What was that about Cg *Not* favoring Nvidia Hardware?

[Doomtrooper]

Well, I suppose I wasn't speaking from personal experience, but I still believe it.

You can believe and mislead all you want, doesn't make you correct in your thinking does it. ATI-ISV's name should explain what he does, and it makes common sense that he knows alot more then some Nvidia forum fan on what ATI does with developers.

 

[Sharkfood]

I also know that ATI's drivers are notorious for not always adhering properly to the spec.

How so?

Every single benchmark, game, tech demo, SDK example, code fragment, program and the like written in DirectX to use shaders has worked flawlessly, without change or modification... and in all cases from code base prior to the release of the 8500, designed to run on the GF3.

I don't know where you get your "facts" from, but they sure arent from THIS world (i.e. the "real" one).

 

[MDolenc]

http://www.beyond3d.com/forum/viewt...&postorder=asc&highlight=texm3x3vspec&start=0

 

[Sharkfood]

(scratching head concerning a thread that argues about NVIDIA's involvement with designing the refrast in DX has anything to do with the fact that all shader programs written by professionals work fine, dandy and compliant on ATI hard has anything to do with the topic at hand..)

 

[RussSchultz]

I think the point was that the ATI version of the PS1.1 instruction wasn't doing what the NVIDIA version did. Though it was never clear in that discussion which was doing what the reference rasterizer did.

Assuming that the ATI version wasn't following the reference rasterizer, that would be one example of a shader that didn't work--which is relevant to the discussion at hand.

Of course, we never got to the bottom of that question in that thread...

 

[MDolenc]

Of course that's a crown of everything on this kind of forum that when you ask "what does refrast do?" you get "it doesn't meter what refrast do it was written by NVidia!" . Your "fact that all shader programs written by professionals work fine" puts you in the same basket I guess, since it doesn't meter if a shader was written by a pro or a pure amateur if passes compiler it should still work fine on every hardware capable of running it (and you obviously missed the first post there).

I guess I should also apologize to ATI for sounding to harsh when I first asked Chalnoth about where he got the idea about GeForceFX not supporting ps_1_4 at all or having problems with it, since my claim about developers getting redirected to ps_1_4 when they encounter a problem in ps_1_3 and below is now interpreted as ATI not providing support for ps_1_3 and below shaders.

 

[RussSchultz]

if passes compiler it should still work fine on every hardware capable of running it

I'd rephrase that to "it should still work the same on every hardware..."

 

[KimB]

I guess I should also apologize to ATI for sounding to harsh when I first asked Chalnoth about where he got the idea about GeForceFX not supporting ps_1_4 at all or having problems with it, since my claim about developers getting redirected to ps_1_4 when they encounter a problem in ps_1_3 and below is now interpreted as ATI not providing support for ps_1_3 and below shaders.

Well, you're right. I really shouldn't have interpreted it as that. But developers are still apparently being redirected to PS 1.4, which would seem to mean that PS 1.0-1.3 aren't working correctly on ATI hardware in some cases. Which may force nVidia to also support PS 1.4 for Cg to properly support ATI hardware (as, apparently, have many software developers).

Said another way, it's more of a semantic point than anything, and irrelevant to the argument I was trying to pose in the previous page.

 

[MDolenc]

ps_1_4 is excellent for fallback from ps_2_0 shaders and that's what will be important for developers in coming months and since ps_1_4 target is (currently) missing Cg is kind of incomplete.

edit: Also redirecting isn't that much of a problem since it takes 5 minutes of your time to convert a shader from ps_1_1 to ps_1_4. However GeForceFX not supporting ps_1_4 or having problems with it would be HUGE problem. There are games out there that currently use ps_1_4 shaders and they will use these shaders on GeForceFX (since it will expose ps_2_0 + lots of other caps as maximum) and you would need a patch to fix this.

 

[LeStoffer]

After a lot of good points the question from the end user still remains:

Hey, I just bought this awesome and expensive piece of hardware - but why the hell won't anybody implement the xtras that it supports in my new, favorite game...?

... and I still maintain that Cg could be a very good tool for game developers to fairly easy - this is my point - make a shader update patch to take advantage of whatever new VS version x.x+ / PS x.x+ some IHV comes out with.

This could well be key in the future when the performance of a game will depend a lot more on shader optimizations. While Microsoft's HLSL compiler is probably sufficient today, they will still be the foot that puts the pressure on the metal so to speak. Maybe it'll be too slow for the rest of us?

 

[KimB]

ps_1_4 is excellent for fallback from ps_2_0 shaders and that's what will be important for developers in coming months and since ps_1_4 target is (currently) missing Cg is kind of incomplete.

PS 1.4 will NOT be important for developers in coming months because its installed base is miniscule compared to PS 1.1 and PS 2.0.

 

[Doomtrooper]

Radeon 8500/9000/and upcoming 9100 including mobility is far from being small Chalnoth..these cards are as plentiful as the Geforce 3 and 4 cards...

Great peformance, great IQ and cheap..in fact cheaper then the crappy DX 7 Geforce 4 MX.

 

[andypski]

ps_1_4 is excellent for fallback from ps_2_0 shaders and that's what will be important for developers in coming months and since ps_1_4 target is (currently) missing Cg is kind of incomplete.

PS 1.4 will NOT be important for developers in coming months because its installed base is miniscule compared to PS 1.1 and PS 2.0.

Actually its installed base will always be higher than PS 2.0 since all PS 2.0 cards support it...

 

[KimB]

Actually its installed base will always be higher than PS 2.0 since all PS 2.0 cards support it...

But what's the point of writing 1.4 shaders instead of 2.0 shaders for DX9 hardware? I'd be highly surprised if any new PS 1.4 shaders were done except as fallback for older cards.

 

[KimB]

Radeon 8500/9000/and upcoming 9100 including mobility is far from being small Chalnoth..these cards are as plentiful as the Geforce 3 and 4 cards...

I doubt it, and you have absolutely no data to back that up, either. nVidia still has a significantly larger marketshare.

 

[andypski]

Actually its installed base will always be higher than PS 2.0 since all PS 2.0 cards support it...

But what's the point of writing 1.4 shaders instead of 2.0 shaders for DX9 hardware? I'd be highly surprised if any new PS 1.4 shaders were done except as fallback for older cards.

Possibly because you can actually do quite a lot with 1.4 shaders, they are a natural and easy subset of 2.0 shaders, and they are a realistic length for real-time graphics on current hardware.

You may well find that many of the shaders in upcoming software can be expressed directly with a 1.4 shader in a single pass for these reasons.

Don't get me wrong - you can do a lot more with 2.0 shaders, but I think it'll be a while yet before 64 instruction shaders become the norm.

HLSL may improve the uptake and usage of 1.4 shaders since you can compile at run time to appropriate models. For each shader you might as well compile to the lowest shader version that is required to get the maximum use out of varied hardware.

 

[Dave Baumann]

I doubt it, and you have absolutely no data to back that up, either. nVidia still has a significantly larger marketshare.

Marketshare with what though?

ATI's data puts the low end, i.e. 9000, at roughly 4 times the sales volume of the performance mainstream ($200 board) which in turn is 8x greater than high end ($300) - meaning that a chip such as 9000 will have about 32 times the volume of high end parts. Presumably NVIDIA will experience similar volumes in the market categories, but they do not have any shader capable parts in the low end.

Given the design wins 9000 and Mobility 9000 has had, even ignoring 8500 before that, there is obviously going to be a large target insall base for ps1.4 reasonably soon.

 

[Doomtrooper]

Radeon 8500/9000/and upcoming 9100 including mobility is far from being small Chalnoth..these cards are as plentiful as the Geforce 3 and 4 cards...

I doubt it, and you have absolutely no data to back that up, either. nVidia still has a significantly larger marketshare.

Hmmmm...

ATI
Tyan
Hercules
Gigabyte
Powercolor
Club 3D
Crucial
VisionTek
Elsa
Unitech
Saphire
Creative

All making 8500,9000 based solutions and you think Nvidia should be the only one supported with PS 1.1

Maybe this graph will help sort you out...

http://www.de.tomshardware.com/graphic/02q4/021025/index.html

 

[KimB]

Show me the market report that more Radeon 8500's (and up) have been sold than GeForce3's and up. That's the only thing that will convince me.

Plain and simple, nVidia has more marketshare. While it is true that much of that is still with TNT2's, I'm willing to bet that most of ATI's is in their Rage series.

 

[demalion]

Your posts seem awfully slippery to me, Chalnoth.

You said:

PS 1.4 will NOT be important for developers in coming months because its installed base is miniscule compared to PS 1.1 and PS 2.0.

The responses you received attempt to refute that. But then you go and demand, as if someone made such a comment about ATI hardware,:

Show me the market report that more Radeon 8500's (and up) have been sold than GeForce3's and up. That's the only thing that will convince me.

No one stated any such thing as far as I can tell. Nor should it be required to be stated to address that first statement.

Further, you try to restate your positions as:

Plain and simple, nVidia has more marketshare. While it is true that much of that is still with TNT2's, I'm willing to bet that most of ATI's is in their Rage series.

As if that was what was being addressed when people responded to you initially. Back to the PS 1.4 issue, do you have a response to the comments that have been made?