Next DirectX to force Standard Precision
Apparently Microsoft?s DirectX team plans to drop partial precision capabilities in the next version of DirectX, not sure if that’s 9.0L or 10. This means both ATI and Nvidia will finally be forced to render using 32 bit floating-point precision, which in turn means many of the unscrupulous tactics used by both to gain performance advantages are now gone ? I can guarantee ATI are smiling right now?
Dropping to different levels of floating-point precision has been a favourite trick of both ATI and Nvidia for quite some years, allowing higher frame rates in important benchmarks – Nvidia arguably being the ?king? of these shady practices.
Partial precision can be used to limit the bit precision of pixel shader processes on the fly. Optimisations can then be achieved by switching to 24 or even 16 bit precision for better performance albeit at a price to image quality if compared against true 32bit floating-point rendering, which apparently ATI now use anyway with there latest cards.
With DirectX 9 the use of partial precision was left to the game or graphics driver, now this ability will be stopped, leaving in it?s place 32 bit floating-point rendering only. All this boils down to better image quality for us all, and no more false marketing..
With Microsoft putting an end to this, both Nvidia and ATI will now have to compete on a level playing field, with some of the major issues of Nvidia Vs ATI image quality now destined to become a distant memory, and I for one am seriously glad.