Former Intel CEO and Chairman, Andy Grove, was fond of speaking about inflection points. In a 1998 Keynote Speech, Mr. Grove’s own words strike a very prophetic tone, given recent Intel history.

“Strategic Inflection Points. They represent, in my description of it, what happens to a business when a major change takes place in its competitive environment. A major change due to introduction of new technologies. A major change due to the introduction of a different regulatory environment. The major change can be simply a change in the customers’ values, a change in what customers prefer. Almost always it hits the corporation in such a way that those of us in senior management are among the last ones to notice.” – Andy Grove

Intel had the foresight to understand that integrating basic graphics acceleration would allow them to position cheaper, more competitive platforms into the business and education markets. The graphics in all of Intel’s integrated core products are descended from C&T (Chips & Technologies) which Intel purchased and merged. C&T was an old graphics chip company back in the day whose comtemporaries included Western Digital (YES, the HDD vendor Western Digital), Tseng Labs, Cirrus Logic and S3. However, hindsight has also shown that Intel did not take the concept far enough. While business and education markets made up a significant portion of Intel’s revenues, it was the consumer side of the business that provided much of the company’s profits.

Intel somehow missed the key inflection point that 3D graphics and media playback/streaming were key drivers in the consumer space, as Intel’s efforts in these areas have been lackluster at best. AMD, long the underdog and a favorite of the enthusiast community, had a better understanding of this market when they merged ATI. And while Intel’s integrated Core CPUs are still the fastest around, they are getting their head handed to them in 3D gaming applications by AMD’s Fusion APU platform. Despite years of careful investment by Intel building up the integrated graphics market, AMD has gate-crashed this party and is making speedy inroads in both mobile and desktop segments.

More recently ARM, an even scrappier upstart than AMD, has started to go head to head with Intel in the mobile segment. ARM has traditionally been limited to industrial and vertical applications where their cheap, low-power consumption processors were best suited for devices that needed to be ruggedized and/or used in demanding environments. Typical uses of ARM processors in the past have included industrial power supplies, industrial control equipment, automotive computers, GPS and more recently eBooks. But with Apple’s game changining iOS devices, ARM finds itself riding on Steve Job’s coattails and expanding their business into the coveted, mass-market consumer space.

This would seem to be another key inflection point missed by Intel. As recently as 2006, Intel’s XScale processors, designed by the StrongARM team Intel picked up from DEC in 1998, were used in digital communication devices many of which found their way into PDAs and early HTC smartphones. But by the end of 2006, then Intel CEO Paul Otellini decided to throw in the towel as the ARM business was not generating anywhere near the volumes they wanted to see – so they sold the business off to Marvell for USD$600m. As history would have it, Apple launched their iPhone at WWDC in June of 2007, barely 7 months after Intel sold off its ARM CPU group.

Intel must be kicking themselves right about now, as they watch ARM gobble up market share in eBooks, smartphones and tablet computing devices. Unfortunately, this is a race that Intel can’t win in the short term. Intel’s CISC (complex instruction set computing) CPUs require lots of additional logic (and die space) which make them consume more power, run hotter and cost more than ARM’s RISC (reduced instruction set computing) processors. As an example, Apple’s A5 CPU is using 45nm design rule (4 year old technology), and it runs the iPad 2 for roughly 8-10 hours of use per charge. Intel has a gaping hole in its product range – it doesn’t have a single product that can do what the A5 does for Apple.

What this means is that Intel is on a very slippery slope. Intel relies on process shrinks to control the power envelope on their mainstream CISC CPU designs. This strategy requires massive amounts of investment into research and capital equipment to enable – and Moore’s Law guarantees at least an 18-month cycle time between process generations. Intel has made very little headway into the tablet space with their ATOM line of CPUs, which don’t have the graphics muscle to stand up against solutions such as NVIDIA’s Tegra 2 and upcoming KAL EL (Tegra 3), both ARM based solutions powered by NVIDIA’s graphics cores. Even the legacy of poor graphics performance is hurting Intel here as these slick new consumer devices have user interfaces that require more graphical power than CPU power.

If ARM licensees wanted to push down to 32nm or smaller, Intel would never make up the power gap using current designs. Even supposing substantial reduction in TDP by clock throttling Ivy Bridge, it still presents a very poor price/performance argument for tablets. Future iterations of ATOM on 22nm process will continue to be outclassed by NVIDIA’s Tegra 3 and Apple’s A5 and upcoming A6 on the graphics front. In a nutshell, Intel has a major imbalance issue – their CPUs are too powerful while their graphics cores are too weak.

So while Intel is finding it difficult to power down to enter the ultra-mobile space, ARM has the luxury of ramping up clock-speeds and performance and moving up into larger form factor appliances. The first step in this battle is being enabled by Intel’s erstwhile partner, Microsoft. The next iteration of Windows (Windows 8), will have support for ARM CPUs out of the box. Which means that ARM is closer to entering the netbook/notebook space, than Intel is at entering the ultra-mobile tablet space.

Given Intel’s two big misses on these strategic inflection points, both seemingly hinging on poor graphics performance, it finds itself fighting a two front war with AMD and ARM. Both are encroching on Intel’s lucrative mobile processor business, while AMD is starting to take ever bigger chunks out of the entry level consumer desktop segment.

Sources: Intel, The Register – AMD merges ATI, AnandTech – AMD Llano performance numbers, The TechReport – Intel sells XScale to Marvell, AnandTech – Windows 8 to support ARM

Share This With The World!