Announcing the world's faste... does anyone really not know what this launch is about?
Internally the GTX 200-series isn’t too distant from G92, or even G80 (the 2, nVidia tells me is to denote that this is the second generation Unified Shader Architecture), but we’ll save the technicalities for out forthcoming full review. What is perhaps more important is that the new cards have, well, more. The 1.4-billion transistor die is, frankly, huge as both cards pack more shader processors than their equivalent predecessors – all of which are also more efficient. An improved Geometry shader is also boasted which should help with newer DX10 games which are starting to take advantage of that feature of the new(ish) API.
Non-gaming performance (read: CUDA) has also been given a considerable re-working with this generation of cards. Video processing has also been given an overhaul and Hybrid Power is supported across the board. Again, these are things we’ll cover properly later.
So the out-of-the-fab specs:
* 240 processor cores comprising.
* 602MHz core clock.
* 1,296MHz shader clock.
* 48.2 Billion/s texture fill-rate
* 1GB, 2,214MHz (effective) GDDR3.
* 512-bit memory interface.
* 141.7GB/s of memory bandwidth.
* 192 processor cores.
* 576MHz core clock.
* 1,242MHz shader clock.
* 36.9 Billion/s texture fill-rate
* 896MB, 1,998MHz (effective) GDDR3.
* 448-bit memory interface.
* 111.9GB/s of memory bandwidth.
As it is, most manufacturers will be shipping cards which vary from nVidia’s reference part, mostly in addition to a stock clocked model. MSRPs are slightly irrelevant at this stage, but if you check the following pages here I’ve listed the major board partner’s bundles and current pricing (roughly and when it can be found) and even some pictures for your viewing pleasure. In alphabetical order to avoid bias, and yes I know I’ve missed a few off.