Let's see the 3D Mark 2003 IQ issues differently, shall we?

[ChrisRay]

Originally posted by StealthHawk
can you explain this? why does the R200 path "have to" use FP16 for the R300?

Actually I believe its 16 bit fixed, Because its using r200 pixel shader functions, I'm not sure how backward compatible the R300 is, Meaning if isd Pixel Shader 1.4 is seperate or if the Pixel Shader 2.0 does the 1.4 Shaders like the Nvidia card,

So therefore I can only speculate.

Originally posted by ChrisRay
Actually I believe its 16 bit fixed, Because its using r200 pixel shader functions, I'm not sure how backward compatible the R300 is, Meaning if isd Pixel Shader 1.4 is seperate or if the Pixel Shader 2.0 does the 1.4 Shaders like the Nvidia card,

So therefore I can only speculate.

yeah, you're right about that, I think. I have heard that R300 supports 16 bit fixed, which I take is INT16. but not FP16.

[Uttar]

AFAIK, the R300 doesn't support INT16. Could be wrong, though.
The R300 is all about massive FP24 performance.


Uttar

[Steppy]

Originally posted by ChrisRay
Since aparently its just 24 bit downsampling to 16 bit (which I think is retarded for any given number of reasons)
Either or, I think ATIS implementation of its floating point precision kinda leaves a little bit to be desired. Expecially when you consider DirectX 9.0 current specification. As ATis card is just a bare minimum for DX 9.0 I'm not quite sure they chose to stick with strict 24 bit precision. DX 9.0 specifications be damned. Probably to save Die space on their already crazily overloaded 0.15 micron proccess.

From a programmers point of view, They leave little room for modification or tweaking, And thats always a bad thing, I can see why John Carmack Stated he has become limited by the r300 programmability. Kinda disapointing to me. Oh well.

You're making some faulty statement here.

1. Both the 32-bit FP(would be 128-bit color as its 32-bit PER CHANNEL) and 24-bit FP( results are downsampled when they are rendered. The FP people are talking about is for color calculations NOT output. It's much the same as the old 3dfx 22-bit color thing...it rendered higher quality 16-bit because all of its calculations were done in 32-bit.

2. It's been shown time and time again that going above DX specs yields nothing as those feature never get supported in games using that API. They may get used in OGL, but few use them.

3. Carnack's comments had NOTHING to do with ATI being 24-bit as opposed to 32...it was about the instruction count being used in situations he was screwing around with NOT game situations, and this has already been addressed in R350 which surpasses the FX's dx9"+"specs.

4. ATI's card being "minimum DX9" is pretty irrelevant as we're probably at LEAST a year a way from games making ANY use of DX9, and TWO years away from games using DX9 as much as DX8 is used now.

Originally posted by Uttar
AFAIK, the R300 doesn't support INT16. Could be wrong, though.
The R300 is all about massive FP24 performance.


Uttar

B3D said it could use INT16 when they did their technical comparison of a gfFX vs r9700. some of that info about the gfFX is probably wrong, as more things became known later, but the r9700 info should be correct.

[AGP64]

Originally posted by Chalnoth
I think the problem is that DirectX offers no way to expose FX12 functionality (12-bit integer).

Since the FX can execute FX12 and floating-point ops in serial, this is a major problem for the performance of the FX.

I would say the other way around. Both Ati and Nvidia knew Directx 9.0 specs when they went off and designed their R300 and NV30. It just seems that the one made some smarter decisions than the other.

[Chalnoth]

Originally posted by AGP64
I would say the other way around. Both Ati and Nvidia knew Directx 9.0 specs when they went off and designed their R300 and NV30. It just seems that the one made some smarter decisions than the other.

Um, no. Both chips were in development for at least two years before the release of DirectX 9. Do you really believe that the specs of DX9 were finalized two years before its release?

The features and specs of DX9 are there because of the hardware that nVidia and ATI developed, not the other way around.

[Chalnoth]

Originally posted by Uttar
AFAIK, the R300 doesn't support INT16. Could be wrong, though.
The R300 is all about massive FP24 performance.

Uttar

The R300 certainly doesn't appear to have any separate execution units, but the FP24 units could be used to emulate INT16 quite easily (since it does have 1-bit sign and 15-bit mantissa).

[RobHague]

and this has already been addressed in R350 which surpasses the FX's dx9"+"specs.

No, the R350 is an optimized R300 core. It has added programabiliy in some area's however it does not surpass the FX for programability. Nvidia have that at least - if you take a look at the comparison charts around the net for the R300/350 against DX9 against the FX then the FX clearley has more programability. But then if it never gets used its not going to matter is it. Im sure it was Carmack that said he was going to be basing his future projects on the things the NV30 made possible.

*edit*

Found the quote on this...

Developers on GeForce FX - John Carmack (Id Software)
NVIDIA Is the first of the consumer graphics chip companies to firmly understand what is going to be happening with the convergence of consumer realtime and professional offline rendering. The architectural decision in the NV30 to allow full floating point precision all the way to the framebuffer and texture fetch, instead of just in internal paths, is a good example of far sighted planning. It has been obvious to me for some time how things are going to come together, but Nvidia has made moves on both the technical and company strategic fronts that are going to accelerate the timetable over my original estimations.

My current work on Doom is designed around what was made possible on the original GeForce, and reaches an optimal implementation on the NV30. My next generation of work will be designed around what is made possible on the NV30.

[Captain Beige]

Originally posted by RobHague
No, the R350 is an optimized R300 core. It has added programabiliy in some area's however it does not surpass the FX for programability. Nvidia have that at least - if you take a look at the comparison charts around the net for the R300/350 against DX9 against the FX then the FX clearley has more programability. But then if it never gets used its not going to matter is it. Im sure it was Carmack that said he was going to be basing his future projects on the things the NV30 made possible.

NV30's "extra" programmability is meaningless due to hugely pathetic performance when using it

[RobHague]

lol in your completley un-bias opinion. If carmack thinks its worthwhile then thats more than enough credentials for me. I look forward to doom3 very much

I think the "fx perfromance is rubbish" thing is kinda old now, stop floging a dead horse lol.

[Uttar]

Originally posted by RobHague
lol in your completley un-bias opinion. If carmack thinks its worthwhile then thats more than enough credentials for me. I look forward to doom3 very much

I think the "fx perfromance is rubbish" thing is kinda old now, stop floging a dead horse lol.

WTF are you talking about?!
Carmack didn't say that the extra programmability is worthwhile for the customers! In fact, he even said he won't really use it before his next engine!

And the NV30's extra programmability on the Vertex Shader front could be an advantage in real-time scenarios. But the extra programmability in Pixel Shaders *can't* - it's pretty much death at 60 instructions already! Maybe less, not sure, been a long time since I seen those numbers.

The NV30 PS performance *is* pathetic, want it or not. That's the main reason while I have, a few months ago, gone from "The NV30 isn't all that great, but it's still quite good." to "The NV30 isn't good."


Uttar