Chasing New Markets With Intel's Atom
First there was Silverthorne, an Atom processor designed for mobile Internet devices. And then came Diamondville, a version of the chip designed for low-cost laptops and desktops. Now, Intel is developing versions of Atom for consumer electronics and other devices.
Belliappa Kuttanna, the principal architect of Intel's Atom architecture, oversees the development of future Atom processors. The best known of these future processors is the Lincroft system-on-chip (SOC) that will go into Moorestown, Intel's next-generation platform for mobile Internet devices, or MIDs. But the company is working on other versions as well.
The Pineview platform contains a version of Lincroft designed for low-cost laptops and desktops, known as netbooks and nettops. Sodaville is a similar system for consumer electronics and Menlow XL is designed for embedded applications.
Kuttanna discussed how the Atom processor is evolving during a recent interview with IDG News Service. What follows is an edited transcribed of that conversation.
IDGNS: During the Intel Developer Forum in Taipei, Anand Chandrasekher, the senior vice president and general manager of Intel's Ultra Mobility Group, showed a slide that listed four new computing platforms that will be built around Atom: Moorestown, Pineview, Menlow XL and Sodaville. What is Intel's strategy for Atom?
Belliappa Kuttanna: In addition to the categories that have received quite a bit of attention of late, which are the MIDs, netbooks and nettops, we are looking to use the Atom architecture and CPUs in other segments, like the digital home, digital entertainment, consumer electronics, and embedded devices. That particular slide tried to capture some of those other usages Intel has in mind for the Atom architecture and these are products that are being developed as we speak, based on the Atom architecture.
IDGNS: What aspects of Atom are you looking to improve with the upcoming chips?
Kuttanna: Power was the first, second and third consideration. In the MID space, we clearly wanted to target the communication MIDs. Clearly, we didn't have a technical solution with the current MID platform to target communication devices. Idle power reduction was our number one criteria. We also wanted to improve battery life for active usage, whether it's video playback or Internet browsing, as the case may be. That was also a key goal for us.
IDGNS: Apart from power, what are some of the other challenges you've tried to address with the upcoming Atom processors?
Kuttanna: Maintaining compatability of the architecture, whether it's with the legacy PC architecture or being able to run an off-the-shelf operating system, introduced some very interesting challenges for us. We had to go in and look at aspects of platform design that hadn't been looked at for a decade or more within Intel to make sure we didn't break any of those compatibility issues.
In the MID segment, in particular, there was a pretty steep learning ramp for us in terms of being able to manage power within the different subsystems of the SOC, given how aggressive the power constraints were. Architecturally, there was a lot of learning and a lot of innovation that we had to apply.
IDGNS: What are your thoughts on multicore for Atom? Is the performance boost worth the increased leakage and power consumption?
Kuttanna: In the MID space, we don't have any plans at this time to go multicore. We believe that, with the simultaneous multithreading support that we have, we get a pretty efficient solution, which gives us a lot of the performance that you would get with a conventional dual-core solution,but not consuming as much power as a dual core would take.
Like you said, leakage management with dual core is a challenge, unless you're willing to adopt techniques that mitigate that and add complexity to the designs. Our plans for the foreseeable future are to stay single core for MIDs. With netbooks, the considerations are a little different because the power constraints are not as stringent as in MIDs. As of now, we are not really saying what our plans are going to be as far as multicore goes.
IDGNS: What effect has the fast-growing demand for Atom had on the design work that you are doing?
Kuttanna: With success come a lot of new and interesting challenges, especially at a company like Intel where we have various options on where to take a particular base architecture. We are in the process of striking a balance, which is not pushing the limits of the architecture and not coming up with suboptimal solutions just because you have something that may make sense for a few product categories, but not other segments.
The best part for me is the amount of technology sharing that happens within business groups at Intel. For a big company, we pretty much act as a startup when it comes to sharing ideas across different architecture groups and product teams. That certainly helped us when we were developing Silverthorne and has certainly helped us in other areas, like SOC designs.
There were other teams in Intel that had a lot of experience with SOCs, and we were able to leverage some of that experience on the fly for some of our products. We continue to do that. It's a balance and very demanding on the team, but at the same time we have set up the infrastructure for sharing across teams and that has certainly helped us.
IDGNS: As an engineer, what has been the most interesting aspect of working on Atom?
Kuttanna: The most interesting aspect to me has been Intel most valuable asset, its process technology. It's amazing what kind of things that our process technologists come up with in terms of making things applicable to each product segment that Intel targets. Combining that with architecture and implementation is something that fascinates me. We have a huge challenge in that we are trying to enter markets segments where we are clearly not the incumbent. Being able to make this a big business for Intel, especially in the smartphone/MID segment is something that we will pay particular attention to and try to do the best we can.
IDGNS: That's still the primary goal for you, to break into the mobile handheld space, even as you target these other producr segments?
Kuttanna: Yes. With Internet usage growing by leaps and bounds in that space, it would be remiss of us not to address that space.