Rumors and speculation build up a lot of excitement in both the video games and tech industries, but confirmation they are true – well, that’s something entirely different. TSMC have finally confirmed that high performance nodes are going to jump from 28nm to a 16nm FinFet+ process. This will drastically impact GPU designs and manufacturing, since both AMD and Nvidia rely on TSMC to produce their silicone.
Cadence are one of the unsung heroes in the tech industry, and their software and hardware help power the automation which creates electronic circuit design (which is… well, kinda important), this makes them one of TSMC’s most important and primary collaborators.
A report on the companies website reads as follows: “Through an ongoing, multi-year agreement, Cadence and TSMC will develop the design infrastructure for 16nm FinFET technology. The collaborative effort will target advanced-node designs for mobile, networking, servers, and FPGA applications.
“Addressing the complete flow from design analysis through signoff, the companies will provide an infrastructure for the design of ultra-low power, high-performance chips. At advanced FinFET nodes, you’ll gain an infrastructure that supports the accurate electrical characteristics and parasitic models needed for successful design implementation.”
The exciting part is that the publication points to manufacturers (using the 16nm FinFet+ node) being offered a choice between either having their products having a performance improvement of about 40% and using the same power draw compared to the older 28nm process, or using 50% less power, but still having the same levels of performance as 28nm.
This indicates we’ll not see GPU’s using this process until very late this year, or early next year (so either AMD 300 series rebrands, or their 400 series, or for Nvidia’s Pascal architecture, which is touted to have huge performance gains over Maxwell). Obviously, this will mean flagship GPU performance will shoot up compared to this generation, and even lower power devices (such as the GPU’s in SoC’s) will perform significantly better – which will be interesting for AMD’s Zen based APU’s – released next year.