in complex shading scenes SM3 works just as advrertised, actually moreso.
moreso compared to what? what the nvidiots were claiming? no...
the main stumbling block many thought SM3 would have was the performance hit of a single shader with multiple instructions. as it turns out the pergormance hit is not that great, at least according to crytek ,but what would they know?
i never questioned that.
the 1.2 patch does not utilize very many SM3 shaders, in a few complex scenes some SM2 shaders were replaced by SM3 for the sake of performance.
how many and where?
can you logically explain to me how judging a patch replacing a few shaders is enough criteria to determine the actual performance benefit of SM3 as a whole?
again, how "few" shaders?
but you miss the entire point of what i said.. it's not about sm3 bashing.. why is that so hard to grasp?
a) nvidiots screamed i'd cry when the path came. it's here. i'm not crying, nor do i have reason to.
b) the topic of the post i ridicule stated 33% increases, which are false, and based on erroneaous tests. the truth is (according to the source you like to quote):
Based on our results with Far Cry, Shader Model 3.0 does bring NVIDIA tangible performance benefits but the results will vary based on usage. In levels with multiple light sources such as the ?Volcano? and ?Research? demos, we saw performance gains of up to 18%. The beauty of this is it?s a ?free? performance increase to the end user, there?s no need to turn off eye candy features to get this kind of performance.
In outdoor areas such as our custom Monkey Bay demo and the ?Training? and ?Regulator? levels present in Far Cry 1.2, the performance benefits are more limited.
At best the performance increase is about five percent, roughly the boost you may get from a driver update. This still isn?t bad, but it isn?t a dramatic difference, we wouldn?t be surprised if some users hardly noticed it.
could these results be improved? sure, it's possbile. does it make nv40 faster than r420? well, not really. it's a mixed bag. the combination of driver (even tho 56.72 is the last "official" nv driver) improvements and sm3 improvements has resulting in the 6800 being almost as fast as the XT, and the GT faster in some areas than the PRO. overall, it's pretty comparable.
as for the judging of criteria, my only criteria was refuting the idiotic claims made by zealots. beyond that, the only claims i can make are those made above. any claims as to "how much benefit" can be gained "sometime in the future" is speculative, at best.
i've always said sm3 will have benefits. i've also said it won't see it's potential till much further down the road. how can you logically argue that?
that makes about as much sense as showing benchmarks where SM3 "makes no difference in performance" and neglecting to mention(or realize) that in many areas no(or very very few) SM3 instructions are even used(thus no difference...)
as i've stated many times, far cry is predominantly sm1.x shaders.. why would it? at the same time, how would it have a significant impact on the game? it won't. it will benefit those limited areas that rely on ps2 shaders being replaced by ps3.
again, i'm not sure what you're trying to argue?
now.. getting off the specific topic, i was just pointed to a new article where crytek stated they will only support fp32 HDR, tho the effects/quality of fp24 would be the same.. interesting they would deny 95% of gamers simply because of nvidia marketing? hmmm.. at any rate, this may be the first actual visual effect which will be sm3 specific (a marketing decision, not a technology one), but i didn't see an eta of when this might happen. will be interesting.
oh wait, plenty of people are doing just that... nevermind!
well, at any rate, it seems to me you are arguing just to argue, since you're really not staying to my specific comments, rather making arguments based on a broader scope than my comments...
and since you're directly quoting me, well... it's seems you're simply trying to "stir the pot"... or you've simply missed the context of what i stated.
