NVIDIA G80 - Foxconn GeForce 8800 GTX review

Somehow, it feels only right that we should start this major article with a confession - NVIDIA managed to throw us (and by that I mean all of us graphics enthusiasts who love to speculate on unannounced products) a massive dummy in the years prior to the release of their first DirectX 10 architecture.  After reading so many interviews and comments from NVIDIA sources stating that a unified shader architecture wasn't necessarily the right way to go circa 2006, we had all convinced ourselves that, unlike ATI's R600, G80 would take a far more traditional route towards creating a DirectX 10 compliant part, with at most only vertex and geometry shaders being unified.  We speculated on a non-unified architecture with thirty-two ALUs, maybe 6x multi-sample anti-aliasing if we were lucky, and so on.

Boy were we wrong.  Last week, the GeForce 8800 GTX and GTS were officially released, and finally confirmed just how misguided our speculation was.  Quite simply, you'd be hard pressed to find many more divergent architectures compared to what had gone before in the history of GPU and graphics board development than what we have sat in front of us with G80.

On the GeForce 8800's launch day, we took a brief look at performance using Foxconn's GeForce 8800 GTX part - Now, we follow through with our promised full review of not just this board, but the nitty gritty of the G80 core itself.  So, sit down, pour yourself a coffee, take a deep breath, and prepare to find out everything you need to know about the PC's first DirectX 10, Shader Model 4.0, unified shader architecture.