Intel Atom, Facts and Analysis Part 2 of 2.

In Part 1 of this report, I wrote about Intel’s Atom and MID master-plan. In Part 2 here, I put forward my thoughts about some of the main features of Intel’s Centrino Atom platform – The MID platform. It is important to remember that Atom-based devices are consumer-focused and not traditional productivity computing devices. They will be power-efficient, small and cheap but processing power is not as the top of the list of features. Pricing and and sizing take priority with Atom along with what I believe is a very important and mostly overlooked aspect of the Atom ecosystem – the software.

Performance.

Performance is a tough one to measure at the moment a there are very few reliable benchmarks available. A lot of the hardware I saw at CeBIT was prototype hardware with beta builds of drivers and operating system and even the tests I performed can’t be taken as final. From the tests I’ve done, the reports I’ve read and all the discussions I had at CeBIT it’s clear that we’re not looking at major performance increases over Stealey. Intel said in 2007 that we’d see about 10% performance increase over Stealey/McCaslin and having tested some of the high-clockrate (1.6 Ghz) prototypes, its feels about right. At the low-end (800Mhz) we’re going to see the Moblin based devices struggle with Windows desktop operating systems which is why Moblin is being developed. Even at these low clockrates though, they will return a far better Internet experience than current ARM-based devices and due to the video decoding hardware, some good video playback performance. The hyperthreading capabilities of the CPU could also help in some areas too.

Graphics.

It almost, but not 100%, confirmed that the graphics core in Poulsbo is a PowerVR part. Possibly the 535. That might not mean anything to most people so let me put it this way. Don’t expect increases in 3D performance but do expect some better video decoding capabilities. H.264 decoding in hardware is just what the doctor ordered for the future of YouTube videos. I wonder if Intel were thinking about Apple when they put that in the requirements list! For some more detailed discussion about graphics performance, see this forum thread.

Battery life.

First let me talk about battery efficiency before I talk about battery life. From what I’ve seen high-end devices (7″ screen, 1.6Ghz) will have a maximum system drain of around 8W with an average down to about 6W. For the low-end, smaller devices (4.8″ screen, 800Mhz) the maximum drain will be a few watts less with averages down to about 4W. Compared to 2006 ultra mobile PC devices that had max power drain figures of up to 20W its a huge advance. On average its about 30% reduction in power consumption across all usage scenarios when compared to 2007 devices. Intel said in 2006 that they would achieve a sub-4W average drain and having asked Intel directly if they were on track with that they said that I ‘would would be surprised.’ Putting these efficiency figures into perspective with the smaller batteries required on the smaller devices, we should expect up to about 3 hours in-use battery life with quiescent, no-load state returning 4-5 hours on MIDs, maybe a bit more if the system and software is well optimised. Considering that the latest smartphones can drain up to 3W in Internet usage scenarios, the power gap has been reduced to a very small and threatening margin when compared to ARM-based devices.

Size.

Centrino Atom devices will all be pocketable. From the Intel Mobility Meeting: For a device to be a Centrino Atom device, it will need to have a screen of 6″ or smaller with a total diagonal dimension of 7.5″ The thickness needs to be 1.2″ or smaller. Which means that you’ll see devices similar to the HTC Advantage and no bigger than a Raon Digital Everun. The smallest devices might reach something the size of a Nokia E90 communicator but I wouldn’t expect anything smaller until Moorestown is available in late 2009 or 2010.

Price.

Intel have been talking about $500-$700 MIDs. OEM’s are talking about $600-$900 target end-user prices. I expect to see early launches at at up to $800 and budget brands coming in at $650 based on what I heard at CeBIT. Considering this price will include 3G radios and in most cases, built-in GPS, its not unreasonable for an early adopters market. Mainstream customers waiting for $400 device will probably have to wait until 2009 before the first sub $500 MID appears but it depends on the competition. If someone chooses to promote their device in a ‘Eee’ style marketing campaign then pricing levels could change very quickly.

Where’s the UMPC?

Intel see the ultra mobile PC as a small market in comparison with consumer Internet devices. Atom is all about focusing on big sales numbers at consumer prices. The smartphone market is obviously the ultimate target with later generations of Atom which means that we won’t see much in the way of marketing for pro-mobile style UMPCs. Intel will support manufacturers building mobile devices on either Atom or the laptop processors which means that the pro-mobile market can continue to develop but its very likely now that we’ll see more Intel-based pro-mobile ultra mobile PC devices being based on notebook silicon rather than Atom silicon. This could leave a nice little niche for VIA to step into. A well-priced Ghz-class VIA Isaiah core processor with decent graphics support could be just what the pro-mobile market is looking for. If the VIA-based HP ultra mobile PC performs, it could really improve confidence in them for the pro-mobile ultra mobile PC market.

Device availability.

Atom launches in the 1st half of 2008 which appears to be a slight slip from the April IDF launch from what I can tell. Indeed, most manufacturers are talking about June/July releases and ‘summer’ is mentioned a lot. Factor-in the inevitable delays and its quite likely that customers won’t be making final purchasing decisions until August of 2008.

Measuring success.

The MID/Atom branding and marketing already seems to be working. At CeBIT there were a number of manufacturers showing devices with specifications based on Atom and behind the scenes there’s a lot of Atom talk going on. If Intel do manage to push out 30 devices this year they will have pushed far more than what we’ve seen coming through the Origami initiative. The real measure of the momentum of Atom during 2008 and 2009 will be based on the number of ISV’s we see jumping onboard. Unfortunately there are a lot of choices in the mobile Internet market for developers now. Apple, Google, Nokia and all the mobile Linux consortiums are all trying to attract developers because everyone knows that it’s all about slick applications and user interfaces.

Summary.

Its clear that Intel have a lot of confidence in Atom and consumer mobile Internet devices. If it fails, it could take the Centrino brand down with it and that would hurt Intel badly. In its first generation, Atom will bring mobile Internet to the pocket in a way that not even Apple or Nokia have been able to do with ARM-based devices. It will enable highly capable portable media players and navigation devices and because of it’s price I would expect to see it being used in Internet enabled TVs within a year. In it’s second generation it will enable the first x86-based smartphone and that’s where the real numbers start. I really wouldn’t be surprised to see Jobs and Otellini launch a Morestown-based Apple iPhone Atom at Macworld in Jan 2010.

Video.

The video below was taken during the meeting I had with Pankaj Kedia of Intel at CeBIT 2008. The sound quality is rather poor, sorry, but you should hear me ask Pankaj what types of operating systems we will see on MIDs in 2008. Pankaj responds: “In 2008 you will see Windows Vista and XP and Linux based devices. All Linux devices will be based on Moblin. Moblin will be distributed distributed by Ubuntu, Redflag, Haansoft and Miracle. All of these OS vendors are making sure that the OS is Moblin complaint. If that OS is Moblin compliant, an application from one OS works across OS vendors.” You’ll also see the Benq, Gigabyte, Toshiba and EB MIDs and 47 million transistors in a 25m2 package! JKKMobile has the YouTube version of this video.

Click for a comparison table of Intel MIDS.