Why SM3.0 != NVIDIA only

[DivotMaker]

i remember the long discussion about the Standars in the past ,how much they were important ,but now suddenly it doesnt matter to follow that and only that. :rolleyes:

I remember all too well the long drawn out discussions from the FX era. Show me where PS 2.0b is NOT part of the DX9 standard?

ask developers at EA and see what they tell you ,later come here and post your findings before judging me as Fanatic. :).

I have spoken to the developers at EA. You really do not want to get into a discussion about this issue with them after the hoops they had to jump through to get TWIMTBP/FX hardware to perform acceptably in DX9.

Secondly, I am not judging you as a person, but your comments here are so blatantly apologetic towards nVidia and what they pulled last year. To compare PS 2.0b with the shader replacement/shader compilers from the FX era is simply not fair and misleading. When you can show me that PS2.0b causes a developer to spend 5X more coding time to get ATI hardware to perform (ala HL2/FX cards), THEN I will understand what you are getting at.

ATI is NOT sacrificing IQ to get PS2.0b to perform. It is a GREAT feature that is capable on ALL R300^ cores.

if im a fan it is about technology and computer graphics ..; (hence why i use Nvidia hardware).

So am I. That is why I feel very strongly that your comments are biased and irrational. PS2.0B is not something that will stall PC game development. It simply adds greater performance and functionality to an era of cards available NOW. I cannot fathom how you can consider this as something "bad" for the industry.


the day ATI produce first Sm4.0 hardware (and fix their opengl problems) ,ask me ,i will be the first to jump the ship .i believe that Companies that push forward for better graphics in games ,should be supported ,whenever $$ its possible. Doing this is helping the future games to arrive earlier and not later . and to move to the Standars is the way the Industry will follow.

Since when did SM 4.0 enter into the discussion? ATI is following MS DX Standards. Just because they are exposing functionality in current cards that offer full compatibility with DX9 and better performance that makes them more competitive with nV, you call this a "bad" thing. Sorry, your line of reasoning makes ZERO sense to me.

[DivotMaker]

It's absolutely mind-boggling that you can say this after Nvidia's '03.

It would appear that his mind is "boggle-proof"....I could not agree more Schuey, but this guy has got some seriously green blinders on....

[DivotMaker]

not at all.. Nv3x supported excelent futures ,ahead of Ps2.0. there wasnt anything better feature wise. (not performance wise). Unreal3 for example use Fp32 everywhere ,with up to 150 shaders intructions long ,what hardware do you think Epic used the first 18months? havent you read JC Nv30 comments which card he use as primary develoment platform and why? game development if one thing ,to play games is another.

Features do NOT mean squat if you can only render in slide show framerates....man, you have lost the plot here.

[ChrisRay]

I have spoken to the developers at EA. You really do not want to get into a discussion about this issue with them after the hoops they had to jump through to get TWIMTBP/FX hardware to perform acceptably in DX9.

Secondly, I am not judging you as a person, but your comments here are so blatantly apologetic towards nVidia and what they pulled last year. To compare PS 2.0b with the shader replacement/shader compilers from the FX era is simply not fair and misleading. When you can show me that PS2.0b causes a developer to spend 5X more coding time to get ATI hardware to perform (ala HL2/FX cards), THEN I will understand what you are getting at.

ATI is NOT sacrificing IQ to get PS2.0b to perform. It is a GREAT feature that is capable on ALL R300^ cores

Do you mean instancing or PS 2.0B in general? The r300 doesnt support PS 2.0B but it does apparently support instancing. The r300 is maxed and limited at 96 instruction. And the big deal about 2.0B is the X800 being able to do 512.

Other than that I could say I agree. Well not in regards to the Hl2/FX cards things. But thats a rather personal issue with me. I think devs should work their hardest to obtain the maximum quality/performance from any hardware, Not work to make one hardware look worse/better than the other.

If that is done on both hardwares I am very happy with the end result. But anyway Sm 2.0B is an X800 exclusive (well Nv3x and Nv40 can both run Sm 2.0B fine) But its not supported by the r300. Apparently the r300 has always supported instancing. Thought there has been some debate whether thast instancing is emulated by the CPU. Personally I have no information for or against that, But its not totally impossible to theory that it could be done by the CPU,


Features do NOT mean squat if you can only render in slide show framerates....man, you have lost the plot here.

I actually agree with him here though, The reason the Nv3x is used as a developer platform is because it supports features, It is completely backwards compatible for NV2x/NV1x, And also has a very good base featureline set that even the Nv4x series can benefit from/use.

I think using the Nv3x as a developer platform might actually make sense if you talking about compatibility sakes, However with the introduction of the Nv4x which supports All the Nv3x did, It makes sense to move forward to the Nv4x from a developer platform standpoint.

But theoretically. Much of what can be done on the Nv4x can be done on the Nv3x as well.

[jimmyjames123]

but the person who composed the comments above must have a very short and/or convenient memory, OR, he knows nothing about the hoops developers have had to go through to try to get FX hardware to run decently with pixel shaders.

BigBertha, the guy who wrote this IS a developer, and has developed PC games in the past! He knows much much much more about the industry than most people here, I can pretty much guarantee it. I suggest that you go to the B3D forum to read more of his comments (and rebuttals) here:

http://www.beyond3d.com/forum/viewtopic.php?t=14392

Ps2.0_a is entirely different as shaders made in that will be unoptimal for other architectures due to trying to work around the oddities of the gffx thus making work in that useless especially as the result may still be that the path is unusable on the gffxs if you really are doing advananced effects that need ps2.0_a

Actually, I believe it was Demirug at B3D forum who stated that the FX cards do have a performance boost when using a ps2.0_a path in FarCry. Remember that their compiler is much improved over what we initially saw at introduction.

[Nv40]

Features do NOT mean squat if you can only render in slide show framerates....man, you have lost the plot here.

and who is saying playing games here? for Game development you dont need realtime speed ,(its helps) but its not neceesary .Nalu was codec in Nv3x line running at ~3fps.. now it runs in realtime thanks to Sm3.0 and the new hardware. its the same thing with *games development*. i guess some people here need more time to see in what way DElays (supporting featurities) will impact their future games . everything you need to know have already been said since the first post..not by me ,but by real gamedevelopers ,so its up to people here to accept it or not .

[DivotMaker]

and who is saying playing games here? for Game development you dont need realtime speed ,(its helps) but its not neceesary .

You ARE kidding, right? If not, then I will refrain from further replies to you.

[mikechai]

and who is saying playing games here? for Game development you dont need realtime speed ,(its helps) but its not neceesary .
You ARE kidding, right? If not, then I will refrain from further replies to you.

I seriously think Game development doesn't need smooth frame rate, and that's why NV3x is a better development platform compare to R3xx for its extra feature set. Of course now it has been superceded by NV4x.

For example, Doom3 development started about 4 years ago, don't tell me you can get it run at > 30fps high quality with the cards back then.

And when a game finally available in retail, 3d cards speed has improved by leaps and bounds.

[Helanic Frost]

Features do NOT mean squat if you can only render in slide show framerates....man, you have lost the plot here.

I actually agree with 98% of everything you've said in this thread except where you call it slide show performance. True the FX series didn't have as good of shader performance as ATI, slide show? I'm sorry i totally disagree with you, not one single dx9 game i own do i have performance that is that abysmal, 97% of all dx titles run very satisfactorily on FX cards with the performance crown going to ATI without question but that kind of comment is bordering on being a cartoonish exaggeration.

[Clay]

@ BigBerthaEA, I don't recall you getting so worked up in defense of NVIDIA back during the FX days.

[Helanic Frost]

Btw when i was playing SWG with my FX5900 i was able to actually use shadows in game and still maintain a playable framerate and ATI users were not able to. Ultrashadow tech is an amazing feature if used right. So yeah the features did mean something to some of us. SWG doesn't support AA and Sony recommends their players not use it, nvidia had better AF and being that i played SWG for a year for 90% of my free time i bought the right card and feel it served me very well. So while you may feel it was a stinker card i won't actually disagree with you. The games we all play vary and when making a card purchasing decision you have to consider what features it supports and what features are in games you want to play. Bottom line here is don't argue, buy a card that plays the games YOU want to PLAY best and ignore all else.

Btw one of the major reasons more games are TWIMTP is because most games are developed on NVIDIA hardware. There's a reason for that. The 9800 pro was the better performer and the NV30 had the more robust feature set, more features = better development platform. Faster hardware = better gaming platform for most people. Simply put you are both right and wrong. End of story.

[Nv40]

Helanic Frost :

btw.. i dont upgrade every 6months.. far from that. i upgrade every 2-3 years ..my last hardware was a Geforce4 ,which amazingly never gave me a single problem worth of mention in that time,best $400 dollars spend :) . now i have a Gt right on time for DOom3 and close to other games. .hopefully i will upgrade more frequently if i manage to buy a PCi 6800 ready for SLi for my main machine.that will be enough for until DIrect10. i think SLI + SM3.0 is the future.

[pat777]

OMG, I was scared that I was banned from Beyond3D but then I realized that I was in Dean(from what he told me his real name has no "o")Calver's section.

Anyways, Real-time performance isn't need for development, it's needed for customers. For example a game could have a 6 million polygon model and have a version of the model with 6000 polygons with a lot of normal/displacement mapping. Will 6 million polygons be real-time?

[DivotMaker]

I actually agree with 98% of everything you've said in this thread except where you call it slide show performance. True the FX series didn't have as good of shader performance as ATI, slide show?

It was an exaggeration to make a point. When I compared the PS 2.0 performance in the few games I played, the FX appeared to be a "slideshow" referring to dramatically lower performance compared to the R350/360 cards I was using. I hope that makes it clearer.

[DivotMaker]

@ BigBerthaEA, I don't recall you getting so worked up in defense of NVIDIA back during the FX days.

Clay,

You have a short memory then. I sparred with DW, Hellbinder, and other ATI users when the first FX cards came out. After seeing nV's performance from that era for myself, as well as some of the driver shenanigans, it became very difficult to support them. I support them now as they have corrected many of the issues they had from before.

I am also defending ATI here because I do not at all agree that PS2.0b is a bad thing for the industry. I really do not care what one supposed game developer says...I use common sense and in this particular issue, I feel I am right. Obviously some don't agree. If I thought there was something wrong with PS2.0b, I would jump on the bandwagon, but I simply do not see how it is a "bad" thing and is really not an MS "Standards" issue.

[DivotMaker]

I seriously think Game development doesn't need smooth frame rate, and that's why NV3x is a better development platform compare to R3xx for its extra feature set. Of course now it has been superceded by NV4x.

For example, Doom3 development started about 4 years ago, don't tell me you can get it run at > 30fps high quality with the cards back then.

And when a game finally available in retail, 3d cards speed has improved by leaps and bounds.

That makes sense....I was combining features with performance during development and I realize we were talking about different things. My confusion. Apologies to all.

[Clay]

Clay,

You have a short memory then. I sparred with DW, Hellbinder, and other ATI users when the first FX cards came out. After seeing nV's performance from that era for myself, as well as some of the driver shenanigans, it became very difficult to support them. I support them now as they have corrected many of the issues they had from before.

I am also defending ATI here because I do not at all agree that PS2.0b is a bad thing for the industry. I really do not care what one supposed game developer says...I use common sense and in this particular issue, I feel I am right. Obviously some don't agree. If I thought there was something wrong with PS2.0b, I would jump on the bandwagon, but I simply do not see how it is a "bad" thing and is really not an MS "Standards" issue.My mistake then but not a short memory...there a lot of you folks to keep up with you know? ;) I don't think that SM2.0b is necessarily a bad thing either. I'm always for furthering the implementation and adoption of new technology. However, I do find all of this very recent hubbub surrounding SM2.0b as somewhat suspect. Why are we just hearing about it now? Why was it not brought up months ago when SM3.0 support was announced for the NV40? One would think that it would make perfect sense to use this SM2.0b capability of the R420/R300 architecture to quiet NVIDIA's SM3.0 marketing buzz before it ever got rolling.

[Drumphil]

er, hasn't sm 2.0b been part of the DX9 spec for ages? (since the beginning I think)

[Nv40]

er, hasn't sm 2.0b been part of the DX9 spec for ages? (since the beginning I think)

nope.. it was added just recently because ATI requested it .it was announced in Gamedeveloper conference 2004 .SM2.0 ,Microsoft Shadermodel Standar Sm3.0 and Sm4.0 in the other hand was released long time ago. ps2.0b path have less features support than 2years old ps2.0a. if there was a hardware close to SM3.0 ,that was the Nv3x line. so thats why is literaly speaking a step back ,more because is introduced 2years later so its an unusual move by an IHV. Ps2b is not an Standar ,its only exist in ATI land ,and it was a stop gap for ATi ,until their Sm3.0 parts.dont confuse emulation with support..special paths can be emulated ,but that doesn means they are supported. if ATI were not working in Sm3.0 NEXT GENERATION hardware for the PC and the XBox and PowerVR were not also Sm3.0 compatible ,things will be diferent . thats the purpose of Standars ,for everyone to support it. its SM2.0 without its limits and everyone needs to support it. the next Sm4.0 will not be released until 2006/2007 and will be the base for "DIrectx10". so it can be say as Biased that it sound that the real Dx9 vision of Microsoft was SM3.0 (hence it can do special effects with virtually unlimited Pixel/vertex shaders intructions count ) and that Sm2.0 was an introduction of it ,the basics. and while speed is not a requirement for gamedevelopment with SLI technolgy there is another level of posibilities. what ATi wanted is to save transistor to focus in better speed ,well.. look at the results .. there is no clear performance leaders in the high end ,it depends of the games tested and settings used. if Nvidia followed ATI way of speed and speed with minor modifications in their hardware without a doubt it will be clocked much higher and faster , but noone will be able to test in 2004-2005 the benefits of what have been already achieved in the Nv40. So you will enjoy earlier True HDR and Sm3.0 in games earlier (a year later) thanks to ->NVidia.

[euan]

If SM3 isn't Nvidia only, then why is there a device ID check to activate the SM3 path in farcry? :rolleyes:

[dan2097]

I am curious, What brings you to the conclusion there are more SM 2.0B than SM 3.0 card? There seems to be a fairly equivalent amount of X800/6800 hardware out there.

Other than that theres more ATI cards which support SM 2.0(Norm) though.


Im counting the 6800s as SM2.0B cards. As sm3.0 is backward compatable they support it. The amount of X800+6800 owners is obviously greater than the number of 6800s :)

Basically support the lowest shader version that you can to get the job done on as many cards as possible. If the effect really isnt feasible in ps2.0b then by all means use ps3.0 though just dont skip straight from ps1.1 to ps3.0 for marketing reasons

Actually, I believe it was Demirug at B3D forum who stated that the FX cards do have a performance boost when using a ps2.0_a path in FarCry. Remember that their compiler is much improved over what we initially saw at introduction.

Was this on patch1.2? Patch1.2 seems to cripple the gffxs performance by removing the overly aggressive optimizations so as you are running the full ps2.0 effects better performance would likely be seen using ps2.0_a. My point is though that in terms of image quality/performance you may be better off using a mainly dx8 path and a higher resolution due to the gffxs lacklustre dx9 performance well atleast thats the way gabe seemed to think, we'll see how other next gen dx9 games handle this.

[Nv40]

If SM3 isn't Nvidia only, then why is there a device ID check to activate the SM3 path in farcry? :rolleyes:

Farcry its just a patch ,there is a big diference between a patch or a game designed around it . pure "SM3.0" is not used at all in Farcry ,Crytek developer told it many times ,the effects demostrated in their "SM3.0" mod they told were also possible with Sm2.0.

[schuey74]

Farcry its just a patch ,there is a big diference between a patch or a game designed around it . pure "SM3.0" is not used at all in Farcry ,Crytek developer told it many times ,the effects demostrated in their "SM3.0" mod they told were also possible with Sm2.0.
And since there is not a single "pure SM3.0" announced yet, why do you need it again? Oh, excuse me, Unreal 3 in '06/07 which your 6800 will absolutely blaze thru @ 320x240. In fact, HL2 would be the closest thing to a "pure SM2.0" game and it's not even out yet. So this year, and quite possibly all of next year, there will not be one game displaying effects not possible on ATI's X800s.

[ragejg]

And since there is not a single "pure SM3.0" announced yet, why do you need it again? Oh, excuse me, Unreal 3 in '06/07 which your 6800 will absolutely blaze thru @ 320x240. In fact, HL2 would be the closest thing to a "pure SM2.0" game and it's not even out yet. So this year, and quite possibly all of next year, there will not be one game displaying effects not possible on ATI's X800s.

old argument, d00d.

[Riptide]

So this year, and quite possibly all of next year, there will not be one game displaying effects not possible on ATI's X800s.
Which is exactly why I don't think SM3 alone is a big make or break feature for the 6800. It's nice, but definitely not a giant advantage @ the moment. Marginal is more like it.

The 6800 has other more tangible assets right now that make it attractive.