Semiconductor IP News and Trends Blog
IP Adds New Twist to Intel-Apple-ARM Triangle
Market forces often create strange alliances. Remember IBM, Motorola and Apple during the height of the PowerPC days. Now a new trinity is appearing on the horizon, this time amongst Intel, Apple and ARM. But this new alliance adds a strong IP twist, especially if Intel begins to make ARM chips for Apple Macs.
Many pundits have commented about Intel’s challenges in the mobile space—from smart phones and tablets to GPS devices. Others commentators wondered about the effect of this relationship triangle on existing strategic alliances, such as those between Apple and Samsung or Microsoft and Intel. Still others noted that Intel’s timely announcement of the creation of 3D transistors might tip the low-power dominance from ARM to Intel. Admittedly, few believe this last conjecture.
One aspect of an Apple-Intel-Arm scenario that remains uncovered is the question of semiconductor IP implementation for any future chips. How well will ARM’s IP play on Intel’s existing notes and the new 3D transistor architectures?
Apple’s move to an Intel fab would be great for Intel, as it would allow the IDM giant to quickly recover revenue from its recent, huge investment in 22-nm manufacturing facilities (see JB). Furthermore, the new fab would support FinFET (or tri-gate) transistor designs.
But how many manufacturing partners does Apple need? The company already has relationships—tenuous ones, to be sure—with Samsung and TSMC. Both Intel and Globalfoundries would be future candidates. But more isn’t always better. Using the same chip and mask sets for these different fabs would be time consuming and expensive.
The question of IP is a bit more interesting. The IP shift to a FinFET is probably much more complex than to UTB-SOI, notes Dave Lammers, editor-in-chief for SemiMD. “For its 14-nm FinFET technology, TSMC stated the need to first develop the SPICE models and then work with the IP vendors.” This sequence is the result of greater separation between the circuit and transistor development.
As for Apple on FinFETs, I think this is a slam-dunk for Macs. In iPhones, it probably won’t be because of pricing, energy consumption, and software. Even if Intel can get the price and power consumption down–and I gather the main thrust will involve the Atom processor–the challenge of porting everything over to an Intel chip isn’t so simple. The alternative is a Microsoft-based phone, which hasn’t fared well for similar reasons.
How removed is the IP architecture of a chip from the physical construct of the actual chip?
Not that much, cautions Neil Hand, group marketing director, SoC Realization at Cadence. “IP has always been tied to the process and is becoming even more closely tied—not necessarily because of the move to 3D transistors, but rather because of the move to more advanced geometries.” The relationship to the process differs based on whether we’re talking about hard or soft IP.
Hard IP has a tightly coupled architecture that’s determined by the underlying process capabilities and physical properties, Hand explains. “It would be impossible to divorce hard IP from the process. As a result, IP companies will be required to have deep process expertise in-house.”
On the other hand, soft IP isn’t as closely tied to the underlying process as hard IP. Still, an understanding of process capabilities does allow the IP architecture to be optimized for better performance and power. Leveraging the process benefits means that the same IP can be used on different platforms—just as the same video-decoding/decompression IP can be implemented in everything from handsets to home theaters.
IP is so closely tied to the process architectures and will increase in complexity at advanced nodes—regardless of advances in transistor technology, such as 3D chips. This adds another dimension of complexity to the proposed ecosystem triangle between Apple, Intel, and ARM.