- ARM the Reigning King of the Smartphone World

Currently, the smartphone CPU market is dominated by ARM, which just designs and licenses their CPU cores instead of manufacturing actual chips unlike Intel (which also doesn’t like licensing their CPU core designs).

By just being a designer/licensee, ARM has created a very vibrant ecosystem around their CPU core and became a de-facto standard in the mobile device market. Many large and knowledgeable semiconductor companies (Samsung, Motorola, Qualcomm, nVidia, Apple etc.) as well as many small ones are building ARM-based CPUs contributing their expertise (GPU, memory interfaces, crypto subsystems, etc.) and diversifying the product offerings. And the variety and availability of compatible ARM-based CPUs offers abundance of competitive choices for the device manufacturers.

ARM has always been power-conscious coming from the embedded device / micro-controller world. They have been improving their architectures for more compute-intensive applications and soon we will see dual-core versions of ARM CPUs and with out-of-order execution (as opposed to with in-order execution). They were replacing 6800-based microcontrollers for the past decade and growing steadily. However, their visibility and popularity exploded with the introduction and huge growth of Apple’s iPhone and iPod Touch.

ARM is also trying to move into a server market where it is increasing getting costly to provide electricity to thousands of racks of servers and to cooling systems. The recent news about ARM and the recent purchase of Agnilux by Google confirm their direction, which is their move into the Intel’s territory.

- Intel Inside in Smartphones?

Interestingly, Intel used to make ARM-based CPUs called XScale. However, they sold that division to Marvell (probably they began to think about their smartphone strategy at the same time).

On the power-conscious mobile side, Intel has been working on the Intel Atom lines of CPUs. The first generation products (codename Silverthorne) were designed for miniature PCs (or MIDs), not really for smartphones. It was underpowered for what it was mostly running on (i.e., Windows), wasn’t as power efficient as ARM, and can’t be made smaller due to its size and the number of components.

So, would this new Intel Atom Z6xx (codename Moorestown) make any difference in the smartphone market? In summary, I don’t think so. But it’s a stepping stone to the 2011 part (codename Medfield) which might make a difference.

Z6xx seems to have improved vastly over its previous generation when it comes to the power consumption. On the paper, it could be even competitive with Snapdragon-class ARM CPUs power-wise, but we would know for sure when someone actually tests and verifies the claim when a working product ships.

However, there are at two problems that I can see with Z6xxx.

First, There are too many components. Z6xx still carries Intel’s PC lineage: it has a separate southbridge and a separate RAM module. With ARM, taking Apple’s A4 for example, it has all that (CPU, southbridge and RAM) included in one chip. This impacts mostly the size and cost of the devices, and somewhat the power/performance efficiencies.

Second, It’s x86-based, and there is no compelling, consumer-oriented, smartphone OS for it. They can’t use Microsoft Windows, which isn’t designed for smartphones. There isn’t an x86 version of Windows Mobile (although Microsoft could decide to port it). It’s not like Apple will port their iPhone OS to x86.

Intel does have a Linux distribution called Moblin/MeeGo, but it’s not really suited for the consumer markets. I think Android is the right fit for it, and Intel is working on an x86-port of Android (most Android apps are Java and run on Java Virtual Machine, so the underlying CPU does not matter as much). However, we don’t know when it will be ready (there is a community effort for x86 Android already, though).

So, I think Z6xx is going to be Intel’s best-effort product, a filler, a stepping stone, toward their eventual goal for smartphone CPU. ARM will maintain its indisputable smartphone CPU dominance for, at least, two or three more years.

- 2011 and Onward

However, Intel’s 2011 offering make things much more interesting. Codename Medfield will be building up on Moorsetown and combine the southbridge with the CPU/GPU. We don’t know if the RAM module will be combined or not yet, but it’s a possibility, even the current embedded device trends.

Also, by the time Medfield comes out, it’s very likely that the x86 Android port would be stable enough and ready. Intel would have improved on the power consumption even more also.

The only question with Medfield is how Intel is going deal with CPU core licensing. If Intel maintains their current stands, they would be the only supplier of Medfield chips and I don’t think that would scale well. They will need to come up with a CPU core licensing scheme that will let others take their design, customize it and differentiate their products.

Surely, by that time, ARM would also have improved their design. So, Intel would need to be very competitive and willing to lose money on this line of products for, at least, a couple of years.

But by 2013, if Intel executes well, they will have a competitive product against ARM although I believe it will be extremely challenging for Intel. They will have to figure out what their business model (mainly licensing CPU cores vs. doing it all themselves) is going to be.

Even with all these, Intel may not make up any significant market shares in the smartphone/MID market. But it’s probably better than not having any presence.