Xilinx 'open for business' at 20nm, but signals the end of process shrinks

The portfolio provides a road map for its Virtex and Kintex families, both of which take advantage of TSMC's 20SoC process. With the move to 20nm, Xilinx is focusing the parts on particular markets. Kintex devices, which will be available first, have more dsp48 slices, suiting those building signal processing applications, while Virtex devices have more transceivers, including some running at 32.75Gbit/s. This suits them to comms heavy designs. EMEA marketing director Giles Peckham told New Electronics that UltraScale will offer designers the ability to move between Kintex and Virtex parts. "Customers will be able to build systems with, for example, high end and low end variants, and get more cost effective solutions," he said. UltraScale will also see the end of using process shrinks for new devices. "The industry is moving to a multinode strategy," Peckham claimed. "With the move to 20nm and 16nm, we're not getting the cost per transistor benefit we used to, so there is no advantage in process shrinks. However, we will do so if the economics make sense." This will see the Artix range, which has been expanded to five devices, stay at 28nm. Kintex will be focused on 20nm, while high end Virtex parts will move to 16nm in the future. * Included in the announcement is the Virtex UltraScale VU440 3D IC (pictured). With 4.4million logic cells, or 50m equivalent asic gates, it offers more than twice the capacity of the Virtex-7 2000T. The device will be targeted at prototyping and emulation, as well as high value, low volume 'number crunching' tasks.