The guys at The Enquirer seem to have some dough about the next generation GPU from  nVidia , In an analyst webcast, Nvidia Investor Relations and Communications VP Michael Hara stated the top end card based on the G92 graphics processor will be ready for Christmas and that it will have computing power close to 1 TeraFLOP .
Time for some cools stats now , Intel’s latest quad core Core 2 extreme QX6700 runs at 2.6 GHz and has a peak floating point performance of 50 GFLOPS, while nVidia’s G80 ( powers the 8800GTX)  has a peak floating point performance of 330 GFLOPS, AMD’s R600 can do a max 450 GFLOPS  and the STI Cell B.E does close to 250 GFLOPS at max throughput . So if the G92 rumor holds it would mean that the G92 will outperform the cell by around 4 times and its predecessor G80 by about 3 times  in terms of computing power.
One problem with GPU’s  is the inability to support 64 bit floating point operations, which is necessary for almost all supercomputing applications . On the other hand GPU’s are far cheaper than supercomputer vector processors making their use in HPC( High performance computing) justified . nVidia had promised FP64 support on GPUs by late 2007 , So is G92 the promised chip , if so its going to have a big impact on the GPGPU movement and on HPC in general.
One little concern for me is the lack of titles for the PC which can take advantage of such horsepower , Crysis is a contender and so is SC-Conviction . TeraFLOP or not it’ll be interesting to see what kind of performance can the G92 can dish out.