ARM announced recently the availability of Osprey, which is a hard macrocell of a dual core Cortex-A9 processor customised for production on TSMCs 40G process. Furthermore, this is open for licensing today, and can be integrated into any System On Chip that a customer wishes to create.
This contrasts highly with the approach taken by Intel. For production of an Atom Soc, Intel has decided to create an archaic 2-chip package, whereby chip 1 (called Lincroft) will house all of the graphics, the CPU and an integrated memory controller and chip 2 (called Langwell) will house all of the IO capability (disk controllers, and however else that is defined). Finally, and if you want to build something with bluetooth, wifi or 3G connectivity then you need yet a third chip, which is called Evans Peak. The key thing here is to notice that as portability goes up, so too does the number of functional packages (Lincroft, Langwell and Evans Peak will be required to make a decent smartphone, only Lincroft and Langwell required to make a larger MID device). This, in the arena of any portability centric device industry, is absurd. The second thing to note is that a device manufacturer is not able to choose what graphics, or which CPU, they have from the intel line-up - whatever Intel says, goes. The only differentiation being given is in the IO Hub (Langwell) to a small degree.
However, the real difference that Intel now faces from the ARM camp is that ARM now have a high performance macrocell of a dual core CPU that outperforms any Atom in a lesser power envelope, that individual companies can take and integrate as they please with whatever IP they fancy, in whatever package they wish. The real trick up the customer's sleeve is that the place and route and floorplanning of the macrocell has already been done by ARM for the customer - all they have to do is take that box and integrate it into their system, and concentrate their efforts on everything else, without having to worry about hitting frequency goals in the processor as well.
Intel's arguement that the mobile industry is fragmented is the main one given for not allowing differentiation w.r.t. the Atom platform. They effectively want to do the same thing in the mobile industry that was done to the PC industry, and create an innovation-stifling set of standards so that system integrators can start differentiating on..... oh, wait, they can't differentiate. A standard foisted on the community that is so restrictive will lead business to only one door - Intel.
The mobile industry has operated under its own steam for years, and has been influenced by very different socio-economic factors to those that drove the standardisation of the PC Industry. Fashion, entertainment and most of all extreme portability are the key-factors driving the industry. And in parallel, standards that enable that to fluorish have emerged. Bluetooth addresses the issues surrounding sharing information in-situ with another person; Wifi addresses the need to connect to the internet in an unfamiliar setting where there may not be a cable that fits your device; 3G is the all encompassing protocol that carries both voice and broadband through the towers of the company that contracted out the device on a payment plan.
Each of the aforementioned standards don't relate in any way to a straight-jacket on how a system can be built nowadays. In fact, companies seek to innovate the way that these standards are integrated into their platforms. And companies have sprung up that create IP for each of these standards in licensable form, with tools that allow system integrators to integrate those standards seemlessly into their systems. A whole architecture of system integration components exists that become more optimal for portability with each passing generation, and introducing a standard integrated platform mastered by only one company will cripple the fantastic work undertaken in these areas by numbers of people across the industry that far exceed the numbers of people employed by any one company.
So it is two very different worlds colliding. One takes a hands off approach, leading to explosive levels of innovation that can only be achieved by such an open model; and the other is driven entirely by the mechanics of domination, to achieve a controlled mono-culture where only those admitted to the party can prosper.
We've seen what happens in the latter case before, and it leads to some decidedly megalomaniacal business practices in order to hang on to your precarious position.
My opinion is that the latter world cannot continue to prosper, and as ARM moves upwards into the higher performance computing envelopes, people will get sick and tired of the strangulation being caused by Intel, and no amount of strongARM (pun intended) tactics will be able to outweigh the power that open innovation models have in getting their foot in the door of new and exciting industries.
Posts
