Shader Models 2 and 4

[Reverend]

The assumptions : That Xbox2, PS3 and Revolution will all be SM3. That ATI will very soon join NVIDIA with SM3.

Question 1 :
Where does that leave SM2 once both (the major consoles plus ATI) are available? We're talking software (=games) support.

Question 2 :
Where will that leave SM4 once both (the major consoles plus ATI) are available, and further assuming SM4 (being still a WIP) becomes a reality (whether with WGF, or otherwise)? Again, we're talking software (=games) support.

I only ask because there have been a whole lot of ATI SM2.0 video cards sold.

Bear in mind the subject of "timeframe".

 

[digitalwanderer]

Q1: As a big base of 2.0 users to pander to. Also I think a lot will depend on how/if 3.0 proves itself advantageous or not.

Q2: Isn't it a bit too early to worry about SM4? Or if not, same as above...it depends on what new features/performance it will bring to the party.

 

[Geo]

Q1: As a big base of 2.0 users to pander to. Also I think a lot will depend on how/if 3.0 proves itself advantageous or not.

Q2: Isn't it a bit too early to worry about SM4? Or if not, same as above...it depends on what new features/performance it will bring to the party.

I think he's after in the long run (i.e. after 4.0 release) does sm2.0 or sm3.0 "win" or is it a draw. Will there be three paths 4.0, 3.0, 2.0, or two --4.0 and (which)?

Thing I can't remember is if WGF 1.0 is sm2 or sm3 --I just remember it being "dx9", which isn't quite granular enuf. I would think that would have some impact too.

 

[Karma Police]

Why worry about SM2.0(b)?

Isn't that the same as worrying about DX7-8 cards when games w/ DX9 come out? All the high-end games (Doom3, Farcry) can still be played with a Ti4600, it just makes the programmers to program for the lower level shaders.

Or am I missing something?

 

[ANova]

I don't see why support for past shader models would be any different when WGF 2 comes out then what has happened in the past and is happening.

It's really up to the developers. Obviously everyone will support SM4; as for the rest, i'm sure most developers would want to support SM3 over SM2 for the greater ease of programmability. Ultimately I think it will come down to how large the SM2 market is as time moves on.

 

[Ostsol]

What is SM4 supposed to introduce, anyway?

(BTW, congratulations on reaching an evil post number, ANova. )

 

[Farid]

Isn't what High Level Shading language are meant for? Dealing with different layers of Shader Model.

 

[DegustatoR]

A1: SM3 will be used more often. SM2 path will probably be made for PC versions of any NG-consoles/PC multiplatform titles. This path will probably have some quality difference comparing to SM3 path or will just be slower than SM3 path.

A2: Same place it leaved PS1.4 and SM2 in the beginning of their life. Up until today we don't have that many SM2 titles released and most games are PS2/Xbox/PC multiplatform with SM1.x effects.

In general i wouldn't worry for SM2 users, they'll probably get their SM2 paths in most of upcoming SM3 games.

The question about SM4 is more interesting. I believe that NV/ATI should concentrate on improving SM3 performance even in their future SM4 parts because it seems that SM4 won't be that important -- at least in it's first generation.

 

[Dave Baumann]

I only ask because there have been a whole lot of ATI SM2.0 video cards sold.

There probably already have been, and will continue to be, a whole hell of a lot more Intel PS2.0 graphics chips sold thanks to Longhorn's reliance on WGF1.0 (DX9L - SM2.0). Given Intel's reluctance to move vertex processing from the CPU it will be very curious to see what their graphics strategy will be from here (especialy since I've heard they "hate" the idea of Vertex texturing).

 

[Xmas]

What is SM4 supposed to introduce, anyway?

"Unlimited" resources, geometry shader, some IEEE754 compliance, integer operations

Isn't what High Level Shading language are meant for? Dealing with different layers of Shader Model.

Theoretically, yes. But SM2.0 is somewhat limited and it's not that hard to write shaders that won't compile to ps_2_0. And "SM4.0" introduces a new pipeline model including a geometry shader, as well as some other new stuff.

 

[Geo]

I only ask because there have been a whole lot of ATI SM2.0 video cards sold.

There probably already have been, and will continue to be, a whole hell of a lot more Intel PS2.0 graphics chips sold thanks to Longhorn's reliance on WGF1.0 (DX9L - SM2.0).

Ah ha! That's what I needed. So three paths it is for quite a long time, as I'd think the console ports would guarantee SM3 sticks around in volume --why make the effort to pull it out if you already have it?

How hard would it be to write an app that would automate SM3 --> SM2 conversion?

 

[Dave Baumann]

Its very much nota give that console titles that make use of the graphics to the full extent will fit into a PC Shader Model 3.0 without conversion.

 

[Geo]

Its very much nota give that console titles that make use of the graphics to the full extent will fit into a PC Shader Model 3.0 without conversion.

But it would be "a little bit pregnant" situation, wouldn't it?

Oh wait, maybe I'm going the wrong way. . you're talking about proprietary extensions?

And I'll ask again the question I edit/added upstream: How hard would it be to write an app that would automate sm3 ==> sm2 conversions? Aren't the on-the-fly shader recompilers in the current drivers already a start in that direction?

 

[Jawed]

There probably already have been, and will continue to be, a whole hell of a lot more Intel PS2.0 graphics chips sold thanks to Longhorn's reliance on WGF1.0 (DX9L - SM2.0).

Ah ha! That's what I needed. So three paths it is for quite a long time...

You wouldn't want to game with Intel's SM2.0. So Intel's SM2 capability actually translates into SM1.1 for games. Intel's SM2 seems relevant only to the Longhorn UI parts of WGF.

Well, that's I how read it.

Jawed

 

[Demirug]

And I'll ask again the question I edit/added upstream: How hard would it be to write an app that would automate sm3 ==> sm2 conversions? Aren't the on-the-fly shader recompilers in the current drivers already a start in that direction?

The shadercompiler inside the drivers primary do instruction reordering to make the shader fit better to the ALU/TMU design.

To answer your question. Yes it is possible but there are many problems and the results of such a conversion will not allways make you very happy.

 

[Demirug]

There probably already have been, and will continue to be, a whole hell of a lot more Intel PS2.0 graphics chips sold thanks to Longhorn's reliance on WGF1.0 (DX9L - SM2.0).

Ah ha! That's what I needed. So three paths it is for quite a long time...

You wouldn't want to game with Intel's SM2.0. So Intel's SM2 capability actually translates into SM1.1 for games. Intel's SM2 seems relevant only to the Longhorn UI parts of WGF.

Well, that's I how read it.

Jawed

For a publisher even people with intel video solutions are possible customers. Because of this and the high number of sold intel video solutions a developer have to take care.

But as you already noticed each 2.0 hardware is a 1.X hardware too.

 

[Geo]

There probably already have been, and will continue to be, a whole hell of a lot more Intel PS2.0 graphics chips sold thanks to Longhorn's reliance on WGF1.0 (DX9L - SM2.0).

Ah ha! That's what I needed. So three paths it is for quite a long time...

You wouldn't want to game with Intel's SM2.0. So Intel's SM2 capability actually translates into SM1.1 for games. Intel's SM2 seems relevant only to the Longhorn UI parts of WGF.

Well, that's I how read it.

Jawed

Yes, but a baseline is a baseline. And not only Intel will make baseline cards/igps to meet the baseline both for desktop and laptop. Whatever level was picked for that would guarantee a path is available for it in pretty much all games for quite a long time. At least that's my read.

After noodling over Dave's observation that "it is not a given" that console titles won't need serious converting, I still think those sm3 paths will survive largely intact. Developers have their own interests too, for one reason, in making ports as painless as possible.

Thanks, Demirug --I do know that. I was just suggesting that kind of automated analysis of shaders is a baby proto-step towards automated conversion of shaders. . .if they want it to be.

 

[London Geezer]

One thing is for sure, in the next few months there will be consoles developers will work on, with full SM3 support, and the engines they will built for those consoles will be SM3. That pretty much means there will be A LOT of SM3 games developed for consoles, and i think the PC market will benefit from this in that there will also be a lot of PC games built around SM3, being conversions of console games. The lowest common denominator will be SM3.
That's a much wider acceptance than SM2 ever got or ever will.
I think there is a very strong possibility that PC games will finally make the leap they need once all the gaming community will move to SM3. It can only be good news.

 

[oddfellow]

I would say there will probably be 3 shader levels.

SM-4, SM-3 and SM-2

I see SM-2b as a bit of a dead end developer/game wise. I think it'll go the way of SM-1.4, in that most developers forgot about it and just coded for SM-1.1

 

[jvd]

THe way i see it going down is like this .

sm 4.0 the next base line after all the consoles are out the majority of games iwill be ported to it

Sm3.0 is a dead end . Consoles will develop for thier feature set which will be more advanced than just sm3.0 but not as advanced as sm 4.0 .


Sm2.0 will get conversions jsut like sm3.0 There are still more cards out there than sm3.0

p.s 2.0b well this is a bigger dead end than sm3.0 most likely few games will take advantage of it and it will mostly be only in big engines .

dx 8.1 and dx 8 will be phased out during 2006 and we will only see games based off the older engines use this