Days later I ventured here to Xilinx to connect with Steve Glaser,
senior vice president corporate strategy and marketing, and an old
Cadence Design hand. I tried to get him to rise to the bait. Altera
is faster out of the blocks on 20-nm; Xilinx is sometimes quicker
to market but Altera gets it 'right,' some say. Blah blah. He
Sure, his slide decks had the traditional compare/contrast of the
two companies in key areas, but in the conversation, which went on
for an hour, he said nary a negative word.
Instead, he said, talked about programmable logic vendors needing to solve larger system-design issues. "It's a corporate
transformation story becoming an industry-transformation story,"
He went on:
"I'm talking about giving an alternative to implement
systems. A lot of our customers have both of us. In the end, it
really comes down to value. If we can't demonstrate unique
value--either of us--we're not going to flip the competitor into
our space, and we're not going to expand the available market.
It's not just about A v. X. A lot of our growth is not versus
Altera. It's about the next-level of integration--in integrating
that CPU or DSP or those traffic-management functions, that may
have been done earlier in an ASSP or an ASIC."
Glaser said today 60 percent of the business is fighting for the
socket, but 40 percent is fighting for "system value," or that next-level integration, and that's a
fundamental difference between now and then.
Cost, perspective, pedigree
That system focus has not come cheap: The company has spent $100
millon retooling its software suite, Vivado, and another $300
million in development of various sorts for 28-nm, including 3-D IC
and SoC technology.
But over the years, Xilinx has pulled in non-traditional
electronics-industry executives to bring a new vision to their brand
of programmable logic semiconductors. Glaser's pedigree not
only is Cadence but the old ASIC powerhouse VLSI Technology; CEO
Moshe Gavrielov came out of EDA (Verisity and Cadence). An earlier,
and successful, Xilinx CEO, Wim Roelandts, brought in a system-level
perspective from 30 years at Hewlett-Packard, the last of which was
running HP's Computer Systems Organizations.
Glaser started at Xilinx last year. It could be he's staying above
the fray and really working with his team to create a different
positioning for the company. Or it could a lingering honeymoon
before the fights renew at 20-nm. Xilinx is expected to start teasing
its technology advancements at that node in Q4.
that is very true @reneCardenas...which is probably why we have Xilinx-Altera duo-poly...Atmel has a small niche and others like Tabula are trying but chances are small that they will join the big 2 because of teh sheer size of the investment required
But doesn't the sheer size and time required to establish a good tool ecosystem prevent others to join in the viable system solutions.
I wish other well, but see their chances very grim to raise to the level require to present any challenge.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.