Tweet
9700 : Faster AA Eh?
*This may have been covered elsewhere, but I couldn't find it in the few other 9700 posts. My apologies if it has.*
Have you noticed something about the percentage/relative performance "benchmarks" in all those 9700 (P)REVIEWS?
Have you see one that DOESN'T have AA enabled? I haven't.
Why? Well this is just my (highly)uninformed guess, but perhaps it's because otherwise the Gf4ti4600 scores higher.
1337 fps p33p : "But how can that be? 9700 is a far Bigger Number(TM) than 4600. More than double!"
Bizzarely enough, I think that perhaps the brand new architecture (with it's unoptimized drivers), designed for DX9, isn't as fast in current DX8 games/benchmarks with NO AA/low detail as the "industry standard" chip. So what to do? Well just turn on AA. Gf4 curls up and dies, 9700's newer funkier hardware makes a difference. Much rejoicing at AtiHQ. Marketing Dept. explodes all over itself.
"Who cares about low detail optimizations? Nobody plays games without AA turned on anymore anyway! AA rulez!" etc.
Ok, maybe. But why was that other new release of DX9 featured hardware, (Speak not it's name, for it is cursed!), deemed "slow" again? Low performance in low detail? Hmmmmm. But it is faster than the current champ when you turn AA and high detail on, yes?
The golden rule of high detail performance scaling on the existing hardware, AA = big hit, has been applied to both the newcomers. But obviously this rule no longer applies : in low detail settings with software that isn't coded to use those optimized designs/increased bandwidth/enhanced shaders they are handicapped. The optimized Brute Force approach wins, as all those funky parts of the new pipelines lay idle. Switch on Highest detail and AA, in each chips differing implementation, and the reverse occurs. Don't we see this happen everytime a new CPU architecture is introduced? Little or no performance gain over the existing standard on exisiting code.
Faster in the new stuff it was designed for.
It could be just me, but isn't it odd that the same outdated performance scaling interpretation, "It's gonna be much slower with AA turned on", has been used to crucify one new chip and praise the other? Even if they do the same thing? Now I could be totally incorrect, wouldn't be the first time, and the benches may have been presented in this manner just to show how much more a$$ a 9700 kicks in current software. Fair enough.
I'm not championing(sp?) one of the new chips over the other : both are far more intelligent designs than the current standard. But I've never owned an ati card, and have had a fantastic experience with the one piece of matrox hardware I own. Show me the other Tech support/fansite forums that are this useful, happy, helpful and prompt. I also appreciate the innovative extra features of the P. Which is the reason I purchased and am still using my G400 MAX - 4 year old features that are only just now trickling into the mainstream. Innovation must be applauded. So I'm probably leaning towards the Parhelia anyway, some bias applies. My point is : benchmarks are like statistics. You can make the numbers dance to whatever tune you like, it's all in the interpretation. Hence they should be given the same credence as a Used Car Salesman, who part-times as a Journalist and dabbles in Politics on the weekend. You can't use the same interpretation of performance to prove a pattern and then disprove it. If you are going to live by the benches, you're going to have to be prepared to die by them too.
Ahh, but I digress.
Not sure if any of this makes sense, but i'm too lazy to proof read it. At least I feel better
Ranting is similar to alcohol : practice moderation and it can be good for you.
Thanks for reading
Have you noticed something about the percentage/relative performance "benchmarks" in all those 9700 (P)REVIEWS?
Have you see one that DOESN'T have AA enabled? I haven't.
Why? Well this is just my (highly)uninformed guess, but perhaps it's because otherwise the Gf4ti4600 scores higher.
1337 fps p33p : "But how can that be? 9700 is a far Bigger Number(TM) than 4600. More than double!"
Bizzarely enough, I think that perhaps the brand new architecture (with it's unoptimized drivers), designed for DX9, isn't as fast in current DX8 games/benchmarks with NO AA/low detail as the "industry standard" chip. So what to do? Well just turn on AA. Gf4 curls up and dies, 9700's newer funkier hardware makes a difference. Much rejoicing at AtiHQ. Marketing Dept. explodes all over itself.
"Who cares about low detail optimizations? Nobody plays games without AA turned on anymore anyway! AA rulez!" etc.
Ok, maybe. But why was that other new release of DX9 featured hardware, (Speak not it's name, for it is cursed!), deemed "slow" again? Low performance in low detail? Hmmmmm. But it is faster than the current champ when you turn AA and high detail on, yes?
The golden rule of high detail performance scaling on the existing hardware, AA = big hit, has been applied to both the newcomers. But obviously this rule no longer applies : in low detail settings with software that isn't coded to use those optimized designs/increased bandwidth/enhanced shaders they are handicapped. The optimized Brute Force approach wins, as all those funky parts of the new pipelines lay idle. Switch on Highest detail and AA, in each chips differing implementation, and the reverse occurs. Don't we see this happen everytime a new CPU architecture is introduced? Little or no performance gain over the existing standard on exisiting code.
Faster in the new stuff it was designed for.
It could be just me, but isn't it odd that the same outdated performance scaling interpretation, "It's gonna be much slower with AA turned on", has been used to crucify one new chip and praise the other? Even if they do the same thing? Now I could be totally incorrect, wouldn't be the first time, and the benches may have been presented in this manner just to show how much more a$$ a 9700 kicks in current software. Fair enough.
I'm not championing(sp?) one of the new chips over the other : both are far more intelligent designs than the current standard. But I've never owned an ati card, and have had a fantastic experience with the one piece of matrox hardware I own. Show me the other Tech support/fansite forums that are this useful, happy, helpful and prompt. I also appreciate the innovative extra features of the P. Which is the reason I purchased and am still using my G400 MAX - 4 year old features that are only just now trickling into the mainstream. Innovation must be applauded. So I'm probably leaning towards the Parhelia anyway, some bias applies. My point is : benchmarks are like statistics. You can make the numbers dance to whatever tune you like, it's all in the interpretation. Hence they should be given the same credence as a Used Car Salesman, who part-times as a Journalist and dabbles in Politics on the weekend. You can't use the same interpretation of performance to prove a pattern and then disprove it. If you are going to live by the benches, you're going to have to be prepared to die by them too.
Ahh, but I digress.
Not sure if any of this makes sense, but i'm too lazy to proof read it. At least I feel better
Ranting is similar to alcohol : practice moderation and it can be good for you.Thanks for reading
Comment