There’s A New Paradigm In Town

I recently wrote an article in the October 9th issue of EETimes that appears to have rattled the semiconductor industry a bit. Entitled “Wake Up, Semi Industry: System OEMs Might Not Need You,” the article conveyed the fact that many system-level OEMs now have the capability — and desire — to develop their own application-specific chips.

This may be news to many. But a simple review of recent semiconductor acquisitions by OEMs in the mobile space provides some insight. While Apple doubled-down with P.A Semi and Intrinsity, and Google acquired Motorola Mobility, there are numerous other back-room examples of chip designers acquiescing to their customer’s specific chip demands.

You might posit that chip companies have always produced two versions of their products, version A with the customer’s special sauce and a stripped down version B designed for the commodity market. So, this is nothing new, right?

Wrong. There are several other parts to the story that prove there is truly a new paradigm in chip design.

Greater Control Equals On-Time Delivery
Arteris recently conducted a survey of its chip design customers to assess their key design challenges. One of the top concerns recorded was time to market, or schedule slips. In the rapidly accelerating world of consumer product launches, having a product on the shelf on time for critical season product cycles can mean success, or layoffs, for many companies. OEMs that have mastered design efficiency are capturing market share and market revenue. Upon evaluation, those winners have increased their number of design starts and have received their chips on time.

In fact, you can calculate the cost of being late, both in terms of market share and loss of revenue. Consider the typical product sales window of 18 to 24 months and apply the amount of anticipated revenue associated with that specific chip. Using a formula based on proven economic research — and actual industry experience — we derive that a company can lose more than 34% of anticipated revenue with a three-month slip.

A three-month slip in chip delivery can mean a missed product window for the system OEM. Better get a tight grasp on that chip design status.

Around the World in a Day
To capitalize on the global availability of talent, and reduce development costs, chip companies now rely on distributed design teams. Each design team is tasked with a specific portion of the SoC, based on their skills and expertise. The basic chip architecture may be designed at headquarters, then passed on to the graphics team, followed by the processor core team, the audio/video processing team, and finally to the connectivity team. In fact, the piecemealed chip design may traverse the globe every day during its creation, only to be cobbled together by the interconnect team upon completion. And the chip is expected to work.

Fortunately, the increasing use of licensed semiconductor IP has eliminated the need to reinvent the wheel every time a new chip is designed. Now, globetrotting chip designs can leverage time-tested IP nuggets to ease the design burden and increase the probability that the chip will “crank” the first time.

A Chip by Any Other Name
We’ve established that time to market is critical and chip design is a global endeavor wrought with potential mine fields. Given this scenario, why would a system OEM even consider treading into the fray? Because the heart and soul of their products are ruled by devices with network-like structures that can truly differentiate the product. “System-on-chip” is becoming synonymous with “network-on-chip (NoC).” And the processing and functionality demands are increasing almost twofold every two years with required design cycles shortening to nine months.

Analyst Jim Tully stated in the recently released Gartner Hype Cycle for Semiconductors and Electronics Technologies, 2013 report that network-on-chip (NoC) technology reduces design time, offers more effective reuse and leads to faster throughput. Given advancements in NoC technology and the ever-ticking product time clock, is it any wonder that system OEMs want more control over their own destinies—and their own chips?