my 9800 article

[Reverend]

Dave, what do you mean by "fixed" in the R350? [edit]And why specifically the R350 and not also the R300? Has N-patch support changed between the two?

[edit]And what exactly are you looking for in a N-patch test?

 

[LeStoffer]

Is there an N-Patch test anywhere? I keep meaning to see if that was 'fixed' in R350 or not.

Dunno, but if you're talking about Truform, ATI have said more than once now that they don't consider that interesting with games now increasing the polygon count greatly (or using normal maps). So I take that as Truform (as we knew it with the R8500) for being dead.

 

[Dave Baumann]

My personal feeling for this shift in 'focus' away from N-Patches was probably likely down to it just being a bit buggered in R300 and wasn't really high priority to go back and fix, thus delaying the chip. If this is the case then it may well have been fixed with R350 (which is basically an optimised / bug fixed R300). I was just wondering if there are any useful test apps to test the throughput.

 

[LeStoffer]

I understand Dave, but I tru[e]ly think that Truform is dead = no commitment from ATI to 'fix' anything like that on R350.

I just found this interview (there have been others):

http://www.driverheaven.net/display.php?page=ben_interview

[Zardon] What type of performance increase can we expect at most from the (semi?) software driver N-patch/PN-triangles implementation on R300? (And since TRUFORM performance is poor on the R300 how will this effect Displacement Mapping).

[Ben] While we are always working on ways to increase the performance of our products through our driver software, this area is considered a relatively low priority. TRUFORM technology was designed to enhance 3D applications with low polygon counts (optimized for older graphics hardware) so they could take advantage of the massive vertex processing capabilities of newer hardware. Recent and upcoming 3D games, which are the focus of our performance optimisation efforts, are being designed with much higher polygon counts, so they tend to benefit less from TRUFORM.

Sounds like a dead end to me.

 

[Luminescent]

Tommti-systems.com has an N-patch test in its archives/download section. Ah, yes, here it is.

 

[DeanoC]

I understand Dave, but I tru[e]ly think that Truform is dead = no commitment from ATI to 'fix' anything like that on R350.

I just found this interview (there have been others):

http://www.driverheaven.net/display.php?page=ben_interview

[Zardon] What type of performance increase can we expect at most from the (semi?) software driver N-patch/PN-triangles implementation on R300? (And since TRUFORM performance is poor on the R300 how will this effect Displacement Mapping).

[Ben] While we are always working on ways to increase the performance of our products through our driver software, this area is considered a relatively low priority. TRUFORM technology was designed to enhance 3D applications with low polygon counts (optimized for older graphics hardware) so they could take advantage of the massive vertex processing capabilities of newer hardware. Recent and upcoming 3D games, which are the focus of our performance optimisation efforts, are being designed with much higher polygon counts, so they tend to benefit less from TRUFORM.

Sounds like a dead end to me.

Actually as the basis of displacement mapping under Dx9 its far from going away, more changing faces. The original magic make thing look better may be dropped but the tesselation properties are likely to become even more important.

 

[LeStoffer]

Actually as the basis of displacement mapping under Dx9 its far from going away, more changing faces. The original magic make thing look better may be dropped but the tesselation properties are likely to become even more important.

Yes, absolutely: Note that I said that the R8500 Truform as we knew it seems dead. Tesselation used for displacement mapping (or HOS) is hopefully key in their present & up-coming hardware (namely VS 3.0).

OT: Deano C, do you know anything about the new nVidia drivers in regard to the FP16/FP32 ordeal?

 

[DeanoC]

Actually as the basis of displacement mapping under Dx9 its far from going away, more changing faces. The original magic make thing look better may be dropped but the tesselation properties are likely to become even more important.

Yes, absolutely: Note that I said that the R8500 Truform as we knew it seems dead. Tesselation used for displacement mapping (or HOS) is hopefully key in their present & up-coming hardware (namely VS 3.0).

OT: Deano C, do you know anything about the new nVidia drivers in regard to the FP16/FP32 ordeal?

I know nothing regarding the new drivers (DaveB asked me a similar thing in a PM). What I do know is that FP24+ is the default precision and NVIDIA are updating there drivers to the 'new' spec. Wether the latest drivers are using the 'old' or 'new' spec I don't know yet.

 

[demalion]

Rev: my basic question was concerning displacement mapping...what implementation of it does the R300/R350 have that is causing ATI to list it as something they support, and the GF FX does not. That evaluation hasn't been supported anywhere (to my knowledge), and my talk about enabled/disabled in drivers was my theory as to how it could maybe be justified.

As for Truform/N-patches: The option to use it as an image quality enhancing feature with minimal speed loss was removed (i.e., scalable geometry compression for certain models). This is something it can successfully achieve on the 8500, and cannot on the newer cards due the severe performance penalty for implementing it in software (this is what needs to be "fixed"). It is most definitely not "worthless" (on an 8500).

 

[Luminescent]

Agreed, damalion.

 

[R300King!]

Althorin that was comedy gold. LOL

haha.. Phil Rocks doesn't he? 8) [/b]

 

[ben6]

Reverend, I had such a disclaimer in the article as originally written. In fact, I also discussed the underpowered system and some of the other issues I had in writing the article . But ah well .