IDF Fall 2007, part 7-- Ultra-Mobility keynote
Anand Chandrasekher hits the high points of Intel's strategy for mobile Internet devices.
I learned today that Intel has a Mobility Group and an Ultra Mobility Group. There's a sensible explanation for the difference: notebook PCs are defined as "mobile"; smaller systems are considered "ultra-mobile."
Intel further divides these ultra-mobile machines into two smaller classes: ultra-mobile PCs (UMPCs) and mobile Internet devices (MIDs). Traditionally, the former have 7" screens; the latter category goes all the way down to the tiny screens of smartphones.
Intel's Anand Chandrasekher, senior VP and general manager of the Ultra Mobility Group, took the stage for the second keynote of the day to talk about "Unleashing the Internet Experience."
His primary contention is that while many people have mobile Internet access through their phones, a survey showed that 61% of the people with this feature available to them have never used it. Intel is trying to deliver platforms that make the Internet not only available, but attractive.
Unfortunately, Intel is still stuck with a basic fact of life. The Internet is mostly written for PCs with large displays, whereas a mobile device inherently has to have a small screen. (Unless you believe in ultramobile projection screens, visor-type displays, etc.; I don't.)
Nevertheless, Intel is addressing those aspects of the experience that are within the range of its product lines. For example, Chandrasekher echoedin saying that performance matters on handheld devices.
He went so far as to imply that current handheld platforms based on ARM processors are inherently inadequate for Web access, although I think the experience of iPhone users shows that an ARM-based processor is fairly effective for mobile Internet access.
On the other hand, he made the valid point that non-PC Internet clients are inevitably going to run into compatibility issues. Even the iPhone doesn't support Flash or Java, and its list of supported video formats is very short by comparison with a full Mac, Windows, or Linux desktop system. Chandrasekher presented a slide showing error rates from tests of popular Web pages on various clients to back up this point; current Internet-enabled smartphones were far more likely to fail on the sampled web pages than a PC or Linux system.
He used the history of the Flash plugin to claim that handheld devices are generally about two years behind desktop PCs (or notebooks) in supporting new Web technologies. I think that gap will close by itself, but he said the best way to close it was to adopt Intel's x86 processors for MIDs in place of the ARM processors that have dominated that niche to date.
So, of course, Intel will be offering chips for these systems.
The first really usable x86 for handheld devices will be Intel's Silverthorne. In a demo, Chandrasekher showed a Silverthorne chip idling at less than 0.2W and consuming less than 0.45W even when running a "power virus"-- that is, software designed to maximize the power consumption of a processor. (These numbers are my estimates from the graphs shown in the demo.)
That's too much power for a smartphone, I think, but very reasonable for a mobile Internet device that isn't also a cellphone. Tack on another 1.5 watts for other components and you'll get a reasonable 5 hours a day of battery life from a 10 watt-hour battery pack.
There was an interesting but too-brief guest appearance by Al Ramadan, senior VP of the Mobile and Device Solutions Business Unit at Adobe Systems. Adobe is launching what I'd have to call a middleware layer for Internet multimedia-- AIR, for Adobe Integrated Runtime-- and an application called AMP (for Adobe Media Player) that runs on top of AIR to present all kinds of multimedia. Both products will be offered for Linux, Windows, and Mac OS on systems from MIDs up to desktops. Betas are available now, and both products will be released in 1Q08.
I thought AIR and AMP showed some promise for breaking this stalemate between website creators and handheld device manufacturers. If AIR-based applications, such as AMP, can be widely distributed in versions that are appropriate for all these different form factors, we may not need handheld devices with 19" LCDs...
Anyway, back to Chandrasekher's hardware content.
Following the 2008 release of Silverthorne, there'll be a 2009/2010 release of a more highly integrated product code-named Moorestown. Chandrasekher said that this product will bring idle power down by another factor of 10 from Silverthorne, putting it into the necessary range for cellphone use. Moorestown consists of two chips: one with a CPU core, graphics, memory controller, and video encoding/decoding engines; the other will have system and disk controllers as well as other peripherals.
To show off the potential of Moorestown, Chandrasekher showed what I can only describe as something that looks like two iPhones joined end-to-end with a single LCD spanning the full length of the device. Think of a 12" submarine sandwich reduced to about half an inch in thickness and you'll get the idea. For the life of me, I can't imagine anyone carrying this thing around, but it was interesting to look at.
In the press Q&A session following the keynotes, both Chandrasekher and Perlmutter took a few potshots at current UMPCs, possibly because most of them tend to be based on competing microprocessors. Chandrasekher dissed UMPCs as being "targeted more at productivity applications" and said MIDs are "much more broad-ranging than that." Perlmutter said "I'm not sure that I love the UMPC because it just takes a PC and makes it smaller." But in any event, I'm sure both men will be happy to sell components into the UMPC market if that's what OEMs and customers want.
And that's it for the second keynote. There are more sessions later, but I may not bother blogging about them unless they contain more actual news...