There is something unique in growing from 20 to 21. "Towering in the confidence of 21," as Samuel Johnson writes, or "When the brisk minor pants for twenty-one," as Alexander Pope writes, a new world opens out.
FPGA and Apple computer, born in the Orwellian year 1984, celebrate their 20th birthday this year. FPGAs are making deep dents into ASICs. iSuppli estimates that in 2003 there were 14,000 FPGA design starts, as against merely 1,400 for ASIC. Xilinx estimates that in 2004 FPGA design starts will number 100,000, as against 1,500 for ASICs, a precipitous decline from a 1996 peak annualized rate of 10,500, estimated by ATR.
ASICs are trying to fight back. Moore's Law advanced ASICs, with cell-based ASICs offering performance and integration unavailable except through full-custom design. With sub-90nm geometries, the very flexibility that made cell-based the preferred ASIC architecture puts this architecture beyond reach of many users and of doubtful feasibility for many systems. Some designers believe that cell-based architecture will pave way for structured ASICs. Among top ASIC vendors who have joined structured ASIC bandwagon are NEC, Fujitsu and LSI Logic, following smaller vendors like Lightspeed and Chip Express.
Perhaps it is too early to say that structured ASICs will replace cell-based ones. Large cell-based ASIC players such as IBM, STMicroelectronics, TI and Toshiba have yet to get into structured ASICs. When the focus is on high volume and high performance designs, cell-based designs are likely to dominate. Though the volume production of structured ASICs is growing fast, it is still small compared to cell-based.
I think issues involved are beyond merely cell-based vs. structured ASICs. The biggest problem ASIC developers face today is design, both in terms of cost and time. While some designers use costly EDA tools, a growing class of designers believe the present generation EDA tools is inadequate, and the new generation will be costlier. In contrast, many FPGA design tools are available as freeware from FPGA chip suppliers, especially at low end.
EDA vendors are going into FPGA tools. Mentor Graphics has rolled out a physical synthesis tool for FPGA. Cadence has invested in Hier Designs, an FPGA tools startup, while Synopsys is getting directly into the market.
There are contradicting reports about FPGA impact on EDA. ATR anticipates trouble for EDA industry. Says a recent ATR report, noting FPGA rise, "...a decent component of EDA software for ASIC designs will vanish from the addressable market space for EDA vendors." Because of this, some analysts fear that EDA industry will not grow at all in 2004, notwithstanding an anticipated 25 to 30 percent rise in semiconductor industry. However, Gartner Data-quest estimates that with the growing EDA SW price, EDA will grow 14.9 percent in 2004, with total revenues reaching $5.5 billion in 2007. Whatever that be, FPGAs are radically altering ASIC and EDA industries.
Like most people in the world, I generally use PC, not Mac. I am glad that Mac has survived this long through various iterations—Mac II, PowerBook, PowerMac, iMac. If I find PC more functional than Mac, I find Mac more inspirational than PC. If I find PC "extroverted" and "global" in its outlook, with Microsoft OS, Intel microprocessor, and other units procured from Taiwanese, Korean, Singaporean and Japanese companies and bundled together to meet market demands, I find Mac "introverted," saying silently that really great designs come not from market demands but from some deep and unknown part of one's own self wanting to express itself to the world. If PC appeals to my intellect, Mac appeals to my heart.
Happy birthday, FPGA and Mac! Unto you I say what Shakespeare says in the Twelfth Night, "Then come kiss me, sweet and twenty."
You can reach Kirtimaya Varma at
kirti.varma@rbi-asia.comkirti.varma@rbi-asia.com
