Xilinx’s Process Advantage
Xilinx’s aspiration to achieve five consecutive years of market gain share, now in its third year, is based on technology leadership, says Steve Glaser, svp for corporate strategy at Xilinx.
Xilinx has made the right choice in sticking with TSMC’s technology, added Glaser. While arch-rival Altera is, of course, going to Intel for foundry at 14/16nm.
“TSMC has a multi-year leadership position in true 20nm technology,” says
Glaser, adding ” what Intel calls a 22nm process is actually a 27/28nm technology.”
One of the reasons for TSMC’s leadership at 20nm is that it was the first company to do double patterning, asserts Glaser.
Despite this, Glaser says: “Our competition is not doing a high end 20nm device at all.”
Asked why not, since Altera also uses TSMC for 20nm, Glaser replies: “The high-end requires special tools and Altera hasn’t got them. We re-architected our tools for the high-end with Vivado. Altera never re-architected their tools by writing the whole thing from scratch.”
At the next technology generation – the 14/16nm node – TSMC is the best choice, argues Glaser, pointing out that, while Intel and TSMC converge on density with both companies using both double patterning and finfet at 14/16, TSMC has three big advantages: it is the only proven foundry source for both 3D IC and ARM-based SoC technology; it offers ‘vastly superior’ design enablement; it has ‘significant’ services and long-term supply advantages.
Asked when Xilinx would start sampling products on TSMC’s 14/16nm process, Glaser replied “next year.”
Asked when he thought Altera would get access to Intel’s 14/16nm process, he said: “Intel’s SOC process typically lags its CPU process by up to a year.” Intel’s first 14nm product is scheduled to go into production in Q1 2014.
The Xilinx -Altera ding-dong on process technology goes back a couple of decades. In recent years, Altera creamed Xilinx at 40nm, but Xilinx beat Altera to the 28nm node.
Now it’s seconds out for the 20nm node.