Although there have been too many rumors without any basis on the Kepler family of GPUs from Nvidia. From SemiAccurate report that Nvidia has finally completed the design (taped-out) of your GPU Kepler GK110.
The chip GK110 has a size of die approximately 550mm ², size slightly greater than 520mm ² GPUs Fermi GF100 and GF110 used in the upper range of GPUs Geforce 400 and 500 Series respectively, and obviously greater than 324 to 361mm ² approximate GK104 Kepler chip, although less than 576mm ² of Tesla GT200 GPU (Geforce 260 to 295).
Nvidia actually ended their chip design GK110 few weeks ago, but such information has been disclosed recently. After having frozen the design stage of the chip, should spend a few weeks polishing some bugs and leakage power, and then finally enter the mass production process, for it is estimated that GK110 will be released during Q3 of this year (July to September), or remain optimistic from May to June.
According to Charlie Demerjian, GK110 continue using a memory bus of 384 bits, denying rumors of a possible use of a 512-bit bus, though you do not have any data on chip performance, this GPU should leave Geforce commercially under the nomenclature Geforce GTX 670 and GTX 680, both models differing by only some units disabled.
It is noteworthy that according to the nomenclature of Nvidia, the chip Kepler of 1 generation of more power should be called GK100, but apparently the GPU had to be canceled by themes of enormous energy leaks that kept their consumption around 300W, so Nvidia decided to replace it with a more evolved version of the chip (GK110: the 2nd generation architecture Kepler) with many of these defects corrected.
GK110 joins other chips ended as Kepler and GK107 GK104, although it is rumored that there is also a GK105, which has not yet have any information. We will keep you informed.