NV30 white paper

[IronH]

white paper document from Nvidia outlining some NV30 technology
http://developer.nvidia.com/docs/IO/3121/ATT/cinefx_whitepaper.pdf

Vertex Shaders

R300:
Max Instructions: Up to 1024 instructions with flow control.
Max Constants: 256
CineFX:
Max Instructions: Up to 65536 instructions with flow control.
Max Constants: 256

Pixel Shaders
R300: Up to 160 instructions with 128-bit F.P. precision.
CineFX: Up to 1024 instructions with 128-bit F.P. precision.

Color Precision
R300: Supports 128-bit floating point. Not sure if ATI supports 64-bit as well.
CineFX: Supports 64 and 128-bit floating point.

[Grrr!!]

Looks good. *drool*. Of course games using these specs wont be out for a while.

[I.R]

Forget the games...... what monitors are gonna be able to show you all those vertex pixel 128bit color thingies :eek: (at a reasonable price :eek: )

[Grrr!!]

Shouldn't all modern CRT and LCD screens be able to show anything produced by the graphics card, as it is given pixel by pixel.

[Volodkovich]

Stop comparing next gen cards, with cards that are comming out even later, theres no point. Thats like comparing a GF2, to a GF4. Also, ATi are releasing a updated version of the r300 when the NV30 comes out :)that will be on 0.13 micron process "vpu", and DDR-II.

[Grrr!!]

I agree with Volodkovich, but this stuff does have a certain drool factor :D.

[KingJackal]

Originally posted by I.R
Forget the games...... what monitors are gonna be able to show you all those vertex pixel 128bit color thingies :eek: (at a reasonable price :eek: )

The 128-bit Floating Point precision is only used internally. The DAC's on these cards ( R300 being 128bit internal as well ) will down-sample to 32-bit colour before output. The purpose of the 128-bit internal precision is purely the elimination of rounding errors.

I R so smart :p