Moshe Gavrielov, ceo and president, Xilinx

NE: Why has the value proposition of the FPGA has been steadily growing? MG: In 2009, Xilinx celebrates its 25th anniversary, marking an important milestone for the company. Not only has field programmable gate array (FPGA) technology stood the test of time – evolving dramatically to meet customer requirements for a variety of applications – but FPGA platforms are also increasingly at the heart of today’s most innovative end products. Indeed, we’re just beginning a new era of electronics in which programmability is an imperative. As the industry moves to 40/45nm silicon process technologies, mask costs will easily be in excess of $1million during the first year of deployment. The complexity of 45/40nm design will not only demand the most advanced tool sets and design for manufacturing technologies the EDA industry has to offer, but also sophisticated engineering expertise. Altogether, this adds up to a tremendous investment in cost and complexity in next generation 40/45nm IC designs – an investment that can be justified in fewer and fewer applications. In fact, over the last couple of process nodes, an increasing number of ASIC and ASSP design groups are implementing designs in older silicon processes, foregoing the performance, power and capacity benefits of Moore’s Law. Unfortunately, this trend comes at a time when OEM and consumer demand for the latest and greatest electronic functionality is at an all time high for myriad end markets, such as automotive, wired/wireless communications, industrial, scientific and medical. For today’s auto makers, electronic features play a pivotal role in boosting sales and increasing margins. If you’ve shopped for a car lately, the dealer may well have tried to sell you the optional ‘basic technology package’ or even the ‘premium technology package’, with GPS navigation, lane departure warning and parking assistance, along with infotainment. Increasingly, ASICs – and even ASSPs – are falling out of favour in this segment. Developing a new chip for a specific automotive application requires manufacturers to meet stringent qualification requirements. From the time of initial design specification to the debut of a new car, the manufacturer’s requirements can change – requiring a level of design flexibility that cannot be easily or affordably addressed with a traditional ASIC or ASSP design. NE: What about wireless communications? MG: Mobile phone service providers across the globe are building the infrastructure to support next generation mobile phone communications. Most of these systems are based upon the 3GPP long term evolution (LTE) standard. Today, the standards body has not completely nailed down all aspects of the standard. Despite this, mobile infrastructure providers are already starting to deploy equipment in the field, specifically macro cell base stations for the LTE market that also support legacy standards such as 3G WCDMA and CDMA2000. Because the 3GPP standards body has not completely solidified the LTE standard, base station vendors have to ensure that the equipment they have already designed – and, in some cases, even manufactured – can be easily modified to support changes in standards. Ultimately, base station vendors would like their designs to be so flexible that they can sell one model of base station for several markets. Increasingly, ASICs and ASSPs, are no match for the job. For example, when LTE requirements change, designers must adjust their designs quickly, even respinning masks to meet the new requirements and risking the possibility of being late to market. Another late change to the standard could require modifications to the software running on an ASIC or ASSP, another redesign and manufacturing iteration of the ASIC or ASSP or the addition of another chip to the PCB, requiring technicians to travel across the globe to swap out PCBs for each base station. When you consider historical trends in wireless communications, the ASIC and ASSP story for next generation base stations is even more tenuous. There’s a high likelihood of several slight variations of the LTE standard, from country to country and from carrier to carrier. In some cases, there will be very large variations. For example, the People’s Republic of China plans to deploy TD-SCDMA, its own variation of the WCDMA standard. Still other parts of the world are planning to run WiMAX. Given the limitations of ASICs/ASSPs, a new design would be required to support each standard at a time when base station vendors are seeking to develop a single model that can be flexibly upgraded to meet evolving standards. NE: So this is where the pitch for FPGAs comes in? MG: absolutely right. If you’ve been in this industry for a while, you’ve observed that the value proposition for ASICs is largely going away, as is the value proposition for ASSPs. Conversely, the value proposition of the FPGA has been steadily growing. Indeed, the winds are in our sails. That said, as ASICs and ASSPs become less economically viable, we need to expand beyond our focus on FPGA capacity, performance and cost to also manage power and to offer an IP portfolio that strengthens our horizontal to vertical platforms. As Xilinx provides more powerful platforms, it will be able to capitalise on this momentum to take the FPGA business to the next level, delivering even more value to customers. In 2009, you’ll see that Xilinx – a company that has built the some of the most innovative programmable logic devices over the past 25 years – is dedicated to becoming the most innovative provider of FPGA platforms over the next 25 years. In fact, it’s our imperative.