So Much More Than Gate Arrays
A couple of weeks after I joined Xilinx in 2008, one of my new colleagues gave me an original copy of the press release announcing our first product, the XC2064, which would eventually come to be known as the world’s first FPGA. The public release date was November 1, 1985 and the announcement came from legendary Silicon Valley public relations firm Regis McKenna. Every once in a while, I’ll pick up the press release and give it another read, because it reminds me how far the FPGA industry has come. The contrast between what FPGAs were like in 1985 and what they can do today is remarkable—so much so that I think many of them have outgrown the moniker “field-programmable gate array.” Well, at least the “gate array” part.
It’s no easy task naming innovative technologies. Back when Xilinx and Regis McKenna were launching the XC2064, it appears that they hadn’t quite figured out what to call the device. In fact, the headline for the press release reads: “Xilinx Develops New Class of ASIC.” Meanwhile, the first paragraph describes the device as follows: “The new device, called a logic cell array, offers a high level of integration together with the versatility of a gate array-like architecture.” The release goes on to describe many truly revolutionary features, including “unlimited reprogramming without removal from the system,” “64 configurable logic blocks and 58 I/O blocks…[and] 1,000 to 1,500 gate equivalents.” These were all quite impressive for the time. Clearly, Xilinx had set its sights on displacing PALs and “bipolar and CMOS programmable products, LS TTL [low-power Schottky transistor-transistor logic] components and gate arrays.” Who knew at the time that this new class of device would one day compete with ASICs and consolidate the functions of many chips into one?
I started covering the EE design space in 1995, and so I missed the point in time when these “logic cell array ASICs” became commonly known as field-programmable gate arrays (FPGAs). What’s a bit puzzling to me is how “gate array” got into the name? In 1985, the public relations team likely called the devices “a new class of ASIC” because even back then, ASICs were starting to rapidly displace gate arrays. By the time I began writing about electronic design, gate arrays were gone, daddy, gone. I get the field-programmable part. That’s an engineering-esque way of saying that the device was reprogrammable. But “gate arrays”? Really?
Certainly, the devices we offer today are so much more than gate arrays that are "reprogrammable". Back in 2008 we began executing on a plan that redefines the possibilities of programmability. At the 28-nanometer mode, we delivered three lines of All Programmable FPGAs (the Virtex-7, Kintex-7 and Artix-7 devices), but didn’t stop there. In its 28-nm generation, we also delivered the first homogeneous and heterogeneous All Programmable 3D ICs, which shatter capacity and bandwidth records, respectively.
What’s more, Xilinx is delivering to customers today the Zynq-7000 All Programmable SoC, which marries an ARM dual-core Cortex-A9 MPCore, programmable logic and key peripherals all on a single device. On top of these silicon innovations, we launched a fresh, state-of-the-art design suite called Vivado to help customers leverage the new device families to achieve new levels of system integration, reduce BOM costs, lower power consumption and speed system performance. Not only do the latest offerings have innovative programmable logic, they also boast programmable I/O, DSP slices and embedded processors, making them also software programmable—they are truly All Programmable, enabling truly reprogrammable systems, not just reprogrammable logic. They certainly are not gate arrays!
Mike Santarini is the publisher of Xilinx’s Xcell Journal. Prior to joining Xilinx, Santarini was a senior editor at EDN magazine, EE Times and Integrated System Design magazine. He holds BA in English from Santa Clara, University (1995).