ARM quashed Apple purchase rumor

Will Strauss, president and principal analyst, Forward Concepts

USA: There was a flurry of excitement when the April 21st edition of the London Evening Standard stated, "Apple is ARM's biggest customer and speculation is that the iPad maker wants to take chip design in house." ARM's CEO essentially threw cold water on the idea, and Apple's CEO was not talking (surprise!). But, it's interesting to think of the possibility.

Apple was one of the original founders of ARM through a 1990 partnership with U.K. computer maker Acorn and semiconductor house VLSI Technologies (later acquired by Philips/NXP). At the time, Apple was interested in adapting Acorn's new RISC processor for use in the company's new Newton Message Pad which was promoted for its handwriting input methodology.

The Newton was not a success (some blame the satire by the Doonsbury cartoon strip for its demise) and Apple divested most of its ARM shares afterwards. Apple returned to ARM processors with the first iPod in 2001 (which employed a pair of ARM 7TDMI chips from PortalPlayer, Inc.), and has since used ARM processors (mostly from Samsung) in its iPods, iPhones, and, most recently, the iPad. Maybe Samsung should consider buying Apple (see below)? The London speculation was that ARM stock, if acquired by Apple, would likely be picked up for about $8 billion.

Apple acquires A4 underpinnings
Although no explicit announcement has yet emerged, Intrinsity, Inc., the company responsible for the 1GHz performance of Apple's A4 processor chip employed in the iPad tablet, appears to have been acquired by Apple.

As I have been quoted in other publications, the A4 is based on ARM's Cortex-A8 core, fabricated for Apple by Samsung and employing Intrinsity's technique for speeding up operation and at reduced power consumption. The Samsung name for the generic 1GHz device is "Hummingbird." We believe that the acquisition is both to improve the Apple product line and to deprive potential competitors of Intrinsity's worthy IP.

The Cortex-A8 core can run no faster than about 750 MHz in a 45 nm process (the Cortex-A8 based Samsung Application Processor in the iPhone 3GS runs at 600MHz), and it would take at least another 18 to 24 months to re-architect the A8 for faster operation using conventional techniques.

Hey, it took Qualcomm at least two years and $300 million to goose their version of the A8 (in their Snapdragon) to 1GHz. The Intrinsity approach required less than 12 months for the new SoC speedup...and that was their first time. In an earlier interview, Intrinsity executives indicated to me that they could do it for another client (ARM has more than a dozen Cortex-A8 licensees) in four to six months.

We believe that the looming iPhone 4 (see below) will also employ the A4 processor. Moreover, we believe that the Intrinsity IP can be employed in the newer ARM Cortex-A9 core, as well. A scenario we predict is that a successor to the Apple A4 will employ the "conventional" ARM Cortex-A9 core (or cores, since it is available with up to 4 SMP cores), and a faster follow-on product could employ the Intrinsity approach, enabling annual faster and lower-power follow-on processors.

Apple's PA Semi acquisition had little input on the A4
Early speculation on Apple's new A4 processor was that it was designed with the talent that it acquired with the purchase of PA Semi appears to have been incorrect. Shortly after Apple's acquisition, a number of key ex-PA Semi employees (including its founder) left and formed a new company, Agnilux. And, as stated above, there's little likelihood that the remaining PA Semi talent could have engineered a Cortex-A8 boost from 600MHz to 1GHz in the few months they were part of Apple.

Just over a week ago Google announced that it had acquired Agnilux. Since Agnilux was a semiconductor design house, it appears that Google has a plan to employ the acquired low-power, high-performance processor expertise for a future chip product.

One could argue that most of the Agnilux employees were from the software side of PA Semi, but over half of the employees in most chip houses are software people. Another possibility is that Google acquired the new talent to design more efficient processors for servers that could lower the significant power cost of its data centers. Such an approach would be similar to that of Austin-based Smooth-Stone Inc., which is employing multiple ARM cores for a low-power-consumption server.

Another possibility is that Google is looking to design its own application processor for smartphones. Google, of course, earlier entered the CDMA smartphone market with the Nexus One, manufactured by Taiwan's HTC Corp.

Since a version of Qualcomm's Snapdragon communications processor provides both modem and application processor functions in the Nexus One, Google may have a plan to produce its own lower-power, higher-performance ARM-based processor. Of course, Qualcomm will still have the CDMA modem socket, since it is highly unlikely that Google (or anyone else) would want to develop a new advanced CDMA modem chip.

Maybe Texas Instruments is correct in its assertion that application processors (like their upcoming OMAP4, based on the Cortex-A9) and cellular modems are on different roadmaps, with application processors updated almost annually, while modems have a much longer design-in life because they have to adhere to type approvals tied to multi-billion-dollar infrastructures.

Samsung lays out ARM roadmap
In a move that has to be sending chills down the backs of both Intel and Qualcomm, Samsung, the world's second largest semiconductor supplier, has indicated a corporate roadmap for a series of ARM-based chip products, starting with a Cortex-A8 based chip running at 1GHz (the S5PV210), slated for 3Q/10 mass production.

That appears to be a variant of the "Hummingbird" chip mentioned above. Introduced as a Netbook solution, there's no reason that the Cortex-A8 couldn't be an application processor for a future Samsung handset (if that side of the company deems it worthy).

Since Samsung is Qualcomm's largest cellphone chip customer, such a move could impact future Snapdragon chip sales. What Intel has to worry about is that Samsung promises "browsing speeds as fast as X86." Clearly, the gauntlet has been thrown down...at least in the Netbook market.

Samsung's "Orion" chip, scheduled for 3Q/10 sampling, is based on the more advanced dual-core Cortex-A9 clocked at 800MHz, followed in Q2/11 by a single-core A9 ("Pegasas") at 1GHz. A dual-core A9 clocked at 1GHz ("Hercules") is slated for sampling in 3Q/11. Note that the Cortex-A9 can handle up to 4 cores, so there is room for continuing product growth. Compare this roadmap against my Apple roadmap speculation, above.

With so many ARM licensees planning to invade traditional X86 markets, maybe Microsoft would consider porting Windows 7 to the ARM Cortex-Ax chip family. Otherwise, Microsoft may lose out to the Linux crowd (and all of its variants, like Android, MeeGo, Ubuntu, etc.). With Windows Phone 7 soon available on ARM mobile platforms, it's clear that Microsoft is very familiar with ARM architectures.

AT&T blocks employee vacations in June--for iPhone 4 launch?
This happened when the iPhone 3GS launched last year, but boygeniusreport.com has confirmed with multiple AT&T sources that the carrier has now put a block on employees taking vacations in June. Since the iPad 3G began selling last Friday, it is likely that the nixed vacations are to support sales of the next iPhone (iPhone 4?), which will likely launch that month. We expect that the announcement will be at the Apple Worldwide Developers Conference beginning June 7.

HP's Palm acquisition: My take
Although Hewlett Packard is the largest producer of personal computers by a large margin, they register just above the noise level in the cellphone market. Since the cellphone market is the only one that can rival the PC market in size, it made sense for HP to beef up its lackluster Windows-based iPAQ smartphone product line through its acquisition of struggling Palm and its more exciting smartphone product line.

Palm's WebOS multitasking operating system is quite good and its Pre and Pixi cellphones are credible offerings. Unfortunately, Palm was late to market and didn't have the resources to build penetration and keep up with a fast-changing market.

HP has more resources, a wider strategic reach and better marketing than Palm and I believe that Palm's innovative portfolio and talent pool provides a stronger basis for future HP smartphones. I think it was a great acquisition for HP.

Samplify: The little company with a better 4G infrastructure solution
We often encounter startups with a better idea, but it takes financial backing, a hard work and good luck for them to be successful. One such startup we've been tracking is beginning to make a difference in the cellular base station market.

Samplify Systems LLC (Santa Clara, CA) began with its Prism signal compression algorithm IP employed in, among other things, contributing to cheaper base station migration from WCDMA to LTE. LTE requires about 6X the fiber bandwidth (over WCDMA) between baseband and remote radio units on the tower, and Samplify's approach reduces the fiber optic bit rate, extending CPRI interface capabilities.

Samplify also offers its own upconverter/downconverter IP, and to complete the base station signal chain, it has partnered with much bigger IDT, which offers CPRI mux/demux chips, SERDES and SRIO switches.

Samplify also offers a 16-channel compressing ADC chip and has recently introduced its own Autofocus ultrasound beamforming IP for improved imaging. Those don't relate to the wireless messages in this edition of the newsletter, but we are impressed by the novelty of Samplify's technologies.