Intel Targets ARM

According to a recent article, Intel wishes to enter the smart-phone market, which has so far been a mainly ARM market. The article correctly points out that the Intel Atom currently consumes too much power (a few Watts) as compared to ARM processors (under 100mW). So, from a technical stand-point, there is still a significant barrier to cross.

However, Intel has always been a chip-fab in disguise as a microprocessor company. So, I won’t be surprised if they managed to use their fabrication prowess to come up with an Atom that approaches the power envelope of an ARM processor. According to the article, Intel is targeting to push down its Atom power envelope around 50 times by 2011. Assuming that the surrounding hardware of both platforms are the same, then Intel will be in the right power envelope to mount a competition.

However, I’m not quite sure if they will succeed.

You see, the strength of ARM isn’t that they are very power efficient. There are plenty of power efficient RISC processors out there and the newest Cortex ARMs do suck quite a bit of power. I believe that their strength lies in their licensing. You see, you cannot buy ARM processors from ARM simply because ARM does not make any ARM processors. ARM’s business model works by having them license out their processor designs to other companies for manufacture and sale.

You can buy ARM processors from a number of different companies, such as Motorola, TI, Samsung and even, once upon a time, Intel (before they sold off their ARM division to Marvel). Therefore, in order for Intel to penetrate the smart-phone market, it would have to out-sell a bunch of other major chip companies, who will very likely gang up of Intel to keep them out.

Hence, this no longer becomes a question of technology. It’s a question of cost.

While Intel may be able to produce a low-power Atom, it could only feasibly do so using very specialised and bleeding edge manufacturing processes. However, the rest of the chip companies can produce an equivalent ARM using yesterday’s technology. This means that it will always cost more to produce an Atom than an equivalent ARM.

Additionally, a mobile phone company would be able to buy an ARM processor from any number of ARM licensees. Hence, there is market competition for price. If one vendor doesn’t give you a part at a price-point that you are looking for, just call up another vendor. It will be nigh impossible for you to buy an Intel processor from anyone other than Intel. So, there is less choice and competition.

Moreover, since ARM merely licenses out its core, there is also more innovation in the System-on-chip market. Each vendor is free to throw in any number of components around the ARM in order to produce highly cost-efficient and compact chips. For example, TI produces the OMAP platform, which marries an ARM core with their own DSP cores provide a boost in multi-media processing.

So, I don’t see how this could translate into a win for Intel. I’m not Intel bashing here.

Ultimately, in the real-world of design and manufacturing, you will want whichever part is cheaper. In this sense, ARM has the numbers on its side. Furthermore, Intel would need to take on not just one company, but a whole armada of large chip companies if it wishes to push ARM out of the market. Unless it has some way of subverting enough of them, it is unlikely to succeed.

There is one possible way though – Intel licensing out the Atom core. That will open up a whole other can of worms for them, but that may be the only way for Intel to compete directly against ARM.