Intel's goal for Penryn: "Ramp and win as fast as possible"
San Francisco (CA) – TG Daily had the chance to sit down with Mr. Banias himself, Mooly Eden, for a face-to-face interview at IDF 2007. And Eden took the opportunity to explain why he believes that Intel has the edge right now – and why he believes Intel will keep it. Of course, that brought AMD into the discussion as well and Eden wasn't shy commenting on AMD's outlook.
Mooly Eden is a unique character at Intel. While he is into serious business - he is vice president and general manager of Intel's Mobile Platforms Group and credited with leading the development of the Banias processor, Intel's turnaround product - he isn't the typical manager you would expect in this position. It is as much pleasure to talk with him about your last vacation as about his everyday work.
Of course, we are looking forward to talking to him everytime we have the opportunity to, as we know that we get straightforward answers without having to try to rephrase the question a few times. We were not disappointed this time either.
TG Daily's Rick Hodgin sat down with Eden and talked about the upcoming Penryn processor, Silverthorne and Moorestown, Intel's relationship to its partners, future clockless processors as well as Intel's take on AMD's competitive situation.
TG Daily: There is a lot of talk about performance at this IDF, especially when it comes to Penryn and all the talk about processors these days is quad-core. So, simply asked, does a quad-core processor in a notebook make sense?
Mooly Eden: Yes.
TG Daily: Why's that?
Mooly Eden: Because I’m going to do it next year in the second half. So definitely, I’m not going to say anything that doesn’t make sense. [laughter] When I was asked if I believe quad-core will go to mainstream, I answered “no.” But there are two categories that might require quad-core. One category is gaming. No matter how much performance we give them, it’s not going to be enough. And some of the games are already so much threaded, such as Microsoft's Flight Simulator; you put four cores to the task and you will get much better performance. The other one is mobile workstations. [This market] needs the performance and it needs the mobility.
For these two categories, we are going to develop a quad-core for mobile and we are going to introduce it in the second half of 2008. Can OEMs take it down to mainstream, if you build the right product? Yes, they definitely can. But, [for the mainstream] I don’t believe there are enough applications to take advantage of four cores. Four cores will also consume more power and more power will hurt the battery life. So, if I hurt your battery life, I’d better give you something in exchange.
When the software and usage will be there, I believe [the quad-core] might go [into the mainstream]. But if you ask me if it will be 2009 or 2010? You’ll see mobile with quad core, you’ll see some people trying to get a Halo. But mainstream in general? I doubt it, at this stage.
TG Daily: That answer surprises me. Don't we always want more and isn't it the indication that we will be getting more and more cores in any of our computers?
Mooly Eden: Well, we can [put four cores in a notebook]. There’s nothing preventing me from doing that. What I will eventually do is that I will fulfill the request from the OEMs. But if somebody asked me as an expert: “Mooly, does it make sense?” I will answer: "No."
And, by the way, I was very unpopular two and a half years ago when I introduced [the 32-bit] Core Duo. People asked me then : “Do you need 64-bit?” And I told them: "64-bit is a marketing hype … by my competitors. A very good one.”
They wondered if I were crazy for not having a 64-bit Core Duo. But there are many things I don’t have. I said, “guys, there’s no software for 64-bit.” I didn't want 64-bit on Yonah. It did not make sense. Many were criticizing us until Steve Jobs gave up on the PowerPC architecture and went to Yonah. He went from 64-bit to 32-bit since he appreciated the performance and battery life more than 64-bit, because he knew that nobody was using it.
Moving forward, I believe 64-bit will start ramping. It’s the same thing with quad-core. We’ll have the technology, we’ll take it down there when we need to.
Read on the next page: Goals for Penryn, the fate of the desktop processor and future clockless processors
TG Daily: What are your goals for the successor of the current Core 2 Duo, the 45 nm Penryn?
Mooly Eden: I want to ramp it very, very fast. It’s a great product. If you think about it, it’s 420 million transistors. It’s got 6 MB of cache.
The interesting thing about Penryn is that when we start designing a product, we are trying to look at several factors. What is the transistor budget? What is the evolution? And where is my competition going to be? Then I’m trying to define something which I call “landing zone.” This is the budget, this is the number of transistors, this is the frequency, this is the performance in F-SPEC and everything. And we start designing the product.
Penryn was [originally] designed to compete with [AMD's] Barcelona in notebooks. The minute we found out Barcelona’s competitive data, we were wondering whether they were actually going to stay with their old technology [for notebooks]. And yes, their [upcoming notebook processor] Griffin [turned out to be based] on their old architecture. I believe we are going to open a huge gap next year. Probably much bigger than in previous years.
So, my objective is to ramp and win as fast as possible. We will do it very, very fast simply to open a huge gap over our competition. You will not hear the word “performance” any more from our competitor, because they will not be able to compete. They will tell you stories about many, many other things. But I intend to get that huge advantage on performance and power efficient performance. [Users] do not yet understand the following: Performance gives you much better responsiveness. If you’ve got great performance and you give me a task, then I’ll do the task, I’ll finish it early, and I’ll go to sleep.
Do you know the optimal situation for a notebook?
TG Daily: Sleep?
Mooly Eden: Yes, sleep mode. So, we designed the microprocessor to be in a sleep mode - a good microprocessor sleeps all the time. You speculatively know when to wake it up, you do the work and go to sleep again. So, with more performance, if you want to download a 60 minute song and compress it and put it into output, I’ll do it much faster than anybody else. But, not less important, not only will I do it faster, I’ll go much faster to sleep so I’ll save battery life.
So I believe simply we’ll try to ramp Penryn as fast as possible because it gives us a huge competitive edge and we are proud of it. We’ve got the capacity. We’ve got the 45nm. And we’ve got a stellar microprocessor. It’s a nice puppy.
TG Daily: So, with so much more performance in the notebook space and notebooks, according to your CEO, approaching a market share of 50% in 2009, the obvious question is whether the desktop processor will become less important or whether it may die, at least from a development perspective?
Mooly Eden: Let’s put it like this. The world for me is very simple. There’s two categories. There is mobile, and there are desktops that are going to be mobile. Not that I’m opinionated on this area, but there’s little use for a desktop for people who do not need it. However, the ability to question it would’ve been much more relevant five years ago than now, because if you look at Core - the Core architecture in desktop and mobile are using the same core micro architecture.
So, yes, the desktop will be replaced by mobile solutions, especially in the living room. In the “all-in-one,” and in all kinds of sexy form factors. Power still matters. Think about an environment where you don’t want the fan to work, because it’s noisy. It disturbs you in the living room. In cases like this, I believe it will overtake the desktop CPU. Overall you can see the trend both in desktop and mobile for low power consumption.
I believe mobile percentage will ramp, but there will still be place for desktop, because desktop, by definition and by nature, will be a more affordable solution. And you still have emerging economies which is still important [for desktop processors].
TG Daily: One key component about the whole power issue is clock speed. We have had recent discussions with Jerry Bautista, who heads up your Tera-Scale project, on clock-less CPUs. Is there any research done at Intel on this topic?
Mooly Eden: Definitely. First of all, I agree with you that the clock network can consume a lot of power. But for that reason, we’ve got something that is hugely extensive, “clock gating.” What we are doing is we are looking at the state of the microprocessor and whenever you do not use it, you stop the clock. You gate it. You will see a very complicated clock gating on the CPU and speculatively we try to stop clock on many systems. Let me give you an example. If you tried to get some data from the cache, and the data is not available and you cannot do some speculative execution, you have to wait for data. The whole CPU is going to be doing nothing, so why do you need to clock the execution units?
So, you’ve got massive clock gating. And we are working on it, and we are developing it, and we improve it as we speak. If you speak about something which is revolutionary, such as an asynchronous microprocessor, which means it’s not the clock that is everything, but rather what it is working with what it is finishing and sending, it will be a very complicated design. But for the near future, the next-generation microprocessor will still be a synchronous microprocessor with clock gating and power down features.
Read on the next page: Mobile Internet Devices, UMPCs, Moorestown, Silverthorne and the $100 laptop
TG Daily: I have to admit, the most interesting device we have seen so far at IDF was not Penryn, but an iPhone-like Mobile Internet Device (MID) that was showing during one of the keynotes. But there wasn't much detail besides the fact that it is base don a new platform called Moorestown. What is the idea behind this product?
Mooly Eden: With respect to this category I can tell you two things. First of all, it’s a fascinating category. We believe it’s going to be huge. First, in our opinion, the real usage that everybody wants is Internet connectivity, so Internet is the usage model [of the device]. Second, if you want to get a great experience surfing the Internet, you need an IA-32 Intel architecture, because the Internet is written around it. If you were to use an ARM [processor] or something similar, there will be many things you are not going to be able to see.
Another assumption, number three, is the development of the ecosystem of the world. People want to have this Internet. Always on, always connected to every place, which means we are speaking about miniaturization.
And assumption four, in order to make it grow very fast, you need to make it available at consumer affordable prices.
TG Daily: Are there any details you can provide on Moorestown? What are the differences between Moorestown and Menlow, which will be the first-generation MID platform?
Mooly Eden: The main difference between Moorestown and Menlow is that I believe it’s just the next generation of what makes sense. There will be much more integration, much more functionality, a much lower power envelope. When you are trying to reach what you know, you do it in steps. [Moorestown] is just more optimized for the sigma that you want to do, which inevitably results in smaller form factor in the same destination as the predecessor. It’s going to be a slick device.
TG Daily: We have been wondering at TG Daily, is the MID what the UMPC should’ve been in the first place?
Mooly Eden: I believe all of it was evangelism. We look at new opportunities as we learned, as we looked at it. When we start developing this stuff, and a lot of it, by the way, is the same, we’ll find out that, when we start talking about UMPC [ultra-mobile PC], the market is not huge. But we thought it was a market we needed to serve.
When we started working with it and we went to the customer, we discovered that there’s a huge opportunity: The MID – with IA-32, very low power, carry-on-you, etcetera, etcetera. We learned that this market is bigger, but learning is part of the crusade. Before you start walking you don’t know it. I believe we have started walking now.
TG Daily: Before Moorestown, we will see Silverthorne MIDs. Silverthorne, however is expected to make it into many other device, perhaps into the $100 laptop?
Mooly Eden: You are more aggressive than anybody else that I heard about, even Negroponte. You are asking me about OLPC or $100 notebook?
TG Daily: Let's take both.
Mooly Eden: I will not refer to $100 or $199. At the end of the day, the decision is done by the OEM who sells it. So you know that Asus got the eePC. Negroponte has got the OLPC. And you know that we are part of the board [of OLPC], right?
TG Daily: Yes, we have heard about that.
Mooly Eden: What is important is that the Silverthorne architecture and characteristics are such that should enable it to go and compete in a very low entry-level product because of two reasons. First, the cost structure is low enough so we’ll be able to price it in order to compete. You can’t come up with a $200 CPU for a $199 device. On the other side, the power consumption needs to be low enough so that it can fit into places without a fan, to enable an overall low-cost system.
On the MID, if you want to ramp this category, it’s a consumer play. Consumer play is not $699. Consumer play is $199, $299. It’s all kind of things, which are cost-affordable. We need to be able to be competitive there, which means that the cost structure should fit such a point. So, if you ask me if it’s $100 or $199, I cannot tell you. It’s not my decision, but are we priced, or are we designed in such a way to arrive at the right place? Yes. It’s a totally different “landing zone” than the mainstream microprocessor, the Core 2 Duo family.
Read on the next page: Relationships with OEMs, the pressure on AMD and Eden's take on the Fusion processor
TG Daily: AMD is known for having close relationships with its partners. How close is Intel's relationship with its OEMs these days?
Mooly Eden: Very close. In the end, it’s my decision. It’s Intel’s decision. But we speak with the OEMs, we speak with the ecosystem. If you speak [with OEMs] the year before introduction, it’s so relevant to execution that you cannot change the product anymore. But if you speak about the earliest change, we’ve got something that is called strategic move. We’ve got periodic strategic engagements where we show them the direction we want to take.
We cannot always do what they want, because many times we can go to two OEM and we get contradicting input. But we do solicit their feedback, because they’re close to the market and we want to win their business. But, at the end of the day, it’s our decision. And you need to bat. You need to make your decision and execute.
TG Daily: With regards to AMD, their product base, and where they’re going to be in a year. Do you have any breathing room? Do you look at things differently now?
Mooly Eden: Do I have breathing room? You should ask whether they’ve got any breathing room. Based on the execution lately, I believe they probably have got their issue right now to deal with. If I look at the mobile space, I believe we have the luxury of being a company that can afford to have a dedicated team to deal with the mobile challenge probably four years before anybody else in the market.
If we look at Banias, or Pentium-M, or if we look at Core 2 Duo, it’s not that our engineers are better than other engineers in the world. More important is that we identified the challenge of the power war. We understood that taking the desktop CPU and stretching it isn’t optimal. And we had the luxury of having two very good design teams working in parallel. Which, by the way, we continue to do with the tick-tock model.
We did the Core 2 Duo. At the same time, our partners from Oregon were working on the Nehalem. Now the guys on Nehalem are going to start ramping. I don’t know how many companies can do that.
The other thing that might not be fair, but I never claimed to be fair, I’ve got the luxury of the technology [45 nm], at least one year before my competition. As a result, the transistor budget, leakage, etc., definitely gives me an advantage. There’s no way anybody in the industry can come up with 420 million transistors. Nobody can come out with 6 MB cache. Nobody can do it, simply because it would not be economic. And I believe the tools that we get for our design [45 nm] puts us in a very good position and I’m proud of that.
TG Daily: What is your take on AMD’s opportunity with Fusion?
Mooly Eden: There’s nothing special about Fusion. The next Fusion, or the next CPU that has CPU + chipset, integrated graphics on the same chipset was mine. It was Timna, which we developed back in 1997 or so. It was a great product, but it eventually it didn’t make it to market. Overall, the idea of integration is the right idea. If you do integration, you can save power because you don’t need the I/O buffer. If I put it on the same chip, you can save space. It looks like AMD will be going towards a Multi-Chip Package and integration. It’s not magic. Can it be done? Yes. Does it provide a smaller footprint? Yes. Would it use less power? Yes. Is it the revolution? No. I believe that going into this direction is the right direction for everybody to go. Eventually it will be an evolution.
TG Daily: So, Intel will not jump on this idea right away?
Mooly Eden: I doubt that [Fusion] is going to change the position of the market and the overall ecosystem. But is it the right trend for technology trend in mobile space? Yes. Is it a risk? Yes.
Here's an example. If you integrate memory controller and graphics that was the problem in Timna you need to gamble which memory technology you go with. You gamble the wrong memory technology and you’ve got too much premium and you might pay a penalty for that. But overall, I believe that going in this direction is the right direction.
TG Daily: On that point, is this a real, formal type of validation that what AMD chose to do a few years ago in their designs is now what you’re accepting and adopting, is that saying that they were right?
Mooly Eden: With respect to Fusion?
TG Daily: Well, there is more than just Fusion, right? There's the on-die memory controller, multi-point …
Mooly Eden: No. Definitely not. Because people are looking at the vendor. What you need to do is to derive. It’s not a good repeater. I had the memory controller on Timna before they had their memory controller their design. So, the technology was there. By the way, the mistake in Timna was that we gambled on RDRAM.
And at the end of the day, we are arguing about many, many things. But, you can do the measurement, and you can see who’s got the better performer, the better product and lower power. I am willing to take Penryn on any benchmark that anybody wants. Whether it be F-SPEC, I-SPEC, PC Mark, SysMark - whatever you want. Open the kimono and do the testing. I believe our product is superior to their product. That is what matters for me.
TG Daily: Thank you for the interview!