tranCendenZ said:
heh my post clearly says "Application AF" right next to the pics, its even in your quote of my post
And performance drops by about 10% with the 8.07 drivers using CP-AF. Therefore, in order to maintain IQ parity in terms of AF quality with Nvidia, it must be considered that scores run using Application AF (most) are inflated by about 10%.
Microsoft actually states that NV40's filtering is superior to refrast, to quote:
Man will you stop saying DTS? DTS is a
home theater sound format. DST is Nvidia's shadow technology.
DTS = Digital Theater Systems
DST = Depth Stencil Textures
And my thoughts on DTS are that it kicks Dolby Digital's butt
We could say the same about 3dc, as nothing official has been announced for it for the next directx and it won't be incorporated until DX10 at least. Like 3dc, its currently under discussion whether DST will make it into the next DirectX. Futuremark claims they use PCF+DST because it is a part of DirectX for Microsoft's XBOX. Anyway, DST can be disabled.
I also question the usefulness of 3dmark05, because I think it is a technologically outdated benchmark. I snipped the rest of your post because I am also disappointed with 3dmark05. It claims to test SM3.0, and fails, not using most of the featureset that would increase performance - heck it doesn't even use geometry instancing, nevermind the longer instruction lengths or dynamic branching. Heck it doesn't even take advantage of SM2.0B. It doesn't use OpenEXR HDR Lighting that current Nvidia and future ATI cards will use.
When you are arguing about this, think about what you are arguing for. You are complaining that it uses DST, but what you should really be complaining about is that it doesn't use *enough* of the new features. It doesn't truly use SM2.0B (dx9.0c), it doesn't truly use SM3.0 (dx9.0c), it doesn't use OpenEXR HDR (dx9.0c), it doesn't use 3dc (nonstandard atm), it doesn't use virtually any technology that was put on a GPU this year. If you want to complain about anything, complain that 3dmark05 measures the potential performance of last year's cards, not this year's cards. 3dmark05 is a good benchmark for the FX 5950 versus the 9800XT, but fails as a useful benchmark for the 6800 Ultra versus the X800XT as it lacks the shader technology to fully take advantage of either of the latter cards. It isn't even a useful game benchmark, again unless you are measuring last years games, as this years most popular DX9 games (i.e. FarCry, Half Life 2, Painkiller: battle out of hell, lotr: battle for middle earth) and next years most promising games (i.e. unreal 3) are using technology that 3dmark05 simply does not incorporate. That should really be the issue - that it fails to predict potential performance for this year's hardware or even this year's most popular DX9 games due to its outdated shader code.
thx
OK, see, now I know what your trully getting at, and parden me folks as this is going to be get really off topic here.
before I start, I'd like to thank you for correcting my typo on DST (DTS...lol).
I was speaking about audio all last night and it was just funny to see your reply and how I made such a funny mistake.
anyway, let's start at the top.
at 4xAA and 4xAF, I get 4614 marks with the app.
I get 4762 marks with the CP.
this is more like a 2% differance not 10%, and did you notice just when I got this boost in scores?
I got the boost in scores when running the CP filter settings.
I repeated both tests to make sure all is good.
now apple740, also tested this, and did get a performance decrease using the controle pannel, but it was closer to 5%.
because I'm sooo OCed in all my parts, I'm quite sure that others will get differant results then both of us, too.
I'm kind of wondering just WHO you are listening to?
and that brings me to the whole "this is not a current card test" as you claim.
(what? because it doesn't use all SM2.0 and SM3.0 it's not current?)
first off, while SM3.0 might have been added at the last moment, it looks more like it was done, or redone with SM3.0 in mind...
and that's just my thoughts, and no one elses.
but really, what are your thoughts on SM3.0?
do you think that you'll always get a performance boost when using SM3.0, such as dynamic branching and geometry instancing?
and your thoughts on HDRL? do you think that you'll always get a performance boost with that increased IQ?
most SM3.0 is not, I repeat, not, a performance inhancment over SM2.0.
and at times, you have to be real carefull as to just when to use SM3.0 type rendering, because it could be a hinderance to framerate.
Farcry doesn't use dynamic branching, it uses static branching. why?
you might want to understand that unlike what most FUD you see each day in forums all around the world, SM3.0 can hurt the 6800 cards more then helps them.
that's a fact.
there will be times, when SM3.0 will help in both IQ and performance.
but if it's not used in the most wisest of ways, it could drop performance into the unplayable state of framerates.
since many people don't clearly understand just how poorly SM3.0 will perform on the latest cards, I advise that you do a search of my name and look for a thread on SM3.0 and Farcry, to see why adding more parts of latest tech, just for the sake of adding it, could HURT your results in both games and futuremark's 3dmark05.
adding some parts or features, to futuremark's 3dmark05, when it does not benifit from it in any way, is not what this benchmark needs, nor what any game, including farcry, should do.....IMHO.
maybe in a year or two, both ATI and nVidia will have cards fast enough to really use SM3.0 like SM2.0.
and then there are games, that may use only tiny amounts of SM3.0 in the next 3 years.
don't expect a game to truly use large amounts of SM3.0 till 2006-2007.
(and even then it will be only used when it helps, not hurts older cards.)
even games such as EQII will only use small amounts of SM3.0, just what makes you think SM3.0 is a "be all, end all" in framerate performance as compared to SM2.0???
it is clear that the NV40's are hindered by longer shader instuctions.
if at times, SM3.0 could have longer instructions, how could it be quicker then SM2.0?
now I know we have gone really off topic here, and you might not like 3dmark05 for differant reasons then me.
yet I still strongly advise that you might want to start a new thread, instead of saying that I should start arguing in a way that you want me to.
after all, again, that is not the point of this thread.
I do want to thank you for your time and thoughts.
(oh, and one more thing. no, you can't turn "OFF" DST in the default test, so it is really an issue that is needed to be discused and fought to correct in this benchmark, IMHO.
I was hoping that we could be able to do an "apples to apples" comparison with this test, and it seems that we might not be able to....that's sad, if you ask me.)
mica