, can drive true competition, and spur the next level of server innovation. We share a common vision that where the industry is going, how we can disrupt it together, and ultimately, how we can lead this transformation. I think it's a pretty exciting day. In summary, you have to think about it. When are these seminal moments in the industry? When do they occur? Maybe they're not as always as obvious as they are. Just like AMD 64, that's this moment right here. AMD is the only computing company on the planet that's introducing this ambidextrous architecture spanning both Arm and x86 ecosystems.
It's the right place, the right teams coming together to create the right solution in this next era of cloud-level server computing at the mega data center. We believe, in concert with our customers and our partners and critical ecosystem members, we will usher in the next era of server computing. It's pretty exciting day, no doubt. I have a good friend at Arm in Warren East, the CEO. Warren wanted to be here, but he was flying from London, interestingly enough, and he called last night about 7:00 as I was in the airport, and he said, "Rory, I'm coming to San Francisco through New York." I said, "Warren, that's probably not a good plan because Hurricane Sandy is there." He says, "I'm not going to be able to make it." I said, "I'm going to get here later in the afternoon.
I really want to be part of this." I said, "Warren, hey, you've gone out of your way. This partnership we've been working on over the past year, we've been building this relationship. We've been trying to innovate and create this next future together. I understand. This is out of your control." Warren said, "Maybe there's a way." I said, "I got to get on the plane." I flew across to California. I landed here, and I got off the plane, and I hear Warren's going to do a video. Warren's doing a video in a taxi at the Heathrow Airport this morning. That's the kind of partner Warren East is. That's the kind of leader he is, and he sees this same future that we do at AMD. I'd like to run a video of Warren East, the CEO of Arm.
Good afternoon. I'm Warren East. I'm CEO at Arm. Before we begin, let me first apologize for not being with you in person. I was on my way to San Francisco last night via New York. Unfortunately, Hurricane Sandy has intervened. 12 hours later, I'm in the back of a taxi at Heathrow Airport. To start with, let me begin by welcoming AMD to the Arm partnership properly. AMD adopted the Cortex-A5 for use in security earlier this year. With today's announcement, we have AMD laying out their plans for Arm technology in the server space. Arm, as you know, has a long heritage in power efficient systems, highly integrated system on chip devices honed in the mobile space.
All of the mobile and non-mobile internet connected devices are today putting huge demands on data centers, driving up the requirement for ever more data capacity in these centers, driving up the demand for more and more servers. All of these servers take huge amounts of energy, require huge amounts of energy for cooling and so on. A new methodology is required. AMD have a lot of expertise in the server domain, and combining that expertise with the highly efficient, highly integrated system on chip technology from Arm looks like a good way forward for servers. AMD's knowledge and expertise of the channel, knowledge of requirements of system designers, and in particular, the software ecosystem required in the server space makes a great match.
We look forward to the transformational effect that AMD's presence with Arm in the server domain is going to have in that environment over the coming years. Welcome to AMD.
Thank you, Warren. There's no doubt that's a true partner. That's an innovator. He doesn't take no for an answer. He found a way to be here with us today. As he said, Warren said, we are excited about this opportunity and expanding our work together. We believe together we are uniquely positioned to truly change the industry. It won't happen overnight, but it's a journey that we've begun over the past year, and we're going to continue step by step to move through that. With that, I'd like to introduce Dr. Lisa Su, our senior vice president, general manager of Global Business Units, who will talk about our product plans in more detail. Lisa?
Thanks, Rory.
Thank you.
Thanks everyone for being here. I'm really, really excited this afternoon to talk to you in a bit more detail about what we're doing together with Arm to really transform the data center. If you really look at what are we trying to do, what's the problem that we're trying to solve? You heard it from Warren, you heard it from Rory. It's that the data center is now being inundated with massive amounts of data, there has to be a way to do it more efficiently in smaller space with less power at a better cost point. That is the problem that we're trying to solve together as an industry, that is what we're trying to lead here at AMD. First, let's take a look at what's happening in the market dynamics. I think the market dynamics are very clear.
It's the explosion of devices with the advent of the smartphone, iPads, new business models, Google, Facebook, eBay. All of these cloud computing workloads are putting tremendous strain on the data center. Just last week, Cisco posted their latest global internet outlook, they showed that data is going to grow in the data center at over 30% CAGR for the next five years, five years plus, we'll be at over six zettabytes of data in the data center. That's a tremendous amount of data that we're processing, it's really changing the way we think about data center economics and data center dynamics. Here at AMD, we have a history of really trying to revolutionize the data center and what really moves innovation forward. Back in 2003, we were the first to introduce 64-bit x86 architecture into the data center.
We were the first to introduce dual processors in 2005. We moved to the petabyte supercomputer in 2008. Earlier this year, when we did a very important acquisition, many of you asked, why is AMD buying SeaMicro? SeaMicro was a systems company. We were a chip company. SeaMicro was a startup. We were known for our consumer devices. We really have this vision of we're really looking at where does technology really need to go as we look over the next five years. The data center is fundamentally where we think the growth is going to be and where the innovation is going to be. SeaMicro brought to AMD world-class leading-edge fabric systems and software expertise, as well as hardware expertise.
Today, I'm really, really happy to talk about the next part of our data center strategy, which is something that we've been working on for quite some time, and it's really the announcement around production-based Arm servers based on AMD and SeaMicro technology. We will be launching these products in 2014. We believe we will be at the leading edge of Arm-based servers in 64-bit. We have something that is very differentiated. In addition to all of our microprocessor and technology IP and expertise, we have the best compute fabric in the industry with the SeaMicro Compute Fabric, and integrating these two pieces together will give us an opportunity to really accelerate the adoption of Arm in the data center. When you think about what that means in terms of the data center, I'm going to show you a set of workloads.
The biggest change in the data center is there is no one size fits all. We will offer Arm-based CPUs with our fabric, we will offer x86-based CPUs with our fabric, and we will also look at opportunities where we can merge the CPU technology together with graphics compute in an APU form factor that will be very, very good for specific workloads and servers as well. AMD will be the only company that's able to offer the full range of compute horsepower for the right workloads in the data center. Let me take you through some of the thinking and how we think this is all going to come together. One of the main points about why things haven't really evolved that much up until now has been that the last two decades, everything has been around, let's call it homogeneous workloads.
It's been around traditional IT. The idea of how do you make a data center more efficient has been, let's put more CPU horsepower, let's increase frequency, let's increase just base processing power for homogeneous workloads. That world is changing now, and that change is what's giving us both the problems as well as the opportunity. The last five years, we've seen the rise of all of these mobile devices and all of these new business models, from the e-commerce to the e-business workloads coming in. What those workloads have done are, as I said, they've shown a tremendous amount of strain in the data center. We're adding lots and lots more data that we expect to move around at a minute's notice.
It's also created opportunity because many of these web workloads are actually highly parallelizable. What that means is we can break up their tasks over numerous CPUs so that we can do them in a much more efficient way. This is the change that allows us to introduce new technology. Arm CPUs are particularly efficient for these types of workloads. Arm CPUs have really had a history in the consumer market, from smartphones through mobile devices through embedded. It is really a standard out there in the industry. What it does is it allows us to really optimize and leverage all of that compute per dollar and compute per watt that comes from the consumer ecosystem. These CPUs are particularly beneficial for these new workloads in the web and in the massive data centers. This is where we think that Arm will excel.
Now, you should say, okay, Arm is a good choice. Why Arm plus AMD? AMD has had a history of innovation in the data center. We believe in technology innovation and technology leadership. We've been shipping Opteron-based AMD processors for over 10 years. We serve every major customer in the server business. We have world-class 64-bit microprocessor technology, and we continue to extend that in both x86 and now in the Arm world. There are many things about servers that are different from the consumer world and the enterprise-class IP. Being able to ensure that we understand the quality, the reliability aspects, the design tools, the platform aspects to serve the enterprise, we think are very, very beneficial to really accelerating the adoption of Arm in the data center. Then finally, and most importantly, the fabric technology is the secret sauce.
This is what will make AMD's solution different from other vendors that will be offering Arm-based systems in the data center. Let me talk to you a little bit about the fabrics and how we think about fabrics, because it really is something that's very, very synergistic to all the problems that we've been talking about today. We said that Arm CPUs have an advantage because they are very, very power efficient. They're small, they're capable. We can put parallelized workloads on hundreds and thousands of CPUs. The problem is, if you connect those CPUs directly to the network, you've just moved the problem from the processor over to the network side, you won't be able to realize the total system cost of ownership reduction that you're trying to do.
The fabric technology really enables you to really connect lots of efficient CPUs together into a cluster so that you're able to link those clusters efficiently into the network. In essence, they really unlock the power of these efficient CPUs going forward. With the SeaMicro acquisition being now part of AMD, we have continued to evolve this fabric technology. We just recently announced, a couple of months ago, our SeaMicro 15000 series, which is the first product that introduces both additional horsepower as well as AMD Opteron-based solutions. We're going to embed this fabric in our Arm-based CPU offerings. This is the only fabric that's production tested. It's been in the field for over two years. We've run millions of hours, we are trialing this at many of the large mega data center customers in the world.
We believe that the technology, together with Arm technology, will give us a unique advantage to accelerate the use of these small, efficient CPUs in these data center workloads. Going forward, how do we look at the data center, and how does it really shape up? It really is the idea of there is no one size fits all. We don't want to say A is better than B or B is better than C. We want to ensure that we have the right product for the right workload for the right problem that we're trying to solve. When we're in these web and enterprise applications, this is a particularly fast-moving part of the industry. This is where you're going to hear from our friends at Facebook a little bit later. You can tell that there's so much innovation happening here.
They're also at the leading edge and able to adopt new technology. This is an ideal place to use Arm and really bring up Arm-based servers as well as x86 CPUs. When you're in the traditional compute with compute clusters where there's a lot of legacy code, you probably will stick with x86 CPUs, and we'll continue to offer leading-edge x86 CPUs in that framework. In some of these specialty workloads like media clusters, online gaming, streaming media, these areas really offer tremendous opportunity for APUs that connect both CPUs with graphics compute technology. Also very relevant in high-performance computing workloads. The answer is, the data center is undergoing fundamental change. It is important to be able to use the right problem for the right workload, and both Arm and x86 have very complementary and important roles in that ecosystem to come.
Probably the largest question that many people have had about Arm is when is the 64-bit ecosystem going to get there? How is that going to come together? We're going to have a very interesting discussion following my talk, which really brings some of the ecosystem partners together. It is true that Arm has a great client ecosystem today, and you'll see that there's a lot of interest by end customers for bringing Arm into the data center. The key is there's disruptive cost, disruptive power advantages for doing that. We need to make sure the ecosystem comes with that. We need to make sure that the operating systems, the tools, the compilers, all of the total suite are there. AMD is committed, absolutely committed, to making sure that that happens.
We've shown that we could do that with the 64-bit x86 ecosystem, and we are, together with Arm and our partners, going to do that for the 64-bit Arm ecosystem. This is one of those things that the advantages will really drive the industry to move faster than what normally you would do. With that, let me summarize. I will say it's a happy day today to be able to really share with you some of what we're doing in terms of innovation in the data center. I think we have really put together a fantastic group of people to talk to you about why we think this is important. There are very few times in industry when you see these inflection points.
This is one of those times for us to lead, for us to bring the industry to a better point, and that's what we're committed to do here at AMD. Thank you very much for your attention. I gave you a little bit of background of what we're thinking, but very important is what the industry is thinking. We've collected a panel here today of several very esteemed members of the industry who are actively working in the area of servers, data centers, and the overall Arm ecosystem. Joining me on the panel today is Simon Segars. Simon is Executive Vice President and General Manager of Processor and Physical IP for Arm. He basically runs all of the processor and IP for Arm. He is both a great partner and a great friend. Simon, thank you for joining us.
Thank you.
Thank you. We also have Jimmy Pike, who's Vice President and Senior Fellow from Dell Data Center Solutions. Jimmy is an absolute expert in cloud computing and data center architectures, and he's made very significant contributions to Dell's enterprise solutions. I also consider him one of the foremost experts, I call Jimmy when I have a question in what to do, and I'm really happy to have him join us here today.
Thank you.
Thanks, Jimmy. Jay Parikh is the Vice President of Infrastructure at Facebook. Facebook has one of the most exciting uses of the cloud and the data center. It is really serving nearly a billion users worldwide, and Jay was telling me before the panel of all of the interesting use cases they come up with. Really happy to have Jay join us. Jay, thank you very much for being here.
Thank you.
Thank you. Tim Burke is Vice President of Linux Engineering at Red Hat. Tim has more than 25 years of experience in the industry, and he runs all the development teams for Red Hat's enterprise Linux operating systems. Tim, thanks for joining us.
Thank you.
Finally, he doesn't need to be introduced. Nathan Brookwood is joining us to moderate the panel. Nathan is a research fellow at Insight64. He is truly respected as being a processor expert, a semiconductor expert, and a system expert, and he's also extraordinarily well respected as one of our foremost industry analysts. Welcome, Nathan. Thank you for being with us. I'm going to turn it over to Nathan.
Thank you. Before we get into the panel, I know that AMD had wanted Amazon to participate in this session, they couldn't be here, but they did send a video. Why don't we take a look at the video of what Amazon's views on this.
It's a great day for the overall server ecosystem. It's a step forward for the Arm server world. I'm really excited to hear this new part being announced, and I'm even more excited about the prospect of seeing it shipped in volume. The most important driver of the cost of a modern data center is power. Anything we can do to lower the power for a given amount of work done is extraordinarily valuable. Nowhere is there a richer R&D stream on power performance and power optimization than in the cell phone world. That ecosystem is driving some of the best innovations possible in power and power performance. In fact, a little secret, I've always used what's happening in the cell phone world as a predictor of what's coming in the server world.
Technologies that show up first in the cell phone world end up in the server world. Knowing that is one of the reasons why I focus on understanding that market and knowing where it's going. It's because that's telling me where the server world's going. There's huge volume economics behind the Arm market. There's huge innovation behind the Arm market. Build on that ecosystem. At the same time, they're leveraging the Advanced Micro Devices experience in building server CPUs for many years. Bringing those two together and building a server-optimized part is taking technology that's there in volume today and is changing and growing and improving incredibly fast with existing server expertise.
Thanks. One of the things I've observed, I've been following the 64-bit computing world for a long time now. As a matter of fact, some people think 64 and Insight64 is when I graduated from high school. I can absolutely say that's not true. It does have something to do with 64 bits. It's interesting that it took a long time, almost 5 years from the time that we started seeing 64-bit, when AMD started talking about 64-bit processors, until there was any kind of ecosystem and support for it in the industry. Yet, with Arm, that's happening much faster. Let me just illustrate. Andrew Feldman, who is the genius behind the AMD Freedom Fabric, I first met about 18 months ago when he was selling dense microservers with Intel Atom inside.
Six months later, I ran into him when Arm was announcing their 64-bit architecture a year ago at Arm TechCon. Three months later, there he was, AMD had just acquired them, and they were talking about Opteron-based dense microservers. Here we are another 3 months later, and we're talking about AMD being in the Arm market. In 18 months, we've gone from little Intel dense microservers to having AMD making this massive commitment. In another year or so, we'll see these products start to roll out. It's really an exciting time if you're an analyst watching this. With that, let me just start, because I've got these wonderful guys here who know so much about the industry. Let me throw a couple of questions out, and I'll start with you, Simon. You're going to be on the hot seat.
There are some people who say this whole microserver thing is vastly overblown, and that it might eventually gain high single-digit kind of market share in the server market. Obviously, Arm must think differently. What do you think that those people who are talking about this being an inconsequential portion of the market are missing?
Well, I think there's a much bigger opportunity than a couple of single digits. I think it really comes down to how you define a server. The conventional microserver is, to the extent we've known it today, has had a small processor with a bunch of peripherals that are shared amongst other small processors. I think what we have the opportunity to do here is bring all of the SoC kind of technology that has evolved so rapidly in mobile to a server, and that's going to allow us to take a completely different approach to architecting the chip, to architecting the system, and delivering radically more power efficiency into the end market. You mentioned that we've been through 18 months of rapid change.
I think the next 18 months, we're going to see a lot more rapid change as the capability just to completely redesign what a server looks like from the ground up is actually made available to us. I think that there's a much bigger opportunity than people have thought about traditionally. Putting an exact number on it, I think I'm not going to do because I'm bound to be wrong, but tens of percent, I think over time is where this market can get to.
Fantastic. Let me ask you, Lisa. I was at IDF last month, and one of the Intel executives there said the role of the CPU in terms of power consumption is vastly overrated, and whether you have an x86 CPU, probably an Intel x86 CPU, or an Arm CPU, doesn't really matter because so much of the power in data centers gets used up- In keeping the DRAM running and running the disks and running all the networking equipment. They didn't think that it really mattered whether switching from an x86 to an Arm would make that much difference in terms of overall power consumption. What are they missing over at Intel that you guys see at AMD?
Yeah, I think that's really a traditional view of what the data center looks like. The real answer is that the data center is changing, the workloads are changing, and the ability to dramatically change what the cost of running a data center as well as the CapEx expense is really unlocked when you do things like put small efficient CPUs like Arm together with our fabrics and systems that don't need as much network topology and really optimize for various workloads. I think that's exactly the problem that we're trying to solve and the exact thing that Arm together with our fabric and the ecosystem can solve.
In other words, it's not just that you're going to be reducing the power of the CPU or the power consumed by the CPU, it's always confusing talking about those factors, but also that you're going to be able to reduce the amount of power consumed by the other elements talking to the CPUs. Is that what you're saying?
Well, absolutely. More importantly, you can reduce the number of nodes that you need from the system to the network. That really changes the dynamics of the economics of the data center.
Interesting. Okay. Well, speaking of the mechanics of the data center, we're all chippy people or software people, there's one guy on the panel who actually spends a lot of time getting his hands dirty with this data center kind of operation. Jimmy, at Dell, that's what your business is, right?
Yes.
Is fitting out these data centers. One of the things that I understand is that people who are running data centers today spend almost as much on power to run the servers that are in those data centers as they spend on acquiring that equipment in the first place. My question is, if we move those kind of workloads to Arm, how does that change the equation?
Well, that's actually somewhat dated. I think you're talking about an era where the power utilization effectiveness of a data center was about 2.0. In modern hyperscale data centers, it's not uncommon to see power utilization effectiveness to be 1.2, 1.1, or even lower. What that means is you're not using up all this power on things other than the IT equipment that we once were. That being aside, though, what that does is that means that all the focus is on the power consumed by the IT equipment. Of the IT equipment, the processor is overwhelmingly the largest consumer of that. Anything we can do to give better power, better performance per power, better performance per watt in a data center is really important. This is especially true if you look at the expected growth rate that we're going to see.
Just the amount of power we're going to need to power these data centers is going to be enormous, so anything we can do to give us a better power footprint is good.
Fantastic. Okay. Now let's change a little bit, talk about some of the software implications of all this. Let me turn to Jay. You guys run large data centers, but you also build lots of software. You've got a lot of software, I suspect, that's now tightly bound to x86 kind of architectures. What's in it for you to make the investments to move to a different ISA, and is it really worth it overall?
Yeah. I think touching on a couple things that were said earlier, for us, it's always about, obviously, power and cost reliability, but also modularity for us is really, really important, right? What I'm excited about, where the industry is one is I think we need to stop thinking about enterprise and HPC environments and sort of porting them back into the cloud. We need to be thinking about our hardware designs across the industry, and basically optimizing for these really, really large-scale kind of mega data center presences. Now, the thing that's happening in these data centers is the workloads are just tremendous, right? I can say that if you all use Facebook, you're used to going to your homepage or to your timeline, and you see one webpage.
There is hundreds, if not thousands, of different services and servers that are involved in generating that one page for you. Some of these things are CPU-heavy, some of them need a lot of memory, some of them have a lot of disks, some of them use all of the above, maybe network-intensive. For us, it's really about trying to not just save energy, that's good, but really what I want to be able to do is, from a modularity perspective, being able to change out and optimize the hardware, and we'll make the investment in the software if the TCO is positive for us, is being able to change out our infrastructure at a right scale or right cycle. I don't want to be bound by an every three-year upgrade because that's what the financial plan says.
If there are things that are coming in the market, I don't want to have to buy a bunch of new stuff and then deprecate the old stuff. What I'd like to be able to do is go pop out those parts, those SoC modules in existing infrastructure where I don't have to touch the network, I may not have to touch the storage, I may not have to touch things like disk or NAND. Maybe it's that I just change out the processor. I change out the processor and the NAND or the NAND controller, and the NAND, and the DRAM as well. Being able to at different times change out the network as well. We'll make the software investments to make this all happen, right?
For me, it's about flexibility, and this modularity is a key part of why I think where we're going, what we're talking about today is exciting.
You mentioned that there are some tasks that require powerful CPUs and others that could get by with lighter-weight CPUs. Do you have any feeling for the mix of tasks that need that kind of big bang that you get with a real x86 versus what you can get with an ARM-type core?
Yeah. I think that's actually, in some ways, maybe orthogonal to the real issue that we face internally, which is, one is, I don't have a real percentage there for you, but I think for us it's like today you have the big, powerful x86. There's very little software outside of sort of HPC environments, and heavily virtualized environments that I think take advantage of these very powerful CPUs, right? Also you try to clock these down or limit the power, you play all these tricks, and then you introduce latency, and you introduce other operational issues, when you do have a spike, or you do have outages, or you have other new software releases. I mean, at Facebook, the product literally is changing for 1 billion plus people every day. We are changing the front-end code.
In fact, as I speak right now, the entire site is rolling for 1 billion users, and we do that once or twice a day. Our product is changing all the time, and the workloads that go behind it are changing literally every hour.
You don't have to stop and have Windows Update take care of all that stuff for you?
No blue screens for us.
No rebooting and all that?
Yeah.
Speaking of blue screens, yeah. Software. Software's always a big issue. I remember in the early days when AMD was out there hyping AMD 64, and the whole ecosystem was still on 32-bit x86, and AMD was beating everybody up, "Move to 64," blah. It took forever. As somebody who's a purveyor of the operating system software, what's in it for you to move your software system, especially as it exists in the 32-bit ARM world, to 64 bits, and how are you going to go about getting that to happen?
Sure. Things are moving a lot more quickly now than they were 10 years ago when x86-64 first came out. If you look at a lot of the drivers of innovation today, it's coming from the bottom up. It's these movements of the Facebooks of the world, the Googles, and things like that. They're really driving innovation in the cloud infrastructure landscape that are moving into the data center. We at Red Hat, we're really all about choice and innovation and flexibility for our customers. Through the collaboration of consortium of industry partners, the broader community, we're able to take developments of these technologies and get them rapidly to market. Take, for example, our cloud infrastructure.
We already have ARM-based versions of our Fedora open source operating system, presently available for people to develop on today because it's really about forming a foundation of layers on which they can build. We're porting and developing a 64-bit version of the Java runtime environment. We've got all the components already in place for the traditional LAMP stacks. In fact, most customers and users of the cloud, they don't know what's on the back end. It's used for a lot of heavy analytics. Like take, for example, the Giants win the World Series, and they want to run all these models to get the right T-shirts at the right spot. Or there's a hurricane ripping up the East Coast, and they want to get water and batteries in the right stores, or get the right types of personalized queries for business propositions.
These are the very workloads that are becoming the foundation both for on-premise clouds and in the data center clouds. There's really a lot of opportunity here for really fast-paced innovation, and it really is the combined power of the open source community in conjunction with our hardware and software partners that are really going to enable the marketplace to move quickly to capitalize on these low-powered hyperscale environments.
One of the things I heard you say is open source is a big part of this.
As I go back and think about what was going on when AMD was trying to convert the world to 64-bit x86, a lot of that depended on Microsoft getting their act together, and the Windows 64-bit versions seemed to keep getting pushed out. Are you saying that because so much of this is open source, that it's going to happen quicker this time?
Absolutely. In fact, we already have 32-bit versions of the ARM port. ARM has been supporting Linux for many, many years, and it's really been this grassroots, bottom-up development. For example, it enables our partners in other companies to participate, whether that's through doing just basic hardware enablement, to doing things like Java runtime environments, to being able to optimize GPU and other floating-point units, to be able to really do more. It's really empowering to let our partners, customers, and the community to drive the technology in places that no one company, because ultimately, we all have our own different priorities and our own different drives and motivations, and it's really the power of open source that's really fueled cloud computing, which is taking over the data center today.
I guess that open source also would make it easier for people who have codes that are based on x86 to migrate them into an ARM environment.
Absolutely, because they can port at their own rate. The other thing, one of the benefits of open source is that the people can drive their own innovation. They can experiment themselves because it really is, as I mentioned before, it's a bottoms-up effort because a lot of the latest technologies are coming out of the social media space, things that are coming out of the mobile front, as well as innovations like in the financial services market that all work very closely together through open source.
It's not just us. Red Hat, although we're building the platform on which this can be built and layered, it really is the combined community that really brings it and makes it pop in the industry.
Jay, how about what you're doing at Facebook in this regard? Are your toolchain set up so that you can easily convert blocks of code or specific functions from one architecture to another?
We have a lot of different systems, for us, it's about picking the right ones where this makes sense, and then putting the team together to make that happen, right? If it's where the TCO and the model makes sense to adopt a different architecture in the web servers, then we will set up a project and build out the toolchain that makes sense. We've done this across the board, not at the processor level because we're mostly x86 everywhere, we've done this everywhere else. In the network and various efficiency stuff where several years ago, we were running our front-end code in Apache, and that was having its performance and cost implications for us.
We basically put together an entire team that wrote its own compiler, and now we have our own virtual machine and compiler that runs all of this code that our product teams build and develop.
Is that compiler x86-based, or could you retarget it for Arm?
Both, yes.
Oh.
It is x86-based, ongoing work to retarget it for whatever we need to.
You control a lot of your own development tools, which would make that easy.
I think that's important, right? We've built up the team and the expertise that we can essentially start at the dirt, where we build our data centers. As you know, we've built our own data centers in Oregon and North Carolina and now in Sweden, we're literally starting at the ground up and building these very highly optimized, efficient data centers where we can minimize all the cooling and all the overhead there. Bringing it up through the hardware, the network, and all the software as well.
Interesting.
If I could add one point to that. As much as people want to roll their own, one of the key things, and this is a big challenge for fostering the Arm ecosystem, is to be able to have a common standard platform on which ISVs and others can layer their applications. One of the most interesting things that I see developing is there's industry consortiums and other collaborative efforts forming between hardware partners such as AMD, as well as the software vendors, to be able to develop a standard commodity-based Arm platform upon which you can build the operating systems, the middleware stacks on top of that. I think that the participants see this together, and we're really trying to work hard to get Arm into a standard platform.
Hmm. How far do you think Arm is from having that kind of standard platform at this point?
I don't think it's not really Arm. That's the beauty of it. It's the industry together, because no one participant can come out with this. It's really the cooperation across the board here, ranging from the chip manufacturers through to the OEMs, through to the software middleware providers, and ultimately into the application tier, that that's what's going to bring it out.
That doesn't happen all by itself. Somebody's got to be driving that.
What's been interesting was, in the early days, we really had to push that. I'd say in this space, that's totally got a life of its own now. We're seeing our partners top to bottom, making contributions, wanting to get involved, and wanting to push this on behalf of everyone. It has totally got a life of its own these days.
Lisa, you started out this session earlier by talking about how AMD was going to contribute to that 64-bit Arm ecosystem. Can you give us a little color as to how that's going to happen?
Absolutely. This is one example, right? Gathering the key influencers of the ecosystem. We're working very closely with a number of the software vendors as well as the OEMs on how we put together the overall ecosystem for that. You'll be hearing more at Arm's event coming up over the next couple of days. I don't want to pre-announce any of those things. There are a lot of efforts to bring hardware, software, systems, OEMs together so that we get the entire framework of what's needed to make this successful.
Great.
It's a multi-year, multi-partner effort.
I'm getting a signal from offstage that they're going to bring out a hook soon if we don't stop. I want to thank you all for your participation here. It's been educational and informative. Thanks.
Thank you very much, and thank you to all the partners for being here. With that, I'd like to invite Rory back on stage, and I believe we're going to take a few questions.
Lisa. I'd also like to invite up Andrew Feldman, who's our Vice President and General Manager of our Data Center Server Solutions team. Andrew. He's also responsible for the SeaMicro Freedom Fabric. Andrew.
Freedom Fabric. Freedom Fabric.
Freedom Fabric. Okay. Questions. Right over here.
John, I have the microphone here. Sorry.
Oh, we have the mic. We'll start here, then we'll go there.
Okay.
All right, we're coming next.
Hey, Rory. It's Glenn Yeung from Citi. My question's kind of three parts. Lisa or Rory can answer. You make the case that AMD is going to be the first with both x86 and Arm on 64-bit. I guess the first part of this question is why is it relevant to have both? Is there somehow a synergy there? Second part is, are you going to be using Arm's technology as itself, or are you being an architect and develop your own underlying core? The third part is there a cost to doing that that we should be aware of?
Glenn, it's good to see you. From a standpoint of why is it important, the ecosystems are going to continue to evolve. Obviously, the point about the Freedom Fabric is very important in terms of applying the server technology to the workloads where it best fits. We also think that this is going to be a very interesting growth phenomena over the next, say, three, five, seven years. As that occurs, we want to be a leader in that space. Because of the uniqueness of the design, we think that Arm 64-bit will play an important role. x86 won't go away tomorrow, won't go away over a long period of time. This is an important opportunity for us to get in there and lead and show the direction, then pull it together with the SeaMicro solution. The second question?
Yeah, Glenn, let me just add some thoughts to that. We are announcing an overall partnership with Arm. I think we've previously announced that we were licensing Arm's Cortex-A5 for security. They've joined our HSA Foundation. Tomorrow, Arm will be talking more in details about some of their new 64-bit core technology. We are a licensee of that technology, and that will continue to expand over time.
The third one was cost associated with it.
Yeah, the effect in 2015.
When we talked about the operational structure that we were creating with the financial reset that we did last week, we considered, as I mentioned, Glenn, in the earnings announcement, where we were strategically investing and where we were going. We've built that into the model going forward.
Thank you.
Okay, let's jump over here.
Question over here.
Yes, please.
Thank you. Shawn Webster from Macquarie. Thank you for hosting this. A question related to Glenn's is on just general affordability. You said it's fitting within the envelope of your guidance, taking into account the reductions that you have in mind. It sounds like this is going to be something that'll require maybe multiple design teams of investment, I'm just wondering if there's other areas within your portfolio that you're thinking of disassociating or pulling back from. My second question, this is a two-parter, is there anything that you would be doing in the Arm space that you can leverage down into clients, like getting into Windows RT environments? Thanks.
Yeah. The first question was in regards to the cost again and the structure?
Is there anything you're pulling back from?
Yeah. No. What everyone should think about is from an investment level, one of the ideas is the idea to reuse the IP that we've created across our portfolio. We have the graphics IP, we have a number of capabilities across our portfolio, that's what creates the richness of AMD. What we want to do is now knit that across, remember we talked about it in the past, SOC15, System on a Chip, reusable IP blocks, agile architecture. This is the idea that each of our components is reusable across systems and across chips. The processor just becomes an IP block in that solution over time, and that processor then slides in and out of the design with the bus communicating across the various IP blocks. That's why we think that's a part of that ambidextrous architecture.
More importantly, we think it opens up a whole another level of growth opportunity to apply our technology. Okay, the second question was?
Leverage into Windows RT environments.
Oh, client. Do you want to take that, Lisa?
Yeah. Maybe the way I would talk about the portfolio is we are definitely targeting our investments at the higher growth segments. The highest growth segment in server is definitely this mega data center, cloud data center workloads, and that's where we're going to be targeting our investments. In client, it's certainly new client and mobility, and that's where we're going to be targeting our investments there. I think Arm has a lot of traction in client for sure. I think Arm has a lot of capability in various dimensions. Today, our relationship that we're announcing is around servers, but we certainly believe that the fundamental underlying principles in terms of microprocessor design, in terms of system-on-chip capability, in terms of system integration, are applicable across multiple markets.
Great. Thank you. Question? There's one here, then we'll go over here next. You're up.
Hi, Mark Hachman with Slashdot. There's been a couple of predictions, guesses as to what the Arm processor will represent in terms of the data center market, 5%, 10%, and so forth. What's your assessment? Then, as a second question, I wanted to know if you could quantify, say, the power savings that the Arm processor will offer you as opposed to an Opteron processor.
Sure. From a growth standpoint and opportunity, we think this will be one of the fastest-growing segments, obviously. In terms of its raw potential, it's hard to put a number on it at this point, but I put it in well into double-digit percentages over that horizon, 3 to 5 years. I do think, again, it's going to be the fastest growth, multiple tens. Andrew, you've had some thoughts on this, and you designed this into the fabric. What do you think about when you think about the mix?
Well, I think the movement towards smaller, more efficient processors is across the board. You heard Jimmy talk about it. You heard Jay talk about it, both from the largest OEM and one of the largest consumers of compute in the world. So you heard that from the builders and the buyers. I think in answer to the second part of your question, I think you can build clusters of Arm-based servers that use a third the power, between a third and a half the power. I think it takes work. I think that's the work that will be worked on with the software community and with companies like Dell that Jimmy represents and teams like the one that I came to AMD with and SeaMicro. That work's on the table and available to us because of this technology.
That new challenge that's on the table is made possible by this.
Great. We got one right here.
Yeah, right over here. You have two.
All right, first
Rick Merritt from EE Times. Three things weren't clear, just specifics. Are you going to design your own cores as opposed to using Arm's? Are you doing merchant chips as opposed to systems of SeaMicro? The third one was, is this one chip or two chips as it is with SeaMicro today?
Yeah. Let me answer those. In terms of our relationship with Arm, as I said, we're announcing an overall strategic relationship with Arm. Tomorrow, Arm will be announcing some details of one of their 64-bit cores, we will license that as part of the announcement. In terms of what will we sell, we will sell merchant chips. We will certainly work with the leading OEMs of the industry, as well as sell systems under the SeaMicro brand as part of AMD systems. The third question was?
One chip or two?
It will be integrated on a single SoC.
Thank you.
Great. Next question right here.
Dean Takahashi with VentureBeat. Sort of related to that question, how much flexibility do you have in designing exactly what you want to design? Did you want this broader architectural license, which would give you that complete freedom to do whatever you want? Or is there some limitation on what you can do?
Yeah. It's fair to say that our relationship with Arm is expanding and broadening over time. We have a lot of 64-bit microprocessor expertise, and we're going to apply that liberally to both x86 and Arm.
Question right here.
Jag Bolaria with The Linley Group. My question is for Lisa. Hey, I know you guys have Opteron six, four, and three. Do you see a role for a lightweight x86 core, potentially, let's say like a Jaguar, towards low-end, low-power, power-efficient SKUs in the server space?
Yeah, I think that's a good question. The question was regarding we have Arm as a small, efficient core. We also have x86 cores as well, our Bobcat family series, Jaguar family series. We do see a place for x86 small cores as well. I think the key thing is the ecosystem will drive the small core adoption. The fabric is absolutely critical to that, and that's the piece that will be common among whether you're talking about small core or big core, and it really is about optimizing for the various workloads as they come true.
That's one of how the question that was asked over here earlier relates to your question is, as you amortize blocks of IP across x86 and Arm, something like the fabric, the Freedom Fabric we designed is the only fabric that supports both x86 and Arm-based processors. It supports both Ethernet traffic and disk traffic. It's designed exactly so we could take one piece of work that we do as engineering and use that piece of work to underpin best-in-class small core x86 and best-in-class small core Arm. That's really one of the tenets of the strategy is to use technology that we invent, in this case, the Freedom Fabric, across the portfolio.
Good. Question in the back.
We got a question in the back over here.
Sergis Mushell from Gartner. Quick question, Rory. On the reusable IP blocks that we've been talking about in the SoC world.
Sure
Could you give us some understanding of what timeframe we're looking at for one? For two, what are those IP blocks more targeted for? Communication, industrial, gaming? What are the IP blocks that we're looking at to build an SoC product family?
Sure. We've been on this journey since many of the management team has been assembled. We've been working on moving toward IP blocks, SoC methodology since back in last year. The idea again is how do we create that set of all of our IPs so that we can reuse it across not only our particular architectures of the ecosystem, but more importantly, across chips. This is something that's fundamental in terms of reducing testing, reducing all the complexity. It's easy to create a complex environment that's every interface has to be retested and tested because of every IP block coming together. What we want to do is move to a structure where those IP blocks have defined interfaces that they allow to be assembled in a much more rapid pace. This will reduce our test cycles. This will reduce our cost.
One of the points to Glenn in the beginning about how you reduce the cost of the overall execution of AMD. This is a fundamental case, as well as moving to a standard set of processes, industry-level mass layers. All of these things make huge difference in terms of the overall cost, moving away from everything custom to reusable. This is something that's occurring now. You'll see it in the Kabini product as it moves to a full SoC chip. You'll see it into the following chips. We're already moving in this direction across the portfolio, and this is a set of journeys, of steps across our product segments to apply that, and that's the foundation of where we're going. Did you want to add a little bit more on that, Lisa?
No, I think you said it well.
Good.
One question from the back.
Okay.
Gil Russell, Bright Side Analytics . A couple of easy knockoff questions here. You announced that it is a 2014 launch, and I would like to know when you expect the return on investment stream to kick in on that launch. Also, have you formed the engineering team? The third part is APU. Where does that segue?
Sure. Go ahead.
Yeah. Okay. Let me take those. 2014 launch, we believe that we will be at the leading edge of Arm-based servers out there in the industry. The market will develop over time. Clearly, we talked about how we accelerate the ecosystem, this will be a couple of years to really get that market to develop over time. Next question was regarding?
Are the engineers.
Oh, the teams. Yes. The engineering teams have been well underway. I think this is one of those things that Rory and I have been working on for quite some time. The engineering teams are executing and working very well with the Arm team.
The last one's APU. Segue.
Yeah. The APUs are really, I would say, orthogonal to some of the conversation because when we think about APUs, I think about the graphics compute as another IP element that we can integrate onto chip. When you think about workloads that can really take advantage of the graphics compute, like some of these media streaming workloads, some of these high-performance computing workloads, I think whether you're talking about x86 or Arm APUs, you'd be able to dramatically improve for these specific workloads, the capability.
We have a question over here.
Okay, I see you.
Yeah. It's Damon Poeter with PC Magazine.
Hi.
To Lisa's point that this is an inflection point, I think is almost an understatement given the fact that the storied x86 house is moving to work with the new architecture. It almost raises.
Pleasure
a lot more questions than are probably possible for you guys to answer. One, just looking at my notes, is we're talking about data center today. When do we start talking about client stuff, and all the way down into mobile with Arm and AMD and x86 and Arm? From AMD's point of view, are there other architectures that you're looking at? Are we looking at MIPS? Are we looking at other things besides Arm for possible future collaborations and stuff like that? You talked about developing developer ecosystems. We've seen that's even tough for the Intel and the Nvidia of the world with all that they've invested into CUDA and their proprietary products. What's your strategy there going forward? That's a lot of stuff to digest, but go ahead.
Yeah
answer whatever you can.
I'll take the first one real quick. From a standpoint of the inflection point, you're 100% right. This is a huge inflection point. Don't misunderstand it. We're a huge x86 business. We'll continue to leverage that architecture. That's an important piece of our business. There's no doubt that the cloud begins to change everything. It disintermediates data and application from device. Device converges, information moves seamlessly across device type, and the cloud fundamentally moves all the data there. We're going to attack that in two ways. We're going to attack it in the server to create the low-power solution and architecture because, as you heard our panel talk about, it's about power, it's about space, it's about huge amounts of data and workload because the applications and data are now out of the client.
The second part we'll do is across the architecture and the client space. We'll start to knit together those because graphics matters. That's why our IP is so fundamental to our strategy going forward. Take APUs across those client devices, drive down the power lower, and then allows us to move into adjacent spaces like embedded, like semi-custom. This is all part of the growth strategy that we see driven by this market phenomena. I think that's how we got engaged with SeaMicro back in the day because they saw that same idea, and how to knit that solution together is, we think, just a fundamental step. Again, this is not a debate about x86. That instruction set has lots of applications and data, and it's going to play a huge role in our future. More importantly, how do we knit that together?
Arm is going to open up so much new opportunity for us, and it's going to allow us to expand across that. The market will ultimately be the final arbiter, we're going to be there to help lead this transition, disrupt, and create a new future.
Maybe Damon, let me just add to that because I think what you're asking is bringing together several pieces of the AMD strategy. I think what we've said is we are a company that has tremendous innovation at our forefront, so high-performance microprocessor design capability, high-performance graphics capability, leading-edge graphics capability, excellent systems capability with software, with our heterogeneous software solutions, and then with our SeaMicro fabric. When you talk about how do you move an ecosystem, I think the way to move an ecosystem is really to be open in how you do it. That's what makes the partnership with Arm so compelling because we're able to take all of sort of the crown jewels and the IP of the company and apply it to an open ecosystem that is thriving to develop.
Whether you're talking to the OEMs like HP and Dell, or you're talking to the users like Facebook and Google and Amazon, there's a desire for that integration capability. It gives us an opportunity to really use the tremendous IP assets that we have to these higher growth market segments. I have a lot of confidence in the ecosystem building. One example I'll give you is our heterogeneous systems architecture foundation that we announced in June. We announced with Arm as a founding member, Imagination, Texas Instruments, ourselves. Since then, Qualcomm has announced that they've joined. Samsung has announced that they've joined. The ecosystem really does go for open systems.
One more?
Hi, James Niccolai, IDG News. Lisa, you said earlier, we believe we'll be at the leading edge of Arm-based servers. You said something about the fabric is the secret sauce that no other server vendor will have. You are still planning to license all this, I presume, or not?
I'm sorry, I didn't catch the last-
Will you be licensing this fabric to other server makers, or is this an AMD technology going forward? I thought originally you said you planned to license this fabric, and other companies would be able to build these servers with the SeaMicro technology. Has that changed, or is that still-
Yeah. What we've said about SeaMicro is we're using it to both advance the systems knowledge, and certainly, Andrew and his team are doing that. We're also working in partnership with the leading OEMs to talk about how fabrics are integrating into their systems. In our own capability, we will be integrating them on chip. Whether there's actually licensing, I think that's a broader statement. I think the key is the fabric together with our microprocessor technology gives us a differentiated advantage for these Arm-based systems. Maybe, Andrew, you want to add a little bit on that.
I'd just say that I think what we're announcing today is a use of our technology, our IP blocks, across a new area, across a new instruction set. I think whether it's working with Jimmy and his team at Dell or with other OEMs, we want to work with them to build phenomenal product that use their advantages and their IP and their insight. If that involves our chips that have the Freedom Fabric embedded in it, we should work with them to help them build extraordinary solutions for their customers. They have a set of capabilities and advantages that are different than the ones that the SeaMicro group bring to AMD. The trick is to use our key strengths and marry them up with those who have different key strengths to build solutions that people like Jay find compelling.
That's what we're out here trying to do every single day.
Okay. With that, we'll conclude the Q&A. I want to finish up the way we started. I want to thank you for being here on this historic moment. I have a passionate personality, and I believe in this inflection point. This is going to occur, and this is part of the DNA that makes AMD who we are. Our engineers look for what's next. How can we create innovation? How can we inspire our customers and partners to think differently? How can we create that next inflection point? How can we change the game? Our partnership today that we announced with Warren East, the CEO of Arm, is a powerful one and an important step forward in this vision of changing the industry and capturing this inflection point.
Warren made every effort to be here, and I think it says volumes about the man who does an interview at Heathrow in the back of a taxi because he believes in what we're doing. This is the first step in a series of steps to build a new and different AMD. This is the future we want to create, and this is the future our customers and partners see. This is our time to lead and apply our engineering skill and intellectual property to make it happen. Thank you all for taking part in this historic moment as we introduce our ambidextrous architecture, leveraging both x86 and Arm technology to unlock the future of data centers in the new cloud mega data center environment. Thank you all. Appreciate it.