Sign in with
Sign up | Sign in
Meet Moorestown: Intel's Atom Platform For The Next 10 Billion Devices
By ,
1. Intel’s Ultramobile Future Arrives

Imagine you’re running a 3DMark graphics demo at perfectly fluid frame rates. Then imagine you’re watching 720p, 8,000 Kb/s video at a steady 30 FPS. And just for giggles, pile on a camera with a little videoconferencing app showing you streaming at the same 30 FPS. Now put all three apps on the same screen. Not earth shattering for one system to pull off, by any stretch, but not bad, right?

Now, imagine all three of those apps running with that level of performance on the smartphone in your pocket.

Impossible, you say. There isn’t a phone in the world right now that can play video at those rates, never mind having the other two tasks running concurrently with no performance impairment. Well, my friends, I’ve seen it with my own two eyes.

Today, Intel goes public with its Atom Z600 processor series. Perhaps netbook performance has left you uninspired. Perhaps Intel’s prior-gen ultramobile platform (meaning smartphones and mobile Internet devices, or MIDs) left such an indifferent impression on you that you’re now asking, “What? Intel had a phone chip?” Rest assured that the Z600 is a different beast altogether.

The company invited Tom’s Hardware to its Austin, Texas ultramobility development center for a pre-launch peek at the platform that has until now been called “Moorestown.” This wasn’t another fluffy press tour. Intel left no doubts that it is serious about this market segment, and was prepared to explain in extensive detail why Moorestown was a game-changer.

So buckle up and give your current phone one last gaze of admiration. You might not be as enamored with it by the time we’re done.

2. Little, Less, And Loving It

To paraphrase vice president Biden, this is a big f—ing deal. After attending this Moorestown briefing, I walked away fairly convinced that I’d just seen the future of mainstream computing. No, I’m not saying that I think 40% of the market will be toting around Moorestown-based devices next year. I mean that, if certain requisite elements are in place, I see no reason why the median form factor used for computing shouldn’t continue its march from the desktop to the pocket.

Look at these recent numbers from IDC. Desktops are done as a growth vector. Intel and AMD can continue their architectural arms race until the cows come home, but desktop PCs are going to become less and less of a market interest as their static sales become an ever-smaller piece of the computing pie.

In contrast, mobile PC sales are going to more than double in the U.S. over the next five years, and the growth rate worldwide is even higher. Mind you, this only extends down to nettop systems. IDC’s numbers don’t account for handhelds or tablets, which IDC is now “keenly focused on,” according to a recent press release.

Look at the 20-year trend. We’ve gone from a market comprised almost entirely of desktops to one now dominated by laptops. Mobile PCs started outselling their desktop counterparts back in mid-2005. Netbooks arrived in earnest during 2008, and now the diminutive form factor seems to be cannibalizing notebooks. Last year, NPD reported that “once they got home, 60 percent of buyers said they never even took their netbooks out of the house.” In July 2009, DisplaySearch released netbook sales numbers showing that notebook sales actually decreased, while netbooks grew 136.9% year-over-year.

If the mainstream computing market is really more about decreasing size than increasing speed or functionality, then it should have come as no surprise a month ago when Bloomberg Businessweek ran an article that said despite netbooks accounting for 26% of all PCs sold during the prior holiday season, “netbooks' popularity may already have peaked.” IDC numbers show netbook growth plummeting from prior-year levels. The article indicates that while several factors may be behind a 2010 fall-off in netbook interest, the arrival of Apple’s iPad and its imminent horde of competitors may be to blame as the industry looks for “the next big thing.”

Add it all up. While no one disputes that desktops will remain important for several applications, particularly at the high-end, the mainstream will continue to place its dollars into smaller form factors. The only reason people haven’t viewed smartphones as computing devices so far is because they haven’t been powerful enough to take over mainstream computing needs. With Moorestown, I believe we’ve reached a crossover point where that’s no longer the case, and if that’s the case, we’re now in a place where your next phone may also be your next PC.

3. Checking Checkboxes

Now, I may be totally wrong about my phone-as-mainstream-PCs thesis. Once I started asking questions in this vein, and about how phones might soon cannibalize desktop and laptop sales, only at much lower price points, Intel reps were quite prompt about trying to divert me from the notion. “That’s not how people use these devices,” commented one rep. I wanted to reply, “Maybe. Then again, they’ve never been able to use these devices in that way.”

Ironically, it was Intel’s opening commentary at the briefing that set me down this thought path. One presenter noted how tomorrow’s smartphones would allow buyers to “get a powerful computer that also happens to make phone calls.” With Moorestown, the company added, “this really hits you in the head.”

As usual, Intel assiduously avoided mentioning rival phone platforms by name, but reps did observe that today’s top three to five smartphones are “increasingly becoming handheld computers.” The company even did an informal study of these leading phones and paid close attention to the first 20 words or so of the marketing on their Web pages. In every case, not one product talked about making phone calls. Instead, the marketing lingo targeted four primary performance vectors that consumers now demand: compute, graphics, video, and Web.

Purists can sit around and debate which phone has better voice quality and reception, but Intel’s point came through loud and clear. Voice functionality is now a checkbox item. All it has to do is be good enough. The public doesn’t demand anything better anymore, at least from a phone. Of course, demands on carriers are a different matter entirely. AT&T, can you hear me now? One of the press managers in Intel’s group noted that her daughter logs in an average of 30 minutes of voice time on her phone per month—and 3,800 texts. That’s not even the future of “phone” usage. That’s now.

The next checkbox item is battery life. The reality is that we all charge our phones every night. Occasionally, some unforeseen adventure or bout of brain impairment might result in needing to stretch for three or four days, but it’s rare to need a phone’s standby battery time to last for more than 48 hours. This is why you never saw a handset based on Intel’s former ultramobile processor, the Atom Z5xx, code-named Silverthorne and part of the Menlow CPU/chipset/wireless platform. Despite having a 2.4W TDP, active battery life under Menlow was atrocious. Given reports of Menlow-based devices like the JooJoo tablet only lasting 2.5 hours under moderate use (compared to the iPad’s 10+ hours), no wonder Silverthorne has kept a microscopically low profile on store shelves.

But what if battery runtime were no longer a problem? What if Intel could check the 48-hour box and move on? OK, a little spoiler here: Intel did. Moorestown smashes the power problems faced by former Atom designs, and in a minute I’ll show you how. Fixing the power problem entailed a lot of rethinking and innovation on Intel’s part, and this leads me to a final thought (for now) on the long-term significance of Moorestown. Look at this:

In general, Intel (and the rest of the processor world with it) has been in the habit of innovating from the top down. The Nehalem launch was another in a long line of examples. New architectures and technologies are made for the top of the market, then they gradually filter down into cheaper and smaller implementations.

With Moorestown, though, I see a development reminiscent of the Pentium M. This is innovation from the bottom up, because there’s a lot more to Moorestown than just a power breakthrough. The Z6xx is also Intel’s first fully integrated System on Chip (SoC). The Clarkdale-based Core i5 and Core i3 cram processing and graphics into the same socketed package, but the Z6xx actually builds them into the same die, a feat Intel has never released before. Did Intel have the capability to do an SoC before now? Of course. But not until now, in this product segment, did it make sense to bring an SoC to market. Intel had to fix its power issues before any other steps made sense, including pushing an SoC up the processor chain into tablets (still virtually a dead segment until the iPad’s release), netbooks, nettops, and perhaps even notebooks.

We don’t know yet how much influence the Atom family will have on the entire spectrum of computing devices, but it’s clear that innovation is no longer only a top-down affair. Breakthroughs are going to be coming fast and furious from the bottom, and their impact on the middle 80% of the market may become much more significant than changes happening at the top.

4. The Moorestown Breakdown

Let’s get down to business. Moorestown is a mobility platform comprised of three key components:

Lincroft: Now known as the Atom Processor Z6xx, this is a 45 nm SoC part integrating a processor core, memory and display controllers, graphics engine, and a video engine. Why go 45 nm when the Core family is already transitioning to 32 nm and the object of the game is to be smaller? Because Moorestown was already in development when the 32 nm process change came online. In short, Intel’s time to market was faster with 45 nm, and the company could still achieve its goals on the larger node.

Langwell: This is the Intel Platform Controller Hub MP20, manufactured on a 65 nm process. Now that most of the yesteryear’s headlining chipset features have migrated to the CPU, most of what’s left in the PCH is I/O-related, although there are still a few surprises lying in wait.

Briertown: This is a dedicated Mixed Signal IC (MSIC) designed to manage power across the entire motherboard. You’ll also see the part referred to as a Power Management IC, or PMIC. Briertown is critical to Moorestown’s power-saving capabilities, but it’s actually manufactured by third-party vendors, including Freescale, Maxim, and NEC.

The fourth component that should probably be lumped in with the Moorestown platform is wireless connectivity, but this part remains too variable to be easily defined. As with Briertown, Intel has worked with numerous providers on several components that dovetail with Moorestown’s needs. Out of the gate, expect 3G products such as the M340 data/voice chip from ST-Ericsson, along with the Marvell 8688 for 802.11a/b/g Wi-Fi and Infineon’s Hammerhead 2 for GPS. More exciting will be Intel’s Wireless Multiconnection 3200, code-named Evans Peak, which will offer the 4-in-1 bundle of WiMAX, 802.11a/g/n, Bluetooth, and GPS on a single adapter.

We’ll dig into each of these pieces in the following pages. As we move through the group, keep in mind that Intel had four objectives for Moorestown, best summarized by this foil:

Among all of the points Intel wants to make with this launch, one stands far above the others: Atom’s power problems are over in the ultramobile segment. With an audio playback runtime of roughly two days and standby time exceeding 10 days, Intel can now play with the other phone chip big boys, especially ARM. Beyond that, Intel devices will be smaller than before, excel in compatibility, and deliver performance (particularly on media) that blows every other option on today’s market out of the water.

If that sounds a bit hyperbolic at first glance, let’s examine the details.

5. Platform And Process

Keep in mind that there are multiple families and architectures within the Atom processor family. In this article, we’re specifically focused on Moorestown and the Z-series, which is aimed at handhelds and tablets. There’s also the N-series for netbooks, the CE-series for TVs, D-series chips for entry-level desktops (D), an embedded series , and a future family “for gadgets” about which Intel wouldn’t even divulge a code-name. The ways in which these series differentiate are largely based on power profiles and performance expectations. We’re not to the point with Atom where one architecture, such as Core 2 or Core i3/5/7, applies to the entire stack. Perhaps it never will.

With Menlow, we had a platform architecture much like the classic PC design—a standalone CPU on top, with an integrated chipset below, similar to the old school northbridge and southbridge being combined into a single Platform Controller Hub (PCH). The Poulsbo chipset crammed in everything but the kitchen sink, and did it all on a relatively giant 130 nm fab process.

The architectural difference in jumping to Moorestown is massive. All that gets retained of the former chipset is the I/O complex. Memory, video, and graphics all migrate to the CPU—and not just in the package but on the actual die. Langwell uses a 65 nm process. Lincroft appears to match Silverthorne’s 45 nm process, but Intel is always careful to note that Lincroft uses a “45 nm SoC” process. It’s not the same process as before, or even a “retweaking” of it. Details here get vague.

While Intel maintains that the rest of the industry is still using 65 nm, Lincroft preserves Silverthorne’s 45 nm process. Recall that Intel’s 45 nm node was notable for its adoption of hafnium high-k dielectric technology, which got a lot of attention when it debuted in the Nehalem microarchitecture. Hafnium high-k, according to Intel, could reduce transistor-level gate leakage by over 100 times compared to the prior silicon dioxide dielectric process used with 65 nm technology. There’s more to Intel’s “LP SoC” process than hafnium, though, but engineers grew cryptic on this point. They stated that with Lincroft there was the “option of multiple transistors” as opposed to Silverthorne having “only one transistor end to end.” Then there were some furtive glances between the engineers and the press crew, and Intel would say no more on the matter. I suppose everyone is entitled to their hard-earned secret sauce.

The dimensional upshot of the Moorestown architecture is that we now have a 30% die reduction, a 40% package reduction, and a 50% motherboard reduction, reflecting significant consolidation across the platform. You’ll often see the Moorestown CPU package specified at 13.8 mm x 13.8 mm (Silverthorne was 25 mm2), but the actual die measures only 7.34 mm x 8.89 mm. One former Menlow reference design for handsets measured 75 x 148 mm. An equivalent Moorestown reference board I saw measured 69 x 130 mm, and that was with over one-third of the board surface sitting empty for an on-board battery.

6. Processor Power

Intel likes to refer to Lincroft as having an “ultra-low-power,” or ULP, processing core. I’m not going to veer off into the basics of Atom microarchitecture or the differences in how it processes data compared to higher-end Intel chips. We’ve covered that business before. For now, I’ll simply note that Lincroft is a single-core part that uses Hyper-Threading to create two logical processors for the operating system. The chip supports 64-bit code and uses the same Intel Virtualization Technology (VT) as the Core 2 Duo. Lincroft features a 24K data cache and a 32K instruction cache at the L1 level. There’s 512K of L2 cache and no L3. As we discuss power and performance from here on, know that we’re talking about a 1.9 GHz Lincroft part. To the best of my knowledge, this will be the top of the Z6xx stack in the near-term.

Now, it’s important to add that the 1.9 GHz clock is a “burst mode” rate, and we need to explain this. There are four primary power states in Lincroft: Ultra-Low-Frequency Mode (ULFM), Low-Frequency Mode (LFM), High-Frequency Mode (HFM), and Burst Mode. The Lincroft model that specs 1.9 GHz is likely to spend a lot of of its operational time in a 200 MHz ULFM.

Intel Burst Performance Technology (BPT) is a bit like the Turbo Boost we’ve seen implemented on the desktop in that it provides on-demand performance when needed, and when power and performance profiles will accommodate it. In the graph shown below, you can see that the HFM is the “sustained thermal limit,” meaning the actual TDP. At no time can the platform exceed its CPU thermal junction (Tj) or external chassis (Tskin) temperature limits as measured by thermal monitors. If these are exceeded, the platform will throttle back to the LFM or ULFM “recovery points” to cool off and then remain within the HFM threshold until enough headroom reappears for another burst. Naturally, these transitions all happen within fractions of a second.

With Turbo Boost, there’s a defined, guaranteed frequency and a set specific limit. When X number of cores go idle, you know the remaining cores will jump to Y frequency, and the BIOS doesn’t know what that frequency is. With Burst Mode, though, frequencies are governed by the BIOS. In fact, as Intel puts it, “Burst Mode frequencies [can be] enumerated as P-states” by the BIOS, and multiple Burst Mode exit policies can be defined.

Another facet of Burst Mode is that it supports “race-to-halt” power profiles as driven by Operating System-directed Power Management (OSPM). Race-to-halt reflects the same concept found in server computing environments: the object of the game is to blast through work as quickly as possible in order to revert to a low power, idle state. While the burst utilizes a high-power mode, total work in CPU-bound loads gets finished in less time than if the load were to run at a “normal” speed at a standard power level, and thus net power gets saved. OSPM functionality is directed through drivers and now support power down modes while the device is still active.

OSPM works in conjunction with hardware-based power controls, acting in a sort of advisory role. Software sets power policies and constraints, but hardware ultimately does the fine-grained power management. As you might expect, power and performance needs will vary depending on the application being used, and part of OSPM involves leveraging middleware profiles based on common hardware and software activities. In the common event of multitasking, where different usage profiles might apply concurrently, hardware ultimately gets the last word.

7. New Power States

Another improvement over Silverthorne is Intel’s addition of “Enhanced Geyserville,” or Intel Enhanced SpeedStep Technology. This is part of how Lincroft chips are able to run at 200 MHz, while Silverthorne bottomed out at 600 MHz.

Just as importantly, Lincroft now uses power gating across 19 “islands” within the processing core. We’ve seen similar power gating employed in recent Core i-series designs, where current is cut to an entire block within the CPU in order to prevent the leakage that increasingly plagues circuits the smaller they get. This is a much more granular and effective approach than the older method of dropping power (and thus performance) en masse across the core, which still left leakage doors wide open. This was the model that Menlow used, wherein power was either in an on, off, or sleep state. There was nothing in between. With Moorestown, we can now control which islands are active or powered down.

The conventional CPU power states of C0 through C6 still carry forward into Lincroft and are now often collectively called S0. But one of the critical improvements in this generation is the addition of two new power states called S0i1 and S0i3.

S0i1 gets utilized during idle periods in which user is still classified as being interactive but may not be doing anything at just that moment, such as when looking at the home screen or reading a Web page. As you can see in the image above, the majority of islands in the CPU are powered off in S0i1. This delivers a 10% to 15% drop in power use while active. There’s about a 600 microsecond latency when entering the state and an exit target of 1.2 milliseconds.

S0i3 kicks in when the user isn’t actively using the device. Essentially, this is Lincroft’s standby mode, the state in which Intel estimates that a handheld or tablet will spend over 99% of its time over prolonged use. Only the SRAM, GPIO, and System Power Management islands receive any power. Entry latency is 450 microseconds while exit latency hits about 3.1 milliseconds. To get a sense of just how much power S0i1 cuts across the Linfield die, check out the the following image. Then imagine nothing but empty, unmarked blue save for two fields in the top-left corner. That's S0i3.

Intel draws the analogy of walking into a house at night. Under the old model, there was one light switch. You walked in the door, flicked the switch, and the whole house lit up, albeit with a rudimentary dimmer attached. Under Lincroft, you get to turn on lights only in the rooms you need as you walk into them. When you exit the room, out go that space’s lights automatically.

Taken all together, the power improvements in Moorestown yield a 2x to 3x reduction in active platform power compared to Menlow, and a 50x reduction in idle platform power. Without this, Intel couldn’t have met its battery life objectives and become a real contender in the ultramobile market.

8. Graphics And Video

A bit of poking around and combining sources led me to make a few comparisons that Intel itself didn’t want to discuss during its Moorestown briefing.

The Palm Pre and Apple iPhone 3GS both use a Texas Instruments OMAP3430 processor based on ARM’s Cortex A8 core and a PowerVR SGX GPU. Both are considered very strong on graphics, and are well-known for their lush 3D interfaces, right? The Snapdragon processor found in phones such as Google’s Nexus One, the HTC Incredible, and HTC’s EVO 4G has half the peak fill rate of the TI processor. Believe it or not, Menlow outperformed the OMAP3430 by 60 percent—you just never heard about it because battery life (or lack thereof) overshadowed everything else. But the kicker is that Moorestown, with its 400 MHz GPU, doubles Menlow’s fill rate, and Medfield, the successor to Moorestown, will double this rate yet again. To accommodate this bandwidth, Intel had to revamp the CPU-to-graphics bus, raising it to 6.4 GB/s on reads and 4.3 GB/s on writes, figures that Intel quipped were “ridiculously high.”

Intel builds plenty of standards support into its “Thalia” graphics core, including OpenVG 1.1, OpenGL 2.1, OpenGL ES 2.0, and DirectX 9.L. The emphasis on OpenGL no doubt explains why Intel loves to showcase Moorestown running a Doom 3 timedemo (at around 100 FPS), but this also pays off in accelerating vector graphics and supporting apps like Adobe Flash Player. This also plays a key part in making the Moblin/MeeGo 3D Clutter UI so compelling.

Just as the graphics support is very similar between Menlow and Moorestown (including their mutual support for SSE3 instructions), the same is true for video. Lincroft can handle simultaneous 1080p30 HD and SD decoding. If you’re willing to accept 480p, MPEG-2 quality, Lincroft can tackle up to half a dozen video streams at once. Such hefty capabilities are possible because of Intel’s integrated acceleration features. On the video side, Lincroft bakes in hardware acceleration for MPEG-2 and H.264/MPEG-4 encoding and decoding. Also tack on hardware decoding of WMV and VC-1 as well as software decoding for MPEG-1, Xvid, Real Video, and Adobe Flash Video. Photo buffs will appreciate that Lincroft adds hardware acceleration for JPEG encoding.

Video addicts will wonder about bit rates, so know that Moorestown can handle 20 Mb/s on every profile, from 720p baseline at 30 FPS to 1080p high profile at 30 FPS. No other phone platform available today can decode 1080p. Only a few can even touch 720p MP at 10 Mb/s. Intel mentioned that the Aava Mobile-produced Moorestown reference design has enough decode bandwidth to blast through 40 Mb/s, although you’d never likely encounter such content. More likely, you might want to decode two 20 Mb/s streams concurrently.

Intel may now own the video crown in this segment—whoever thought we’d use that phrase in print?—but competition is coming up fast. In early 2009, TI announced its OMAP 4 series based on ARM’s Cortex-A9. The A9 will reportedly deliver a 7x performance gain, enabling 1080p video recording, capture of 20-megapixel images (good luck affording the sensors for that), and battery runtime able to play back MP3s for one week straight. If it seems strange for Intel to be hinting strongly at Medfield before Moorestown even arrives, the A9 would be why.

9. Display And Memory

In the notebook world, most displays are based on low-voltage differential signaling (LVDS). Lincroft supports LVDS at resolutions up to 1366x768—plenty of pixels for the sub-notes in which you might expect Atom to appear. In the handheld and MID spaces, though, displays are gravitating to the Display Serial Interface (DSI) put forward by the Mobile Industry Processor Interface (MIPI) Alliance. The purpose of MIPI is to establish industry standards for wired and wireless interconnects in the ultraportable world, which often have different priorities than those in the desktop or server worlds. DSI uses a form of LVDS serial bus but seeks to specify lower-cost approaches to LCDs specifically in the ultramobile market.

Today, Intel specifies that Lincroft can handle MIPI-DSI output at up to 1024x600. So, even if someone were to create a Moorestown-based device with a big enough LCD to accommodate native 1080-resolution video, Lincroft tops out at 720p to the screen via LVDS, and even less via MIPI-DSI. Why, then, should we get excited about Moorestown’s ability to handle 1080p video when no other phones can? In part, because the horsepower able to decode 1080p can also decode multiple 720p or lower streams. But also keep in mind that if your video collection is standardized on 1080p content, you don’t want to have to waste time transcoding everything you transfer over to your ultramobile device. Just copy and go—the device will take care of the rest.

Lincroft’s embedded controller supports two memory formats: LPDDR1 at speeds up to 400 MT/s and DDR2 up to 800 MT/s. Why two formats, given that memory is hard-mounted on the platform board? The answer has to do with market segments. Moorestown currently targets two device application groups, one based on communication and the other on entertainment and productivity. The communication models are the ones with 400 MT/s LPDDR, and they don’t presently span up to 1.9 GHz with BPT. That falls to the entertainment/productivity group with its 800 MT/s DDR2. Interestingly, only the latter group will support 1080p decoding. As of this writing, we still don’t have confirmation that the communication platforms will decode 720p, although it appears likely.

Typical power consumption for Lincroft and Langwell combined in standby is about 3mW. This applies to both the communication and entertainment/productivity groups. However, under active use, the higher clocks of the latter group start to take a toll. Whereas communication platforms land in the 300 to 500mW range, entertainment/productivity platforms run between 450mW and 650mW, at least on pre-release hardware.

10. Langwell Platform Controller Hub MP20

I said earlier that the MP20 chipset was pretty much an empty, downsized shell of its former Poulsbo self, reduced to being a collection of I/O blocks. That’s not exactly true. Yes, Intel builds a USB 2.0 EHCI controller with three ports into the 14 x 14 mm package. There’s also a USB On-the-Go (OTG) host/device port for letting the Moorestown ultramobile connect to other USB devices. Intel’s inclusion of HDMI support is all but overlooked in the Moorestown documentation, but its presence is obviously a key point for those who would like to use their devices as a video player for their TVs.

How about storage? There are four ports for SD/MMC/SDIO devices. More importantly, Langwell supports SLC and MLC NAND “drives” with up to 4KB page sizes and a raw capacity of 64GB. Note that while Langwell supports the CE-ATA spec common for 1-inch Microdrives, there’s no SATA connectivity. Most likely, the feature was considered unnecessary bloat given the part’s target applications. However, Intel does have implementation guidelines for developers that want to employ a third-party USB-to-SATA bridge. The same guidelines also detail how to connect a full-sized keyboard via an I2C interface.

Camera support and image processing also run through the MP20, both for still images and video. Moorestown allows for two cameras—one 5-megapixel and the other VGA (640x480) resolution. The 5MP channel is a dual-lane MIPI interface able to support RGB, YUV, and RAW color schemes, although how this capability gets exposed will be left to developers.

Langwell’s audio engine is more complex than you might expect. Intel uses its in-house Smart Sound Technology (SST), based on a 24-bit DSP, for voice processing and audio codec accelerations. Specifically, Langwell provides hardware-based encode acceleration for AAC-LC and PCM (WAV). On the decoding side, there’s hardware acceleration for AAC-LC, HE-AAC, MP3, PCM, and WMA9. This optimization, along with other elements of Langwell power management, is part of how Intel is able to hit its week-long MP3 playback time target. When the audio engine needs system resources, it generates a power management event, but only the path to memory is enabled. The audio engine refreshes the buffer and the platform quickly returns to its low power state.

While arguably less of a concern for consumers than businesspeople, some will be interested in Langwell’s integration of a security and cryptography engine, which includes the functionality of a Trusted Platform Module (TPM) chip. Intel accelerates AES, DES, 3DES, RSA, and other crypto operations, but the real benefit lies in enabling a secure boot environment, as with vPro systems. This is likely to become increasingly important as mobile devices become a larger target for malware and hijacking. Of course, content providers are keen on the ways in which DRM can be enforced with a security engine, as well.

11. Briertown Mixed Signal IC

Briertown may seem like the extra wheel of the Moorestown platform, but its role is critical for Intel hitting the necessary power targets. There is no one official Briertown design. Rather, Briertown is an architecture specification for peripheral support and power delivery, including battery charging now managed by hardware instead of software. Look at Briertown on a Moorestown motherboard and you’ll see that it’s considerably larger than the CPU and chipset combined. Despite this, Briertown requires almost half as many components and consumes about one-third the board area as the equivalent power delivery circuitry on Menlow platforms. Not surprising, Briertown also costs about one-third the price of its predecessor.

The MSIC is part of what makes Moorestown’s power gating possible. The Briertown complex runs multiple voltage rails to the CPU and chipset, all of which can be controlled by the operating system. Without this, there couldn’t be Lincroft’s Burst Mode because Briertown is what enables those lightning-quick transitions in and out of the various C- and S0ix-states.

Briertown manages device and subsystem power delivery at the system level, each of which get integrated as “jellybeans” within the larger complex in order to keep management more granular. For example, the TPM block occuplies about 6 x 6 mm and consumes up to 85mW. The USB OTG block from Philips measures 20 x 14 mm, consuming about 300mW, and the 2MP USB camera chip from ST Micro measures 7 x 8 mm and draws roughly 400mW.

One can wonder if the Moorestown devices we see at mid-year will be as power-efficient as the knock-off brands and designs we’re likely to see several months later. Intel was emphatic about the fact that when it comes to its new power states, hardware management technologies, and Briertown design, there is no room for flexibility. The platform must hit certain minimum performance levels, and Intel isn’t willing to compromise on this. We may see more “cost-effective” products creep into ODM designs over time, but it won’t be from modification of Intel’s core platform. How much impact future OEM corner-cutting will have on final power and performance levels remains to be seen.

12. The Experience

Hardware without software makes for a very nifty doorstop. As mentioned earlier, Intel has developed a tight allegiance to the Linux-based MeeGo OS, formerly known as Moblin before Intel and Nokia partnered to develop and promote it more heavily. MeeGo gives Intel the ability to have one environment span across the embedded, handset, MID, and netbook/nettop segments with a unified software stack. Of course, these days you can’t touch a phone platform without first asking about app support and a foofy app store front end.

Well, Intel does have an app store for MeeGo: the awkwardly named AppUpCenter. When AppUp was announced last January at CES, Intel promised that netbook OEMs would be supporting it. However, Intel’s site is still in beta, and it remains to be seen how much third-party support materializes. According to Austrialia’s TechWorld, AppUp passed the 200-title mark last month, and that was with Intel motivating developers with some monetary incentives.

Am I going to declare MeeGo a sure success in the making? Only about as fast as I’ll expect an effective overhaul of American healthcare. In both cases, you’ve got a big player fighting uphill against a much larger, deeply entrenched, and competing infrastructure. I wish Intel all the best on this front, but my money is on Android being a bigger hit. I’m not sure that Nokia’s allegiance to MeeGo, which is now effectively Maemo 6, will be enough to propel into the top OS ranks.

As of today’s launch on May 4, 2010, it’s anybody’s guess what will happen with MeeGo. On one hand, just a week ago, we saw the LG GW990 phone officially vanish into pre-release hell. The GW990 had been the Moorestown poster child ever since CES in January. I should have suspected something amiss in the air when Intel only focused on the Aava Mobile and OpenPeak designs at the Moorestown briefing, almost totally ignoring the GW990. Could this be fallout from Intel’s bosom-buddying up with Nokia and the fact that Nokia released its first MeeGo port for the N900 to developers a month earlier?

Of course, some users won’t care. So long as devices sync properly with user data, whether local or cloud-based, and run the applications the user needs in a way that’s compelling and effective, that will be enough for plenty of buyers. I got to fiddle about with the Aava Mobile smartphone and OpenPeak tablets for a few minutes at Intel’s briefing, and my first impressions of MeeGo were very positive. Is it better than iPhone OS or Android? Am I going to marry it based on the hands-on equivalent of a MySpace posting? It’s way too early for answers. Let’s do some premarital cohabitation with a review unit or two and see where things lead.

I will say this: MeeGo delivers a solid, impressive graphical UI. I’ve seen it multitask first-hand, felt the responsiveness of its touch interface, and admired its wicked 3D navigation models. Yes, I also saw it hiccup once or twice. Despite its Linux roots, MeeGo still has plenty of refinement in its future. To give you an idea, I saw the Aava Mobile reference handset running MeeGo and one photo management app in which one thousand 5-megapixel images were arrayed in a grid on the screen, represented as tiles so small they only looked like colored dots the size of a few pixels. With a few finger drags, the user could zoom into this grid however much he or she wanted, even up to 100% full size. It was like Google Maps meets My Photos.

That was cool, but the part that really blew my mind was when the presenter zoomed all the way out and said, “Now, what if I want to resort these images by their dominant color?” With a couple of taps and the time it took for a quick screen wipe, all of those tiny tiles reorganized themselves into a continuous progression organized by tone gradients. Imagine having this kind of filtering capability according to other content variables, such as the presence of blue sky or human faces—on a phone!

When I asked Intel reps why they were going to market first with a relative no-name like Moblin/MeeGo when Android was seemingly so close to being ready, one team member answered with a wry grin that in a process as intricate and critical as a new platform launch, it made sense to focus first on a software infrastructure that offered Intel the most direct control. No doubt, control and timeliness were also key factors in Microsoft’s omission.

The Archos 9 tablet supports Windows 7 on Silverthorne, so we have a strong suspicion that Moorestown could shake hands with Redmond if developers were so inclined. However, there remains so little momentum behind Microsoft in the consumer ultramobility spaces and so little track record of Windows-on-Menlow success that it’s no surprise Microsoft remains wholly absent from Intel’s platform launch. On the other hand, we did get to see Android running on a handset, albeit one that was dissected and mounted on a lab test bed much like the one below.

13. Why Moorestown Matters

How big of a deal is Moorestown to Intel? Two of the first factoids Intel dropped at its briefing were these: 1) Globally, there will be one billion more new “connected” users by 2015 than there are today. 2) By 2015, there will be 10 billion connected devices in use. How much of this will be PCs versus non-PCs? In June of 2008, Gartner declared that “the number of installed PCs worldwide has surpassed 1 billion units” and that “it will surpass 2 billion units by early 2014.” Even assuming that Intel’s projection is overly optimistic and that a large swath of these future “connected devices” will be things like gaming consoles, connected cars, or whatever, we’re still talking about multiple billions of connected handheld devices in use. Is this feasible? Considering that the International Telecommunications Unions’ 2009 annual report pegged global mobile phone usage at 4.6 billion units, yeah—I’d say billions of “PC-like” handhelds is totally feasible. Intel might just sell more ultramobile processors in the next five years than it has sold into the PC market over its entire history.

Looking back across the last couple of weeks, taking in the specifics of Moorestown, its evolution, and what information Intel has fed (and not fed) to the press, I believe that this launch is roughly equivalent to the arrival of Conroe and the Core 2 family. Conroe put a final stake through the heart of NetBurst and confirmed Intel’s commitment to abandoning frequency as the central measure and means of processor performance. It led Intel down a different path for desktops and notebooks and allowed Intel to keep up a pace of innovation that competitors have been unable to match.

I believe Moorestown, and especially the Lincroft SoC architecture, will do the same for Intel’s ultramobility pursuits. Silverthorne was simply a warm-up, a prelude. Are there still blemishes waiting to be discovered in shipping Moorestown devices? Almost certainly. It seems very unlikely that even Intel could go from a chip as maligned as the original Atom to a miraculous revolution in just one generation. I suspect if the platform were that good, I’d have a unit in my hand right this minute. It would rival the iPad, cure the sick, game the stock market, and draw the fairer sex like moths to a flame. Heck, I’d settle for any one of those things.

Who needs a sportscar when you have the hottest new superphone platform? Pankaj Kedia, Intel's top marketing mastermind for Moorestown, struts his wares for the womenfolk. They seem impressed.Who needs a sportscar when you have the hottest new superphone platform? Pankaj Kedia, Intel's top marketing mastermind for Moorestown, struts his wares for the womenfolk. They seem impressed.

No, the message of Moorestown is not that Intel is suddenly the best mobility platform on the planet. Even company reps admitted that the Moorestown scorecard is mixed. Compared to its rivals, Moorestown is allegedly on par for browsing and standby power, trailing on audio playback, and excelling on video—and excelling so much that direct comparisons are often impossible. Moorestown debunks the common belief that IA is too power-hungry to succeed in ultramobility. As of now, IA is ready to fuel the rocket in your pocket.

With power needs met and performance at least on par with the competition, Intel finds itself with a familiar challenge. If the company can scale Atom on ultramobiles in the same way it scaled Core 2, then the future of ultramobility and perhaps mainstream computing seems all but sure to remain in Intel’s corner. One engineer commented, “Intel sees the Internet as a primary means to the end...and the end itself.” If that’s true, if the race in mainstream hardware is really a race to enable the best browsing experiences possible, regardless of size, shape, or location, then Moorestown seems likely to bring that end within Intel’s reach.