TSMC deal offers glimpse of Intel future

Intel is making a bid to become a force in smartphones. This will test its ability to compete in arguably the most important chip market outside of PCs.

The deal struck this week with Taiwan Semiconductor Manufacturing Co. will put the Intel architecture into the same factories that churn out chips for companies like Qualcomm and Texas Instruments, which use an alternative architecture called ARM--the choice for many small devices, cell phones, and most smartphones, including the Apple iPhone, BlackBerry Storm, and Google-based Android phones.

ARM has always been a thorn in Intel's side. So much so that Intel acquired the StrongARM architecture in 1997, turned into Intel XScale, and aimed it at handhelds (most prominently iPaq handhelds sold by Compaq and then Hewlett-Packard). Before that, StrongARM had been used in the Apple Newton (a primitive precursor to the iPhone) and other small devices.

But Intel sold the money-losing XScale business to Marvell in 2006. And so ended Intel's attempt to compete ARM to ARM in the small device space.

Intel processors are not a player in the market for smartphones like T-Mobile's G1, which uses an ARM-based Qualcomm chip

(Credit: T-Mobile)

What happened? The small consumer device and communications chip business is not the PC business and, consequently, not an area where Intel has historically been competitive. But that doesn't mean Intel can afford to ignore this space. Handheld personal computing has arrived (if you hadn't noticed). The iPhone, Blackberry, and Android phones are virtually handheld PCs--with Intel processors nowhere to be seen.

So this time instead of coming up with an ARM chip, Intel is trying to shoehorn its successful x86 architecture into the ARM universe of smart phones, consumer electronics, and the amorphous, though typically profitable, "embedded" market. TSMC excels in building chips for all of these markets. The world's largest contract chip manufacturer operates successfully on gross margins much lower than Intel's enviable x86 PC margins, typically north of 50 percent (as this CNET Blog Network piece points out).

And one market where Intel would like to succeed (and some would say must succeed) is smartphones because of its sheer size and because "that's where the PC functionality is moving toward," said Doug Freedman at Broadpoint AmTech. Though markets for hardware that goes into, for example, industrial or medical hardware, will be important, it's the smartphone market that will test Intel's ability to compete profitably in a consumer space outside of PCs.

Just how big is the overall cell phone market? On a unit basis, it is about five times the size of the PC market. There were about 1.22 billion handsets shipped in 2008, while the PC market is forecast at 257 million units in 2009, according to Gartner.

But Intel cannot operate the way it does in the PC world--where its credo almost seems to be: if we build it, they (HP, Dell, Acer) will come. This won't work in the cell phone industry. Service providers and handset makers are center stage, hardware is at best a side show. So, hooking up with TSMC is a way for Intel to make itself more palatable to cell phone companies, which are not used to dealing with the 800-pound PC chip gorilla. "By going through a TSMC, it is perceived less as an Intel move and more as, hey, I'm just another source for you the handset maker because you're already used to buying stuff from TSMC," said Ian Lao, an analyst at In-Stat. "It's insulating the gorilla thing."

And it's none too early. Qualcomm is now pushing the performance envelope with its Snapdragon platform, Nvidia is hawking its graphics-intensive Tegra technology, and Texas Instruments is revving up its OMAP chips to achieve better performance per watt.

In other words, while these chip companies are not wavering from their longstanding strong suit of power frugality--an imperative in the cell phone world--they are also beginning to ratchet up chip speeds to 1GHz and above and add more processing cores. And that's Intel's strong suit.

"For ARM developers, multi-core implementations will address much of the performance differential," said In-Stat's Lao. Look no further than Qualcomm. The future Qualcomm QSD8672 chip will be a dual-core Snapdragon that features two CPU computing cores capable of 1.5GHz performance, 1080p high-definition video, Wi-Fi, mobile TV, and GPS. The graphics core is based on Advanced Micro Devices' ATI unit's technology.

Hmm...Dual-core processor, ATI graphics, high-definition video? Sounds a lot like a PC. Indeed one of the burning questions is whether PC makers will begin running Microsoft's operating systems on ARM-based devices, according to Lao.

"The next 9 to 18 months will be quite interesting to watch," he said. "Can Intel get down to the cost and power levels needed? Will they be able to get the carrier and handset makers aboard? There will definitely be a market shakeup."

[slickuser]

x86 will make developers life easier. No more expensive debuggers required for
embedded system development. No more custom development platforms...

designers can use now build their components (add-on cards etc) using Atom based PCs.
This will make their R&D cheaper.

There are plenty of x86 debuggers, IDE and other developer tools available
for cheap/free at developers disposal.

[tasehagi]

"one of the burning questions is whether PC makers will begin running Microsoft's operating systems on ARM-based devices"

that's for Micro$oft to decide, not PC makers...

[pithenumber]

and given MS's is quite open to anyone running it on any hardware, MS will probably give it the green light. What would they lose?

[nanikore]

Microsoft operating systems ALREADY run on ARM-based devices... remember?

Windows CE
Windows Mobile

[pithenumber]

*cough*
he means a "full" pc version of Windows, like XP, 7 and Vista

[skrubol]

MS may make Windows Mobile or CE more like full Windows, but there's no way they're going to port Windows to Arm. Arm devices are not as powerful as desktop CPU's, so Windows Mobile/CE is a much better fit, plus there is tremendous cost to porting something like Windows. Not to mention there will be no apps to run on the new OS.