So, 16 x 4 = 96 Pixel processor's? Nope

ATI really pushed this 'similar' R300 Architecture to R520 now, the major difference what I see is increased Pixel and Vertex Pipelines, the layout of those Pixel or Vertex Processors may be different (should be) and ofcourse Core speed, Memory speed / Interface and fabrication process. This 'Quad' (block of 4 pixels) isn't bad mind you, this is actually a strength. Although, from what I've read that most games (current games) only uses Pixel Processors, but not Vertex Processors, are they increasing more Vertex processors just for the sake of 3D Mark?..I do feel though.

Although the Quad Architecture does look promising (example LOD operations), as each Pixel Pipelines divides itself in a group of quads or four. Although I would like to know more about R520 Architecture (Vector, Scalar etc.,),I would also like to know more about Texture Units on R-520 more, as in R420 it comprise of two Shader Units (72bits) and a single Texture Unit (24bit) which can do work in Parallel, hence the advantage .. and how wide these units are in R520? the operations per Pixel Pipeline per clock really enhances the efficiency, but how many operations those blocks can capable to push? as what we have seen in R420 Architecture, although it will be hugely dependent on and F-buffer also.

The branch prediction logic being used in R520 does look promising also, it does means minimal errors, and it can massively increase the performance.

Now, the Core speed and Memory speed has been increased also from 600Mhz & 700Mhz to 625Mhz & 750Mhz respectively, so it certainly giving 10,000 Fill rate, just a tad slower than Geforce7 7800GTX, but since the Memory speed has also been increased, so yea I would say the final revision of R520 does hold some promise. 512mb of Memory is certainly giving it finishing touch.

I would say it's quite a nice Article Sander, thanks for providing us some pictures and info on the core.

Kudos !