Intel's Atom(ic) fallout

YOU MIGHT NOT SEE the first Intel Atoms in any smartphones - the size and power usage, especially in idle mode, is still well over the allowable limits for such device. In the case of idle power, we're talking about another 15-20 times reduction before we see it in a really compact slide phone able to run for a week on standby.

Most of these problems - in particular integration and idle power - should be solved in a years' time with the Moorestown platform refresh. The CPU will then integrate memory, graphics, video acceleration and interconnect - kind of a mini-Nehalem, while the I/O, wireless and power management sides will be customisable, including non-Intel add-ons. The whole shebang will support everything from 3.5G, WiMax and WiFi to 3-D and HD video playback, TV out and more - all that at less than half the size and one-tenth idle power of the existing Atoms.

But hold on, some of the current ARM-based smartphones support most of these features right now, and still last nearly half a week on standby. Why should we even contemplate waiting for Moorestown and its more compact successors?

If you ask Intel, their answer is one word, better say acronym - X86. The argument that a smartphone must also be able to comfortably browse this PC-optimised Internet with all the X86-optimised flashes, players, plug-ins and all other things that make Internet browsing an irritation even on a proper PC these days. No matter how many custom ports for specific smartphone ARM-based CPUs are done for a galaxy of these Net plug-ins and such, they will never be on time or match the PC X86-ones.

The impact, according to Intel, is that Nokia's old dream - the Mobile Net, separate from the PC Net - is long gone now, and there is only one Net now: the PC X86-optimised one, of course. Therefore, to fully enjoy the Net experience natively on a smartphone, it should share at least the instruction set with the bigger PC cousins.

Assuming the Moorestown really allows us a 4-inch widescreen X86 iPhone and other similar-sized devices - yes, the rumour mill is spinning wildly on that upcoming deal, and we've all seen Intel's mock up at the last IDF - we could see the very high end of smartphones start seeing a noticeable X86 presence by mid 2009.

However, out of several hundred million 2009 smartphones, only a few percent will be these large and uber-expensive geek status symbols, used more for show off then for actual phone calls.

If Intel really wants X86 to become even with ARM, or even dominant, for the phones that are actually used for calling by the millions of normal user everyday, they will need to scale down the size and idle power further by another half at least.

Therefore, a 32 nm successor to Moorestown may be the right thing then, for say a standard 2.7 to 3-inch screen smartphone, like the Glofiish M800 we reviewed yesterday.

What will all that mean for the Taiwanese vendors entering the smartphone market on a global scale against Nokia, Samsung, Motorola and SonyEricsson? Well, if the ARM is sidelined in a long run, it means less choice and less variety - we all now how "widespread" is the smorgasbord of X86 choices, after all.

So, from that viewpoint, ARM should be encouraged to live and thrive, especially if its smartphone CPU vendors can pool some funds together to have up-to-date common code ports of all major Net and mobile office applets, plug-ins, players and such - not to mention faster CPU speed-ups to keep pace with Intel.

On the other hand, an X86-based smartphone will ultimately look more and more like an Ultra Ultra Mobile PC - UUMPC, with all the architectural, BIOS, driver and other approaches similar to a PC. Now, that is an area where Taiwanese do have an advantage over the Nokia and the rest of the big gang. And, Asus, Acer and the rest aren't short of the money either - so, we could be in for a more exciting stuff in our pockets soon... µ