let see what will really come to the market. Couple years ago when Altera planned to go with 32nm, they were offerred 28nm by TSMC. That's why there's no 32nm chips from them. People are trying to make Moore's law sustainable. Who knows in next-next-year, what technology will be proposed!?
This is a war between two different approaches of designing transistors: FD-SOI (AMD, ST, GlobalFoundries) vs FinFET (TSMC, Intel, GlobalFoundries)
FD-SOI (Fully Depleted Silicon On Insulator)roadmap has renamed the 20nm node as 14nm.
The argument is that FD-SOI transistors at 20nm node match in performance with 14nm FinFET ones.
The point is 3D geometry in transistors: now they have a significant "volume" and not only a "length".
Just to make things more confusing, I've heard rumors that some companies (not FPGA companies of course, because they would not stoop to this sort of thing) are re-braiding their 20/22nm offering as a "16/14nm equivalent" ... I'm not too sure about who is doing this or how they justify it ... if you have any information on this please post it here...
I'm losing the ability to keep track of what's going on, because there's so much of it.
Altera and Xilinx are currently at the 28nm technology node, and they've both announced a roadmap to the 20nm technology node.
Meanwhile, the folks at Achronix have just announced that they've begun shipping their Speedster22i FPGAs that are built on Intel's advanced 22nm, 3D Tri-Gate transistor technology.
TSMC have only recently started talking about their own FinFET technology at 16nm, while Intel's second generation 3D Tri-Gate transistor technology will be at the 14nm technology node.
Now we hear that Altera plan on introducing a future generation of ultra-high-performance FPGAs using Intel's 14nm 3D Tri-Gate transistor technology.
I may be wrong, but I think this is the first time an FPGA company has announced a "next-next-generation" technology before their "next-generation" technology has hit the streets. We truly do live in exciting times!
Blog Doing Math in FPGAs Tom Burke 13 comments For a recent project, I explored doing "real" (that is, non-integer) math on a Spartan 3 FPGA. FPGAs, by their nature, do integer math. That is, there's no floating-point ...