Will nomad make Xilinx nimble?
08 July 2008
Moshe Gavrielov, CEO of Xilinx, hopes to make the leading programmable logic company nimble and fighting fit.
Speaking to Moshe Gavrielov, the new president and CEO of Xilinx, brings surprises. First of all, the Israel-named, Californian looking 54-year old, reveals that although he was born in Israel he grew up in Hendon, north London. “I am a Spurs fan,” he admitted, “I have a high threshold for pain.” After graduating from the Institute of Technology (Technion), in Haifa, Israel with a bachelor’s degree in electrical engineering and a masters in computer science, he stayed in Israel, joining National Semiconductor as a young engineer. Designing high-performance 32bit microprocessors “was a dream” he wistfully recalls, reminding me that this was in 1979. After six years, he moved to America as a microprocessor designer for Digital Equipment on the east coast. Then he joined the ASIC work with LSI Logic, working in various roles from engineering and management. He had a sojourn in England again with LSI, working in the Bracknell European headquarters. He fondly remembers the old building with a clock tower. His route to the programmable logic world was via Cadence where he was general manager after the EDA company bought Verisity where he was CEO in 2001.
He began 2008 with a new job as CEO of Xilinx. The quietly-spoken man has already made a mark with a streamlining of the company’s workforce creating what he calls a functional business, instead of several independent business units. The move to a more holistic approach made headlines only when 250 job losses were announced. These were shared among IT staff, finance and marketing as the new model meant that several roles were duplicated across the former business units.
It seems that the technology and market-oriented groups overlapped in the model he inherited and he believes that the company will operate more effectively, using resources that are more easily moved around as a result of moving from a business unit model to what he terms a ‘functional’ model. He often refers to getting the company to be ‘nimble’, anticipating perhaps a tough fight ahead as programmable logic’s target markets, consumer and automotive are hit by the credit crunch and subsequent slow-down in consumer spending.
Now, there are three groups in Xilinx; the software and IP, known as the solutions group, the silicon group and a single marketing group to prioritise the market’s need. The solutions group delivers and the silicon group implements the products, with the solutions group, explains Gavrielov, making it clear, despite the use of the vague ‘solutions’ term. “We revamped the portfolio and cancelled low ROI projects, moving resources to the critical 45/40nm generations.” The company’s Virtex-5 is already available at 65nm node and an announcement on the next-generation of 45/40nm technology is promised for the end of this fiscal year.
The separate teams for Virtex, the high-performance FPGA, and for Spartan, the economy version, are no longer separate teams. Some capabilities on Virtex are drifting to the Spartan world, explains Gavrielov. Although he makes the point that, despite that, unique elements have to be introduced. There are three design teams at the remodelled Xilinx, a Virtex team, a Spartan team and one where elements for both groups fit. The goal is to make sure as much as possible falls into the third category, smiles Gavrielov. In this holistic approach, the engineers can prioritise the customer’s requirements in a design.
Looking at the figures, the company’s $1.8bn revenue has stalled in recent years and with 20 per cent allocated for R&D, there is clearly a need to grow. The way to do this is to grow the markets using programmable logic. Seven years ago, Xilinx’s majority interest was in telecomms, an unequivocal 80 per cent. Now it is 42 per cent, with military accounting for 15 per cent and the combined consumer and automotive sectors accounting for 17 per cent. Industrial, including test and measurement and industrial controls and ISM (industrial, scientific and medical) are also significant areas now for the company. Interestingly, Gavrielov describes automotive as an emerging technology for the company. It concentrates more on infotainment “rather than ‘under the hood’ – or as you would say, ‘under the bonnet’ electronics,” he grins, reminding me that he is part Londoner under that north California tan.
There are 15,000 active Xilinx customers and programmable logic now has to be cost-sensitive and power-sensitive. This is a huge challenge, says the CEO, and there is lots of fertile ground for creating specific, programmable capabilities.
“I call it the Programmable Imperative,” he says of the need to capitalise on programmability and broaden the offerings to more markets. “I coined that phrase and think it’s good... although no-one else seems to...” he trails off, reminding me of a dad’s wounded expression when his children don’t appreciate his corny jokes.
He is eyeing up ASIC’s and ASSP’s $130bn market which dwarfs FPGA’s $3.6bn and believes FPGA’s programmability is the key to address new areas with technology. That is why Xilinx is concentrating on power and costs alongside the shrinking geometries.
How about Structured ASICs, I ask. After consideration, Gavrielov is candid: “I thought it was a bad idea and it has failed.” In eight years’ time, he believes that Structured ASICs will be left to small companies or start-ups. “I would put money on a programmable world – I AM actually putting my career on it!” he laughs.
The Programmable Imperative encompasses his hope that the company’s mission statement of programmable silicon in every electronics product extends to a recognition of serving business through programmable logic as a means to enable customers to get differentiated solutions for target markets based on programmable logic fabric. How the company does this is of no concern to the customer, believes Gavrielov. The customer experience is driven by tools, the breadth of IP and ease of getting the product to the market. Even if the architecture is critically important, hiding it with software and interfaces is the right way to go to achieve this Programmable Imperative.
I wanted to reassure him that if others did not appreciate his marketing concept of Programable Imperative, he must keep his chin up! We must be democratic - I have plugged Programmable Imperative concept here and now it is over to the industry to see if embraces it as wholeheartedly as it appears to have embraced the technology.
Contact Details and Archive...
Most Viewed Articles...