 Okay, we're back here live in New York City for a special CUBE presentation of HP Moonshot Announcement. I'm John Furrier at SiliconANGLE.com. It's theCUBE, and we're going to explore and continue to do a drill down on kind of what is all this mean. Moonshot obviously changing the data center dynamics, changing the world around computing, software, software-defined data center, and all the exciting mobile and cloud computing action. I'm John Furrier with SiliconANGLE.com. I'm John Furrier with my co-host. I'm Dave Vellante at Wikibon.org. Brian Glinsman is here from Texas Instruments. He's the Vice President and General Manager of the Processor and Embedded Processing Division of TI. Brian, welcome to theCUBE. Well, thank you, Dave. Very happy to be here. It's an exciting day. I'm really excited at what HP's done and what it opens up for the entire community. Yeah, we've been seeing all kinds of innovations today. I mean, this notion of lots of little processors and systems on a chip and this idea of tuning the system for the workload as opposed to saying, okay, this is my workload and I gotta change it to match the system characteristics. Really seeing a sea change of the thinking that goes into infrastructure. So what's TI's perspective on this? What's your angle on this opportunity? Well, TI has historically been very strong in embedded processing where large OEMs would build proprietary architecture. And by combining our technology with the Moonshot Program, we enable customers to take our technology and put it into a cartridge, just like this, where in this cartridge, you have enough horsepower. Sure. This cartridge has the equivalent horsepower of an SOC, but what you get out of it is you get 16 ARM processors and 32 DSP processors, which is enough to do, for instance, if you were thinking in media processing, it could do anywhere from 2000 to 10,000 channels of converged VoIP in one little blade and the power and cost and space savings is amazing. So compare it to what would a conventional infrastructure take? So I've got this system here. What would a conventional infrastructure take to replicate what you just described? Well, it's all come a long way over the last 15 years. We've been designing SOCs and moving them up. We used to be very DSP centric. Now we bring in all the other types of processing with ARM and packet processing. And the HP Moonshot infrastructure enables people to get to market very quickly with not only one device from us, but from devices from the entire ecosystem so you can build a server that makes the most sense. In terms of what this does, typically the proprietary systems are large ATCA blades that are custom built that consume 200 watts per blade. And while the capacity could be achieved the same as this, it's a very expensive proposition for our customers to do that. And this way they can use a common base and their time to market is reduced by years. So you call this Keystone, right? That's your platform? The SOC is Keystone and what we mean by that is we designed, we're on our fifth generation of SOC and what Keystone means to us is giving you the tool set of ARM, of DSP, of acceleration, whether it's for radio or media or packet processing and blending it together where you can, the user, the end customer can get entitlement out of the silicon. And by enabling this platform, we're opening up markets we've never seen before because those customers weren't going to build a custom server. Now, just by populating cartridge blades they can have their choices. What markets are you opening up? Be specific, because that's really important that we've been talking earlier about kind of enabling platform of this disruption. Well, sure. I mean, I think you had CTG on here. They do a lot of mathematical computations and so what we're looking at here is being the math coprocessor but easy to use. And what that enables is they can do whether it's in a small truck or whether it's in a large farm, far more computations per watt, far more computations per square meter of floor space and it enables them to go. Other areas that we've seen a lot of traction, financial data modeling for Wall Street in genetic mapping, genome, anything to do with high performance. So when you say new markets, you mean that you're enabling the customers because of the footprint and the cost savings on the energy, they get more servers, thus they can do more relative to their app, whatever that app is. And in that case, it's computation farm or whether it's real-time transactions, whatnot, right? Sure. I mean, in the day of a generic server, the X86 architecture has been really good and we don't see replacing that. But when you have a purpose-built server that can meet your needs because you do a lot more media processing or a lot more math processing, this type of platform will enable you to do a far more efficient power and space and cost. David, I always talk about it ever since, you know, we've been covering Oracle Open World now for three years, going on a fourth year about purpose-built as being kind of like, you know, proprietary. And if you look at it that way, you can argue that. Here what's interesting is you have purpose-built capability in an open environment because you have an ecosystem, you have some software programmability. So that's kind of how we see the world going. Do you agree with that? And if not, what? Yeah. So, and not the Oracle part, but they open, you know, they come in on Oracle. So I think the Open Platform, bringing it together so you have, HP takes care of the infrastructure, they take care of the booting, the security, the environmentals. And now you have the ability to bring multiple cartridges of what you, what the customer desires to make their best server, I think is a huge advantage. For my traditional customer base, what it brings is time-to-market reduction of two to three years, because they don't need to go out and revalidate Neb standards on their hardware, reliability, it's done for them. It's like having a UL label on a device. Yeah, and it's just, it opens up to so many more smaller customers where, you know, unless you were going to be selling a billion dollars worth of product, building these type chassis is really expensive. And so now you've got a standards-based chassis that whether, hopefully, it's TI, but you know, whether it's one of the other eco-vendors out there that enable a technology that you just, you know, the X86 architecture is great architecture, but there are other architectures that get you more horsepower, more performance for purposes, and what we want to do is enable both. Now you guys have always been good and specialized, right? I mean, you've never sort of tried to take on the, well, maybe maybe always. I wouldn't say never, we did in the old days. Yeah, that's true, and I remember, I'm old, so I do remember that, but then you learn your lesson, I guess, is how I would say that. But no, the interesting thing is you've learned how to thrive, you know, by tucking into these various places. This moonshot really opens up Pandora's Box in a good way, doesn't it, for you guys, and opens up huge new markets. It does, and when you look at where things are going, you know, you've got data manipulation, but more and more, everything's becoming about analytics, whether it's analyzing user trends, whether it's analyzing video, whether it's analyzing voice, whether it's however you want to do it, and to be able to blend in large data farms with highly computational matrix math gives you a huge lift in what you can do. I mean, if you move forward, instead of having a video camera that just says there's someone moving, having a video camera that says there's someone moving, they're in the 7-Eleven, they've got a hood on, they've got a metallic object in their hand, please call somebody to look at this. That's a huge step forward, you know, just look at everything we're doing whether it's around cars driving themselves, bringing all that capability into the server so that you can utilize all the specialized embedded processing, again, not just here with our cartridge, but with any cartridge, it gives you a very large base of things we've never thought of, and it ultimately will result in much better products, much lower power, much lower cost. So take that example that you just gave us. Is that in production today? No, we have customers working on it, it's very high-end today, but as you start cost reducing it. Yeah, they say how much does that cost, right? As you start to cost reducing it. I mean, the devices that we're putting into this server, they can take a radar map of the earth and turn it into a video map, right? That's what we use them for. They can be used for a Cloud RAN solution where you put the entire base station in a server. It's a vision that people have. Who knows which way things go, but it enables anything to be put in a server that can be put in embedded processing. So, but coming back to that example, are we five years away, 10 years away, 50 years away, not 50, of course we don't think out that far, within a decade to have the cost be at a level that is sort of consumable by an average business? I think we're getting very close to be consumable by the average business and that it really comes down to getting the development base out there and getting the ability of the developers to, you know, purpose-built is a little bit harder. It's a smaller ecosystem. Requires some talent, I mean. It does, but at the same time, the benefits of it are tremendous in what you can do in a power cost area position. So we want to do both of those. I don't see either one going away or winning. I see the- They're not mutually exclusive. I mean, we see custom definitely the way to go, but it's not necessarily purpose-built in the way that people used to kick purpose-built as being proprietary. I mean, even Oracle's moving to purpose-built open to some extent using industry standards and stuff to get Java. So, you know, we see that. People want the best solution for their offering. And that's why I like the software angle on this. And I think, you know, the software developers, the lambstack guys and other folks out there, not the software that's on the devices, that still needs to be innovated and you guys are doing that. So, you know, the question is, yeah, no, it's a no brainer. Mobile is causing more action, forcing pressure on the infrastructure guys, no doubt. Where are we on that in terms of this kind of new hardware hitting the scene? Honestly, I think it's really the top of the first inning in my opinion. It's just shipping. Moonshot, 15 months ago, was first generation. This is the second generation. I mean, I call it first generation because it's gonna have to use shipping code. It's early. How much more work do you see happening quickly? And how accelerated do you think it will be? Well, I think, at least from the cartridges we're providing, what we'll see is very high computational folks moving quickly because they have large teams, they have a big need. And some of these things, they have a power space budget of 8kW or something of that nature. And if they can triple the performance by changing a cartridge out, well, that's great. So they've got some confines. But as we move along, I mean, there's no question the mobile world is bringing a lot more traffic, a lot more video, and a lot more analytics to the party, as well as just base stations, right? And so, all of these things could be done in a Moonshot server. In many cases, I think we will start seeing them be done in Moonshot servers or that type technology. There still will be purpose-built boxes. I mean, it's a blend, obviously. That's just the way it is. I mean, that's not the wrong of having purpose-built. It may happen in the mainframe days, too. Next five years, just your last question, just your personal perspective, given the historical views you've had in the industry and now kind of where we are, we've always talked about this as an inflection point transformation, et cetera, et cetera, you're seeing it on the consumer side now in the emerging tech and enterprise. Up and down the stack. Next couple of years, where are the big things gonna be happening in your mind? Just on the general landscape of the industry. What do you see as really put in the pressure of the drivers for more change? Well, I think, as I mentioned earlier today in the Moonshot launch, power consumption per function has to come down. No matter how you achieve that, Moonshot is clearly targeted at 90%, 85% power reduction. That has to happen because as you turn on more and more of the world to the things we're just used to every day, watching a YouTube video on your cell phone or whatever, the power consumptions of seven billion people do that is huge, so power is the number one thing we need to attack. But if you just, it's kind of like telling people you can lose weight by not eating, okay? Great, but at the end of the day, you know, that's not gonna be the solution, right? So we need to figure out how to do what we're doing today at 90% less power or 99. Are you saying this begins the strategy of this is low power? Me and me. Yeah, no, that would, no. Red meat, you know, more energy, protein? No, I'm just saying. I don't want to be the Adkins guy, we're in New York, we're in there. We need to, power is the number one way to address things. By enabling purpose-built, along with more general purpose servers, you'll see the power-hungry applications moving to purpose-built to save the power and to enable a broader ecosystem that gets us here quickly. I think it's an exciting time in the industry. Brian, thanks for coming on theCUBE. We'll be right back. This is theCUBE here, live in New York City for a special presentation of theCUBE. Extracting the signal from the noise at HP Moonshots, big announcement here, changing the game in the data center, cloud mobile, social, great low energy, high performance. Here's theCUBE, we'll be right back at this short break. Stay with us.