D3D RightMark and 3DMark03 Texel Fillrate
The X700 XT's performance in Rightmark, especially in the pixel fill tests, are a little strange. The pixel fill tests are short of their expected values. The RV410's geometry performance, by virtue of the same six full MIMD vertex shader units as R420, is shown off here, eclipsing even the 6800 GT. In pixel shading terms, it's neck and neck with the 6600 GT.
Overall, disregarding the odd results in the fillrate tests (more on those soon), the X700 XT is as good in Rightmark's pixel shader and geometry tests, as 6600 GT.
Futuremark's 3DMark03 Fillrate Tests
When Rightmark's pixel fillrate tests showed odd results on the RV410, I turned to 3DMark03's texture fillrate test to double check that things were actually OK in another simple raw fillrate test (no trilinear filtering, this time a texture fillrate test with multitexturing). That test sees it as follows.
In the 3DMark03 case, X700 XT's fillrate appears to be optimal, for its clock speed. Plotting the fillrate numbers against their possible peak figures gives a percentage efficiency.
NV45 and NV43 show close to peak efficiency in the fillrate case in 3DMark03's test. NV43's written pixel fillrate limitation due to ROP count is ignored in this test due to its setup (it's a texture test using alpha blending, so you can count processed texels before final ROP output).
X700 XT is highly efficient in this test too, not far from that of NVIDIA's recent architectures. Maybe ATI has some memory access work to optimise with X700, to edge closer or ahead?
