TSMC’s 20nm node will be used by Apple, Qualcomm and their few VIP customers, most of the manufacturing companies are looking forward to the 16nm FinFET technology. TSMC’s one of the company will give the brief demo this month on the 4th Annual Conference of the supply chain, then the progress of the company’s 16nm FinFET and 10nm process will be announced. And it seems like the previous rumors come true as the first to adopt 16nm FinfET technology label has been received Nvidia. Nvidia’s Maxwell architecture Erista processor will be the first to use the 16nm process node. Erista is the successor of Tegra K1 and would be released in Q2 of 2015.
NVIDIA first to use TSMC 16nm FinFET technology, Maxwell Architecture Erista Coming in 2015
Currently TSMC’s 16nm process has been fully certified by the reliability and by the end of 2015 nearly 60 customer solutions would be using 16nm process node. TSMC as expected in July of next year start the mass production of 16nm FinFET technology. Capital spending for next year from $9.6 billion this year to $10 billion.
According to information released by TSMC company, Avago, Freescale, LG Electronics, MediaTek, NVIDIA, Renesas Electronics and Xilinx (Xilinx) and other companies will be the first to use 16nm FinFET technology.
As mentioned the NVIDIA, GTC General Assembly in March this year announced a new generation of the NVIDIA GPU and Tegra roadmap successor after Maxwell architecture is Pascal which will come out in 2016, using 16nm FinFET process. And after its successor Tegra K1 is Erista, this is the son of Wolverine (Logan) and logan is just Tegra K1 code.
The Erista Tegra processor architecture will be upgraded to Maxwell architecture, but did not mention the CPU architecture, but 64 Dever core NVIDIA has been applied to a dual-core version of the Tegra K1, Erista should also use the Denver architecture CPU.
For now the Specifications of Next-generation Tegra processor are unclear, but according to NVIDIA announced before the news, Erista the GPU will be upgraded from the current Kepler to Maxwell architecture, more outstanding performance per watt ratio. Judging from the indicators given NVIDIA, Tegra 4 per watt ratio (single-precision floating-point GFLOPS /W) than 20GFLOPS / W, Tegra K1 can reach 40GFLOPS / W (Tegra K1 floating point performance is said to reach 326GFLOPS) while Erista will exceed 60GFLOPS / W, under the same power performance than Tegra K1 upgrade at least 50%.
By convention, Erista most likely to be published in CES2015 which will be held in January of next year, but the actual product will be available from the middle of next year. After TSMC’s 16nm FinFET technology will be in the mass production in July.