How many sm3 games were compared?

[russo121]

I' ve read a lot of articles saying x1800 is better in sm3 compared to g70. The only game I actually think that uses sm3 is Splinter Cell - Chaos Theory - in 17 different games http://www.xbitlabs.com/articles/video/display/radeon-x1800.html - even Fear is SM2.
Well, one game is not enough to prove that. I would like also that reviewers write some information regarding which sm was used, or opengl.
Maybe I'm wrong in that list, but I would like to see more sm3 games compared just to arrive to any conclusion.

Edit : Far Cry was tested using sm3 - "...despite of the fact that both ATI and NVIDIA hardware works using Shader Model 3.0 code-path, NVIDIA’s products manage to grab the lead away from ATI..."

 

[Mintmaster]

For FarCry, scroll down a bit:

The eye-candy mode changes the situation by putting the RADEON X1800 XT impressively ahead of the GeForce 7800 GTX due to tremendous memory clock-speeds. The model X1800 XL cannot beat the model 7800 GT, even despite of allegedly more efficient memory controller.

Although I don't know why the put the last sentence there. The X800XL beats the 7800GT at 1280x1024 and 1600x1200 by a bit. It just loses at 1024x768.

You can look at Tech-Report's SM3 FarCry numbers as well:
http://techreport.com/reviews/2005q4/radeon-x1000/index.x?pg=8

But yeah, we'll probably have to wait until Unreal Engine 3 for many more PS3.0 titles. Given that PS3 is making that engine popular, I'm expecting quite a few next year.

 

[Skrying]

From what I've read Unreal Engine 3 is really only a SM2.0 engine using some SM3.0 to increase peroformance.

Really, I cant think of any games that really use SM3.0 for much. The SP:CT patch showed that indeed everything that could be done in the SM3.0 path could be done on a SM2.0, and dont forget, HDR is NOT an SM3.0 feature.

 

[John Reynolds]

From what I've read Unreal Engine 3 is really only a SM2.0 engine using some SM3.0 to increase peroformance.

Same goes for Oblivion according to its devs.

 

[Mintmaster]

Don't get me wrong, I'm no PS3.0 cheerleader. I think it really has pretty much no real benefit over PS 2.0 with the exception of dynamic branching, and with G70 offering very limited advantage with its use, I never bought the whole PS3.0 advantage for NV since 2004, especially since the stencil test was generally much more effective, like in Humus' demo.

I just remember Epic guys talking about using branching for speeding up soft shadows, just like in shadermark and ATI's upcoming demos and SDK samples. Carmack is also talking about using shadowmaps for a while, so it's likely he'll also use branching for speeding up shadows.

Funny that this time ATI will have the shadow advantage. It may be too late, though.

 

[Skrying]

Unreal 3 Engine speed will be very important, IMO. I've always seen Unreal engine games, I'm guessing because from what I've heard its just way more easier to use than any other major engine.

Cant wait till the next Unreal game comes or games using the engine start appearing.

 

[jvd]

Sm 3 is nice don't get me wrong but we have


p.s 1.1 , 1.3 , sm 2.0 , 2.0a , 2.0b and now sm 3.

I would say the largest base of cards is p.s 1.1 and sm 2.0 cards . So of course most of the engines for the next 2 or so years will be programed for that using sm3.0 for a speed increase. I don't see sm3.0 as anything more than a stop gap till sm 4.0 or its sucesser comes out.

 

[wireframe]

I don't see sm3.0 as anything more than a stop gap till sm 4.0 or its sucesser comes out.

Hopefully SM 4.0 will be a good stop-gap until SM 5.0 is published.

 

[jvd]

Hopefully SM 4.0 will be a good stop-gap until SM 5.0 is published.

I'm thinking sm 4.0 will be around for a long time . Perhaps 2-3 years and will be the defacto standard for dx 10. These little baby steps are very annoying and it splinters the market so much .

 

[wireframe]

I'm thinking sm 4.0 will be around for a long time . Perhaps 2-3 years and will be the defacto standard for dx 10.

Well, I am pretty sure SM 2.0 and SM 3.0 will be around for 2-3 years as well. While I think that SM 4.0 might be good in that it will consolidate GPU programming, I have my doubts thats it will be around longer or be more popular from that fact alone. Just because the shaders will be unified doesn't mean that features will not be added and the model incremented.

Ultimately, I see this as very unimportant because what will matter is performance. Assuming your 2-3 year lifespan, little doubt can be had that a part released at the beginning of that cycle can perform the things one released at the end will, shader models aside.

These little baby steps are very annoying and it splinters the market so much .

They also help generate revenue for the industry. Nothing sells like an incremented number. As a consumer I agree that this is annoying. This is one advantage closed systems like gaming consoles enjoy. It is also very evident how the software created for those platforms really shine whereas the PC counterparts often come through as experiments in rendering rather than games.

 

[Reverend]

I think it really has pretty much no real benefit over PS 2.0 with the exception of dynamic branching,

Yup, that and 32-bit floating point as minimum.

 

[jvd]

Well, I am pretty sure SM 2.0 and SM 3.0 will be around for 2-3 years as well. While I think that SM 4.0 might be good in that it will consolidate GPU programming, I have my doubts thats it will be around longer or be more popular from that fact alone. Just because the shaders will be unified doesn't mean that features will not be added and the model incremented.

I meant as the highest sm version out . HOpefully they make one shader model for dx 10 and thats it .

Ultimately, I see this as very unimportant because what will matter is performance. Assuming your 2-3 year lifespan, little doubt can be had that a part released at the beginning of that cycle can perform the things one released at the end will, shader models aside.

Well that is normaly true anyway . The exception is relaly the r300 series and sm2.0 . But the geforce 3 was quickly droped and replaced with the geforce 4s for decent sm1.1 performance. The nv30s were dogs at sm2.0 (As was the rest of the nv3x line) even the geforce 1 sdr was poor at dx 7 .

But if ms was to introduce sm4.0 developers could start developing for it right away. It would still take 1-2 years for games to take advantage of it and early adopters of the first cards capable of this are already used to having to upgrade to actually use those features or lower thier res to take advantage of it .

However for the consumer you don't have to worry about buying a sm4.0 part and then having sm4.0a coming out and then 6 months later sm4.0b and then a year after that sm4.5 or whatever .

Gamers will still update as games will take more advantage of the shaders and the hardware will need to be replaced.

They also help generate revenue for the industry. Nothing sells like an incremented number. As a consumer I agree that this is annoying. This is one advantage closed systems like gaming consoles enjoy. It is also very evident how the software created for those platforms really shine whereas the PC counterparts often come through as experiments in rendering rather than games.

I think it pisses people off more tha nanything . There are way to many shader models and the lead in time from buying the card to getting to see the feature sused are way to long and by the tiem they do come around your most likely going to have to update anyway . If the industry starts making goals to hit as a whole i think things will be better . Instead of sm2.0 , 2.0a , 2.0b there should have just been sm2.0 .

I think users will be more willing to upgrade when they see that thier old cards were taken advantage of . If the average consumer could go into the store and buy a x1800xt the day its launched and pickup in the months around its launch games that take advantage of its hardware i think the industry itself would be in better shape . As it is the sm3.0 hardware has been on the shelves for over a year now and we have barely any games that use these features and many that do use it for speed . Not for visual quality and many wont notice that anyway when they buy a new card to play on . They are buying a faster card to begin with .

 

[Demirug]

Sm 3 is nice don't get me wrong but we have


p.s 1.1 , 1.3 , sm 2.0 , 2.0a , 2.0b and now sm 3.

I would say the largest base of cards is p.s 1.1 and sm 2.0 cards . So of course most of the engines for the next 2 or so years will be programed for that using sm3.0 for a speed increase. I don't see sm3.0 as anything more than a stop gap till sm 4.0 or its sucesser comes out.

SM3.0 is the final destination for Windows XP. It does not matter how good your GFX will be you will never see more than SM3 for Windows XP.

This can be a showstoper for SM4 because as long as you need to suport Windows XP and older hardware (pre SM4) you have to use D3D9. If you want to use SM4 too you have to write all your graphic stuff a second time for D3D10.

SM5 is very unlikely at the moment as SM4 say that they driver have to work with HLSL and every shader with valid syntax have to run.

 

[Reverend]

I see some of you talking about SM4 and commenting about it.

What would SM4 be?

 

[Demirug]

I see some of you talking about SM4 and commenting about it.

What would SM4 be?

The new Shadermodel for D3D10 aka WGF2.0. If anything goes right we will see it in the Beta2 of Windows Vista.

 

[Reverend]

I don't see sm3.0 as anything more than a stop gap till sm 4.0 or its sucesser comes out.

SM3 is important. Ask multi-platform-target developers why I say so.

I know this forum is more or less PC-centric but with consoles making more money for developers than the PC, it really wouldn't matter how good the next SM is supposed to be if we have the XB360 and PS3 basically being SM3-based. And coming out end of this year and foreseen to be *the* consoles at least 1.5 years after Vista. Note the relationship between PC and console games for the same title(s) and how the PC folks complain about "lack of progress". Royalties rule.

Or can you back up what you just said?

 

[Reverend]

The new Shadermodel for D3D10 aka WGF2.0. If anything goes right we will see it in the Beta2 of Windows Vista.

Ohhh, thaaat. Gee, I must have been watching too much of "Ice Age" with my son.

I ask again, what would SM4 be? For it to be mentioned and commented on by all these folks?

 

[Reverend]

SM5 is very unlikely at the moment as SM4 say that they driver have to work with HLSL and every shader with valid syntax have to run.

Oh bull... SM5 *is* likely, just like DX20 is likely.

MS will make sure of that when they target the next XBox.

Most here are offering opinions as facts.

Let's just take stuff like Vista and the latest IHV offerings as what they are, stuff you can/will be able to buy while hoping your purchase is justified.

 

[Xenus]

Out of the loop Rev. SM4.0 is supposed to be the big unification of the shader instruction set. So while they may not run on the same units. The intruction set is the same between pixel shaders and vertex shaders.

 

[Reverend]

I'm not too far out of the loop wrt SM4 (even though I admit having been generally disinterested in 3D for the past few months... it's been boring, don't you agree?). Some folks have (apparently) given me the benefit of the doubt when reading my comments in this thread but I was just being sarcastic.

I'm not even sure if shader unification is a good thing. But if it is a "must" to meet one of Vista's requirements, then, hell, everything about SM4 will be more important than a "stop gap" SM3 or such.

There isn't much in Vista/SM4 (that I know, from what I have been given and what I learned myself) that dramatically enhances 3D for the PC compared to pre-Vista/SM4. My opinion of course.