GPU Technologies - Hyper Z HD
Hyper Z, ATI's technologies that cover Z-buffer optimisations and performance improvement have also had an update from what's present in R3x0.The Hyper Z engine is fed by the geometry setup engine which then performs per pixel optimisations before passing on pixels to the rest of the render pipeline.
Hierarchical Z is again present, now able to discard up to 256 hidden pixels per clock cycle, before feeding non-hidden pixels to the fragment shader pipeline. Like NV40, X800 XT PE can do 32 Z/stencil buffer ops per clock, which should help to accelerate games that do a depth/stencil pass before full rendering. The big recipient of such a speed boost should be obvious.
While ATI doesn't offer up any actual numbers for its Z buffer compression ratio in R420, with R360 peaking at 8.8:1, assume it's either identical or more capable. Of course, the compression scheme is lossless.
Fast Z clear is a given, but it's unclear to whether it's actually any faster or more efficient at clearing the Z buffer than any R3x0 chips, but it's safe to assume it's either equally as good, or better. The picture being painted of R420 being R3x0 on steriods is becoming clear.
