Why is SM3.0 so important?

[Rugor]

I keep hearing how important SM3.0 is, and that everyone should implement it immediately and that ATI is making a huge mistake in not implementing it immediately.

What I want to know is what makes it so important to jump on the SM3.0 bandwagon, I understand why Nvidia is saying this, since it's a feature they support and ATI doesn't, but why do so many people agree with them?

Is there any reason, other than the fact that this time it is Nvidia that is introducing it, why it is so important to go to SM3.0 much more quickly than SM2.0 or PS1.4? Why is it different this time?

Nvidia spent millions trying to get people to use PS1.1 rather than PS2.0. But now there's a new model they want everyone to support PS3.0.

I understand why it's important to their PR, but why to anyone else?

[Bigjakkstaffa]

At this juncture, it isnt really important in the practical sense.

The only real benefits at present that we're going to see is increased speed in shader operations, however, as many people have pointed out, SM2.0 is still nowhere near being used to its full capabilities yet. Indeed there are only a handful of games at present that even utilise SM2.0, nevermind SM3.0. By the time SM3.0 becomes a mass technology (i.e. lots of people have cards capable of it) we''l probably find that SM2.0 based games are only just reaching widespread level and that SM 4.0 is already availiable.

Had SM3.0 been to SM2.0 what SM2.0 was to SM1.1 the story may have been different (ATi would probably be wetting themselves).

SM 3.0 at present represents little for the average gamer to get his/her knickers in a twist. The only people to whom SM3.0 looks to be of great significance to is Nvidia in their PR material, indeed what will make or break how widespread SM 3.0 becomes is how much money Nvidia is prepard to shovel at games developers to 'persuade' them to utilise it.

Overall, SM3.0 can probably be summed up as a technological step forward, but one whose timings is unfortunate, as with SM2.0 still for from widespread usage and SM4.0 coming in the not too distatnt form of the R600, it could easily become one of those technological features which is noteworthy for its existence, but never really utilised on a large scale as a result of its timing and the fact that when it becomes a practical mass market solution/technology, bigger and better things will have already arrived and wowed the consumer.

--Jakk:t

[bob05]

SM 3.0 might not be big, but it's not insignificant. Along with SM 3.0 brings true centroid sampling. Essentially it makes sure textures are only rendered on the triangle they are suppose to be on. This was the feature Gabe Newell was talking about when he claimed that AA will only work on ATi cards when running HL2. But eventually Valve got them working on both cards. Another feature new to the table (which no doubt you've heard of) is that Dynamic flow control that ATi "was trying to steer people away from". It prevents the execution of unnecessary shaders and such, thus cutting down on the amount of instructions sent to the GPU. In addition to all those, SM 3.0 will add AA for shaders, which could eliminate artifacts from some shader driven special effects and such. To top it all off, it lets the length of shaders be very long, something along the lines of 65535+. This is a much larger number than SM 2.0's 96.

SM 3.0 should improve on many aspects of development, and should let game developers focus more on making the a game good rather than trying to add more FPS to a game and/or use tricks to bypass limitations on currently available shader models and/or hardware. :t

[Rugor]

Bob, I understand many of the things SM3.0 is doing, including enabling centroid sampling (the HL2 AA issue), but that's not my real question.

Why aren't people seeing Nvidia's rush to immediate SM3.0 takeup as anything more than a crass marketing ploy because this is the first time since the GeForce 3 that Nvidia has introduced a shader model before ATI?

I find it strange that the same people (and I am not limiting this to Sysopt by any means) who could support the FX series and the replacement of PS2.0 shaders by PS1.1 shaders wherever possible are now jumping on the must upgrade to SM3.0 now bandwagon. Especially since both SM2.0 and SM3.0 are floating point formats where PS1.1 is still integer precision. SM2.0 and 3.0 are much closer to each other than either is to PS1.1

It looks pretty hypocritical to me.

If you want to go for the 6800 why not pick some of the features that really matter: It has better OpenGL performance, depending on the driver version it can enable angle-independent AF, and Nvidia's nView is better multi-monitor software than ATI's Hydravision. Those are all more valid reasons for saying you want to go for a 6800 based product over an X800 based card.

[bob05]

Originally posted by Rugor
It has better OpenGL performance

Exactly, it comes down to what you want to play by September, Half Life 2 or Doom 3. I wanted Doom 3 since 2001, but many people want Half Life 2. It's your call. :t

[Magua]

I don't think you can simplify Nvidia vs. ATI into a Doom3 vs. HL2. The performance of both games on both companies top cards will be outstanding. Even though each of the games is supposedly 'geared' towards one of the cards its only a few FPS, which dosen't make too much difference. I would base the decision on overall performance over several games and applications, not a HL2/Doom preference.

[bob05]

Originally posted by Magua
I don't think you can simplify Nvidia vs. ATI into a Doom3 vs. HL2. The performance of both games on both companies top cards will be outstanding. Even though each of the games is supposedly 'geared' towards one of the cards its only a few FPS, which dosen't make too much difference. I would base the decision on overall performance over several games and applications, not a HL2/Doom preference.

Of course you shouldn't base it on just one game, maybe I should have said which game you want to come free with your card. :p j/p, but it does just come down to preference of brand since essentially they are both neck and neck overall (surely you could get into specific instances where one card will out preform the other, but most people wouldn't notice specifics anyways). A smart buyer is an informed buyer. :t

[jamil5454]

Most people will play other games besides HL2/Doom3, but if you have them both, why? I know that when I get HL2 this summer I'll get 7.1 speakers and maybe a new monitor so I can have an awesome experience. I mean the game is at least ~30 hours long which is exceptional for a FPS. Plus, they're porting Half-life 1 and all its mods to the Source engine, which will give you many more hours of gameplay. Doom3 will be a much better game with surround sound than without surround sound, because it will be scarier. These two games are some of the biggest games ever, and even if Nvidia edges over ATI in most games/benchmarks, I'll still go with ATI because of better HL2 performance. Obviously I'm excited more about HL2 than doom3. If it were the other way around, I'd be persuaded to buy Nvidia.

[Magua]

Forgot that you get one of the games for free, that will be a huge selling point. Im sticking with my 9800pro for now, then will probably pickup an ATI next gen card when those roll around.