NV40 vs R420 Extreme Pipelines

[Ostsol]

I was under the impression that Displacement mapping required vertex texture capabilities (a mono texture that is used to offset the vertex mesh).

The other way to do it, as was proposed in OpenGL, was basically to render to a vertex buffer, converting framebuffer data into vertex data. With MRTs this is made even better as it allows one to define multiple vertex attributes in a single pass. Unfortunately, the extension that is supposed to allow this has yet to be finalized.

 

[KimB]

Did we ever reach a consensus on whether the "FP32 requirement" for PS3.0 meant that the internal precision must be totally FP32 and that the final output can still have a minimum precision of FP24 or whatever, or is it FP32 everywhere (except with _pp)?

The requirement is for the storage of temporary registers and the processing amounts.

There is no output format at FP24.

 

[radar1200gs]

I'm not really concerned with OpenGL or proprietary extensions here. I'm trying to work out what makes R420's pipelines "extreme" under DX9.

EDIT for Ostol: please read what I wrote: "OpenGL or proprietary extensions".

 

[Mintmaster]

I thought PS2.0b would more directly map to the ATI Rv380/Rv370 or perhaps even 9x00XT series.

I don't think so. There's nothing in ATI's XT product page about increased instruction limits in the pixel shader, and there was nothing about it in Wavey's B3D review.

I think he's referring to the F-buffer idea. I don't know if ATI ever plans on exposing it for the 9800, but the differences between 2.0b and 2.0 could theoretically be explained by the F-buffer - longer instruction limit, no texture instruction limit.

However, I would have expected the dependency limit to be removed, and having more registers doesn't seem very easy for the F-buffer to solve (but may be possible).

Just a thought. Anyway, I don't think R420 will have ps 3.0 support with all the buzz around. But you never know, really. I think during the time of R200 rumours I read somewhere that ATI wouldn't use pixel shaders because it had truform (or some BS like that), so until R420 is actually released, everything is just a rumour.

 

[Ostsol]

I'm not really concerned with OpenGL or proprietary extensions here. I'm trying to work out what makes R420's pipelines "extreme" under DX9.

Well, just for informational purposes, the extension wasn't to be proprietary -- and in any case, ATI never wrote any proprietary extensions --, but was to be an ARB extension.

 

[T2k]

It looks to me like R300 is doing displacement mapping in software, not hardware going by that diagram (if it truly is capable of displacement mapping). nVidia could do that too.

FYI: look for "Pre-Sampled Displacement Mapping"...

 

[radar1200gs]

In other words the work is still done by the CPU, not the GPU.

 

[Dave B(TotalVR)]

Or it could be done by a 'tesselation unit' on the graphics card perhaps?. Either way its certainly not done by the vertex shader.

 

[radar1200gs]

Or it could be done by a 'tesselation unit' on the graphics card perhaps?. Either way its certainly not done by the vertex shader.

Yes, but if there were a tesselation unit present in R300 we would probably know about it by now, and the fact that TruForm is done on the CPU strongly indicates there is no such unit.

http://www.beyond3d.com/previews/nvidia/nv40/index.php?p=6

Vertex Shader:

Vertex Texturing - Vertex Texturing is the capability of supporting texture access from within the vertex shader. Typical usage of this is simulation of physics effects where the physics model is run on the pixel shader and the results are shown on screen by reading back this result and displaying it as geometry. For example, particle effects and cloth animation can be achieved using this principle. Displacement Mapping is also possible and this is especially interesting when combined with higher order surfaces or tessellation.