What annoys me with Nvidia and Ati video cards, reg option

[K.I.L.E.R]

that allows you to enable tiling.

Here is a quote from Radeonator's help file:
Tiling

Different forms of tiling. It enables different forms of tiling, tiling is used to render the screen in parts (tiles) which saves bandwith (Kyro cards are totally based on tile-based rendering). So enable all forms of tiling.

Also there is a reg option with Rivatuner on my previous Geforce 3 for tiling as well.

Now from what I understand is that my brothers Geforce 3 (was mine) and my Radeon 9700 Pro are not DRs but are IMRs. I understand you do not need to have a video card that does DR to support tiling.
What I would like to know is if the option works and what benefits I will recieve on my IMR card?

Thanks

Tried it and these are my results:
OpenGL
2003-01-01 17:11:46 - war3 (no tiling)
Frames: 2308 - Time: 47599ms - Avg: 48.488 - Min: 28 - Max: 76

2003-01-01 17:14:11 - war3 (tiling)
Frames: 2331 - Time: 48820ms - Avg: 47.746 - Min: 29 - Max: 73

Direct3D
2003-01-01 17:16:27 - war3 (no tiling)
Frames: 2935 - Time: 61919ms - Avg: 47.400 - Min: 29 - Max: 62

2003-01-01 17:18:44 - war3 (tiling)
Frames: 1991 - Time: 40798ms - Avg: 48.801 - Min: 31 - Max: 71

(don't worry this isn't the framerate I play with normally)

 

[Tagrineth]

Tiling really just means the frame buffer is arranged in an array of tiles rather than one big 'chunk'.

It means slightly more efficient memory access and organisation, in theory.

Look at both D3D and OGL - the minimum frame rate is increased in both.

Interesting though that OGL's max frame rate went down... looks like some kind of driver bug?

 

[OpenGL guy]

Interesting though that OGL's max frame rate went down... looks like some kind of driver bug?

The OpenGL results look identical to me. I.e. the difference in the numbers is just noise.

 

[ogldriverman]

The word "tiling" in graphics rendering has traditionally meant that rendering is done in blocks or "QUADs" as opposed to scan-line rendering. It is an oft misused/ambiguous term.

Tiling also refers to memory arrangement. Textures, Z, display, or any graphics memory is (or can be) tiled, i.e. arranged so the QUAD access is sequential instead of sparse.

Modern GPU memory controllers allow different levels of tiling. MacroTiling for display memory to avoid DRAM page crossings (relatively expensive). MicroTiling for textures to linearize QUAD rendering (not QUAD polys but 4 pixels).

The Neon 3D single chip processor from DEC implemented one the first TILED renderers, and that means it rendered triangles in blocks to avoid page crossings.

But, there is also full scene deferred tiled rendering or CHUNKING, and this is what people typically mean when they say tiled rendering (they should really say a CHUNKING architecture). That is, a scene is rendered a section at a time (all polys in a tile or a chunk) to allow rendering to on chip memory (ala Kryo).

- DRIVERMAN