Musings and Questions

As ever marketing material doesn’t always clear everything up; we at Beyond3D are inquisitive fellows so here are few of the musings and questions over what was presented to us on GeForce4.

References throughout the documentation were made to the “up to doubling” the effective bandwidth over the previous generation for the new Ti variants, however most of the operations listed were already present in the previous generation (Z-Compression, Crossbar, and occlusion culling) so how is this achieved? The occlusion culling is now said to be ‘second generation’ but what does this mean, that there are improvements to it or that this is just the second generation of chip that utilises it? From the ‘factor of 4’ improvement it would seem that its level of effectiveness has not been increased from the previous generation.

I also have a query over ‘Quad Cache’; as said there are four different caches for the various data types, however wouldn’t this have been done previously to some measure? Caches are usually highly optimised for the type of data they are going to handle so making them specific to a particular task is normal. In some cases, such as with a T&L engine and a pixel cache, the two operations are so far down opposite ends of the pipeline it may not be even physically possible to fit a cache that they both share on the die. I would expect that in at least some measure this would have been happening before, even if not to 4 caches.

Auto-Precharge also seems to be a feature of the RAM being used as well, exactly what is the GeForce4 core doing to enable this to occur?

Onto GeForce4 Ti’s Pixel Shader abilities. There is no mention of the level of support for them; the only new feature that appears to have been added is ‘Z-correct Bump Mapping’ – is this part of the Pixel Shader functionality, and if so to what level of PS functionality does it fit in with DirectX? nVIDIA also makes reference to ‘nearly two times the pixel shader performance of the previous generation’, beyond the hike in clockspeed how is this achieved?

Antialiasing is another area where some technical clarification would be useful. Numerous references were made to ‘no penalty 2x and Quincunx’ modes, some clarification on what is actually occurring here, and how it is achieved, would be beneficial; this could well be linked into the “elimination of Antialiasing steps to improve performance”, which, in itself, could be explained better. If colour combines occur on chip and the front buffer is removed then this would mean that 2x/Qunicunx modes on GeForce 4 would have the same bandwidth requirements as no AA on GeForce3.

I am also unclear as to what mode ‘4XS’ antialiasing actually is. The name could suggest the mystery multisampling/supersampling mode that had been discovered in GeForce3, and the fact that the documentation states there is more texture information would also tend to suggest this. However if this is the case, how is it achieved? Also the documentation states that the mode contains 50% more sample coverage, in which case what is the sample pattern used for this mode?

Although there appears to be no changes to the level of Anisotropic filtering available in GeForce4, they did mention it in the Accuview Antialiasing document. This makes me wonder if they are going to link the level of Anisotropic filtering in with the FSAA selection utility. Currently it is outside of Microsoft WHQL certification rules to supply DirectX controls through the driver panels for the level of Anisotropic filtering; by tying this in with FSAA selection it could be a neat way to circumvent these rules for official certified drivers.

And for technical interest, the documentation does state that GeForce4 Ti consists of 63 Million transistors, but it doesn’t state what silicon process this is achieved on.

Well, there’s a fair bit there to ponder, however on Wednesday 6th of February the European launch of GeForce4 is occurring in Brussels, which I am fortunate enough to be attending. This article will be updated with any pertinent details upon my return and I shall be attempting to clarify as many questions as I can whilst there. Stay tuned for more…