SM3.0, Far Cry and Opengl.

[MrBored]

From what I have read patch 1.1 for Far Cry adds SM3.0 support, but it cannot be used as DirectX 9.0c hasn't been released. Now according to the Far Cry readme there is a beta opengl renderer for the game that can be enabled through an ini file. Surely this would allow the SM3.0 shaders to be used ( as long as the Nvidia drivers allowed it ) and therefore benchmarked?

I would also be really interested in seeing the Opengl performance of the Nvidia cards and ATI cards through this renderer, Deus Ex in Opengl and KOTOR ( although the later might be too CPU limited ). I think the Nvidia cards have better Opengl performance in the older games because they use the old TNL core that Nvidia left in their chips rather than the VS ( I think they spoke about this when the Geforce 3 was launched ). I am pretty much basing this on the large difference in the TNL performance and the VS speed from the 6800 review and the X800 review. The ATI seems to use the VS for all tnl calculations ( hence no difference in speed ) while Nvidia falls back to the old core. Therefore new games that use the VS for all tnl should loose this advantage ( i.e. Doom III ). I may be completely wrong as I'm not sure exacly how vertex shader is used by coders and how much impact this could have, but anyway.

Nvidia Vertex Shader performance
ATI Vertex Shader performance

Just a thought...

Oh, forgot. Hi.

 

[Tim]

I am pretty much basing this on the large difference in the TNL performance and the VS speed from the 6800 review and the X800 review. The ATI seems to use the VS for all tnl calculations ( hence no difference in speed ) while Nvidia falls back to the old core.

Actually the 6800 emulates TnL with the vertex shaders just like X800.

 

[MrBored]

Actually the 6800 emulates TnL with the vertex shaders just like X800.

How does this account for the double in TNL peformance over the VS in the link I provided when there is no such jump for the X800?

 

[PeterAce]

Actually the 6800 emulates TnL with the vertex shaders just like X800.

How does this account for the double in TNL peformance over the VS in the link I provided when there is no such jump for the X800?

Any relevance :

NV40 does not have fixed function vertex pipeline. The FFP results come from the fact that hand-tuned vertex shaders to implement FFP T&L are more efficient than the on-the-fly compiled ones. This shows you some of the growing room left for the vertex program compiler.

 

[MrBored]

Ok thanks.

I'm still interested in seeing some shader orientated opengl games being benchmarked to see if Nvidia still holds the crown there. Benchmarking Quake 3 based games just blows my mind. I used to run Q3 on my PII 333 with a 3DFX Voodoo 3500 just to put into perspective how old that engine is... obviously there have been slight improvements, but come on.