by Anton Shilov
10/20/2011 | 03:40 PM
Although Apple is developing chips for its ultra-portable products like iPad or iPhone, it will barely be able to develop central processing units that are powerful enough against x86 micro-architecture. As a result, Apple’s chips will not power Macintosh computers, but Intel Corp.’s products will remain inside Macs.
Throughout several decades many companies used to develop microprocessors for personal computers and/or servers. Such companies employed from hundreds to thousands of engineers, but at this point there are just two companies designing competitive chips for PCs, five companies developing processors for servers and a number of designers trying to make competitive chips for mobile, embedded and other industries.
Apple has never attempted to develop its own central processing units and used off-the-shelf chips for many reasons, the main of which is state-of-the-art technologies required for competitive microprocessors along with experience, patent portfolio and so on. But Apple started to develop its own system-on-chips for ultra-portable products like iPhone or iPad.
With its low-power A4 and A5 system-on-chip devices, Apple saves great amount of money as it does not have to pay premium for SoCs from Nvidia, Qualcomm, etc. Moreover, designing system-on-chips does not necessarily mean creating from scratch and then hand-tuning all the units inside, but involves licensing and tweaking them as well as adding certain unique logic. The company has a lot of low-power processing experts and continues to hire chip design specialists, thus the A-series SoCs will only get more competitive.
“Apple is bringing in-house some of the design aspect in the form of SoCs, to get exactly what they want and not a carbon copy of some other (e.g., Samsung) design. They will never ever sell that part, it will be exclusively Apple, and we're going to have a devil of a time trying to figure out what's in it and what its specifications are - all part of Apples master plan to keep its secrets secret,” said Jon Peddie, the head of Jon Peddie Research.
But it is not only much easier to develop a mobile SoC than to design a fully-fledged microprocessor that will be able to power a desktop or a notebook, it also makes a lot more economic sense for Apple. The company presently sells around a hundred of millions of iOS-based devices that are also powered by A4/A5 SoCs per year, but the volumes of its Macintosh offerings are much lower (around 20 million per year). Hence, per-chip development costs may be too high even in case Apple actually creates CPUs competitive to Intel’s or AMD’s.
“Apple's unit volumes for notebooks is way smaller than for iOS devices, and the R&D needed to field a competitive notebook CPU would be way higher, so the economics don't work,” said Nathan Brookwood, the principal of Insight 64.
There are a lot more reasons for Apple not to develop chips for its Macintosh desktops and notebooks.
Many microprocessor technologies are patented by AMD, Intel, IBM, Sun/Oracle and some others. Hence, Apple risks running into patent wars against established chip designers. It will take years for Apple to avoid such conflicts or develop its own patent portfolio and ink cross-licensing agreements.
“There's a huge IP hurdle for anyone trying to build an x86-compatible CPU. Any chip that claims x86-compatibility but lacks licenses from Intel and AMD would be a sitting duck for patent infringement claims,” said Mr. Brookwood.
Even if Apple manages to avoid legal disputes with other chip designers when creating its own CPUs, it will have to hire additional engineers and develop numerous technologies that will outperform off-the-shelf chips from AMD and Intel. To date, no company has managed to leave Intel behind in terms of performance leadership for a significant amount of time.
“Apple would have to out-engineer Intel on Intel's playing field. Of the long list of companies that have tried to do this, including IBM, TI, AMD, NEC, and STM, only one, AMD, has attained any success, and their journey has been uneven at best,” reckoned the analyst from Insight 64.
The success of modern high-profile microprocessors depends not only on technologies that are under its hood, but also process technologies that are used to make them; it is impossible to pack a billion of transistors into chips that are produced using an outdated fabrication process since it will be too expensive to manufacture it. In around three years time Intel will have 14nm fabrication process, whereas other players in the industry will be one step behind with 20nm process technology. This can easily transform into overwhelming performance advantage Intel’s chips will have over everything else, including potential Apple’s CPUs.
Finally, even in case Apple manages to develop an ARM-based chip with performance comparable or higher than that of leading-edge x86 microprocessors, it will barely be able to transit all Mac OS clients and developers from x86 to ARM. The transition from Power to x86 in the mid-2000s was painful, but allowed the company to eventually gain market share. With ARM, nothing is guaranteed.
To put it shortly, it not only does not make sense for Apple to kick off a competition against Intel – one of its primary suppliers these days – but it will be impossible for the company to achieve any success on Intel’s playground.