We’ve heard that chips such as GT200 and G92 were Nvidia’s investment in the future. If you look close enough you will see that GT200, a 240 Shader chip, is nothing more than an advanced and stronger version of G92 architecture and this is the path that Nvidia is taking for the time being.
At the same time, Nvidia's harsh lesson was that R700's innovative architecture was enough to beat Nvidia’s GT200 in the performance and high-end segments and that their architecture was simply not enough for the performance crown.
At 55nm, we believe Nvidia can and will put even more Shaders at higher clocks in the new GT200 and this might be enough to get some performance increase and end up slightly faster than R700, Radeon HD 4870 X2.
You've seen that removing one shader cluster makes a GTX 260 out of GTX 280 and if you remove even more from the top end chip, you will end up with a performance chips for different market segments. The big problem of this strategy is that if you have 10 clusters each with 24 shaders and if you enable only seven of them for a product, the chip will cost the same as if you had all 10 clusters enabled. The safe net is that you might have some chips where one of the clusters won't work properly and you should be good to launch GTX 260 and use these wisely. The way to get rid of this issue is to design a brand new chip with seven clusters making it smaller and cheaper for some other market segments.
Eventually mainstream and entry level products based on the same architecture will end up with slower clocks and less Shaders. The only big problem for Nvidia is that ATI’s RV770 has better performance per watt and this is something that Nvidia has to fix in the next generation.