As you might expect Nvidia's 9800 GTX has a lot in common with all the recent cards they've released in that it is based off the G92 core that the 8800 GT, 8800 GTS 512MB, and more recent 9600 GT all share. To sum it up quickly the major difference between this and the G80 core the previous generation of DX10 cards were based on is the die shrink to 65nm to cut down on die size, costs, and power consumption. It features 128 shaders with a default clock speed of 675MHz for the core and 1.69GHz for the shaders with a 2.2GHz memory clock speed with 512MB of RAM using a 256-bit memory bus. This might seem a little odd to some people who know the GeForce 8800 GTX specs as that card featured 768MB of RAM with a 384-bit memory interface, so higher in regards to the GeForce 9800 GTX. That said there were more than likely some efficiency issues in regards to texture management which have been resolved with better drivers and revised hardware but it still seems on paper to be one of the few times a newer product has lower specs than the previous generation. ATI also did this with their Radeon 3870 moving from a 512-bit memory interface to 256-bit and that card was faster than the 2900 XT. The stream processor number holds steady at 128 though the clock speed is 100MHz faster than the default of the GeForce 8800 GTX which was 575MHz compared to 675MHz.



Link: AMDZone