The graphics card market is hard and rocky and requires time and again for new ideas and innovations from AMD and Nvidia - yes, certainly Intel and S3, but we can at this point in exceptional cases. In any case, Nvidia CUDA in things apparently smelled a rat and now uses every effort to further expand this area before competitor AMD's Stream and Advanced Parallel Processing can establish in the first place.
The codename of Nvidia's latest investment is "Denver Project" and what is behind it is quite interesting. This is based on a) the licensing of the Cortex CPU A15 from Nvidia and b) the licensing of the ARM architecture, which allows Nvidia to devise a powerful ARM CPU itself. This should be combined eventually with an Nvidia GPU, the target audience here is clearly on supercomputers, but also to other classes of devices are operated.
Back in September showed a roadmap for future Nvidia GPU generations. As suggested by Tesla and Fermi were already there continues with substantial physicists. For the year 2011 is "Kepler" is provided, which made significant discoveries in astrophysics, 2013, "Maxwell" to reach the market, who formed the basis for theory of electricity and magnetism.
The first products based on Project Denver will come 2013, matching the Maxwell generation. This could also explain why can be reached on the CUDA GPU roadmap such a large jump in the GFLOPS / watt efficiency.
Unfortunately, there are - as so often - not exactly as that, and we must for now be content, but we wonder whether the project a success or is it a bad investment. By the year 2013.
source:pcmaster/ComputerBase