Computerbase.de averaged 11 modern games. They got 16% at 1920x1200 8AA, and 18% at 2560x1600 4AA. So your estimate of 10-15% is not correct. It's rather 15-20%. In Anandtech's case, the performance difference was also around 15% or more.
In this article is saying 5970 remains the fastest graphics card in the world. Correction in that article its the fastest dual gpu but not the fastest single gpu.
It is the fastest CARD in the world.
That must have either included truffles or at least 1 bottle of expensive wine? ^_^
Release date and performance remains to be seen but this smells like "damage control" since NV released the 580 . Fudzilla and and an "insider" as sources? I'm surprised that no one is questioning these reports ala when the 580 was first rumored, the tone from some was that it was all a ruse to rain on AMD's upcoming 6xxx release!
Have you seen the reviews lately for the gtx 580.
Have you? The 5970 is still the fastest single card solution money can buy, but a lot of people think that multi gpu setups have cons that outweigh the marginal performance advantage the 5970 has over the 580.
Sorry, who's doing the damage control? I'm a bit confused by your post.
I don't think either company needs to do damage control right now.
That doesn't mean associated parties won't be carrying out their own black ops regardless.
I see what you did there.
It is the fastest Direct x 9/10 CARD in the world.
At all resolutions, it was 11% faster. At 2560 it is 16.3% faster. At 1920 it is 14.9% faster.
OK, that makes sense. I see hwo you arrived at the 10-15% then. I am not sure 1024x768, 1280x1024 and 1680x1050 are that good to test though for a $500 GTX580 (esp. the first 2 resolutions).
I'd love more tests done at 5040x1050 and 5760x1200. Perhaps the least costly way to do that is to ax the 1024x768 and 1280x1024 resolutions. Maybe keep those low resolutions for lesser cards, but at midrange and higher, transition to the Eyefinity resolutions.
fixed
In direct x 11 games it loses, that more important.
Depending on the resolution and depending on the game.
For example, the 5970 is better than the 580 in Metro at any resolution higher than 1680x1050, but is slower in Dirt 2 at all resolutions (though both cards still play the game well).
You can't say "it loses in DX11 games" part and parcel because that's just false. It wins some, and loses some. Overall, it's still the faster card.
Depending on the resolution and depending on the game.
For example, the 5970 is better than the 580 in Metro at any resolution higher than 1680x1050, but is slower in Dirt 2 at all resolutions (though both cards still play the game well).
You can't say "it loses in DX11 games" part and parcel because that's just false. It wins some, and loses some. Overall, it's still the faster card.
Whats the point of buying this card when it only performs in x10 games. You can to play futures games right based on the right software.
It's not going to be 15-20% faster than the 580. The 5870 was slower than the 480, so why will there be a sudden turn around this time?
Not because AMD new what they were up against. They try to make the card as fast as possible within power limits and cost constraints - knowing how quick the opposition is doesn't suddenly make them do a better job. It's not like they'd normally not bother trying.
Whats the point of buying this card when it only performs in x10 games.
Look, I'm not trying to say that people should buy a 5970 over a 580, and that's because I'm not a fan of multi-gpu setups anyway.
Ehh? The 5870 came 6 months before the GTX480, beating everything NVIDIA had and even with GTX480 the difference between the two is pretty small.