20% figure as an average is impossible. Its closed to possible with a brand new architecture. Something like 50% only occurs when their are issues to begin with. Fermi is performing somewhere around 8% better and it is a brand spanking new architecture.
When techpower tested between launch drivers and 10.4 catalyst(the one that almost got wizzard to quit). There was only a 2 percent improvement. A complete average 20% is pretty much crazy impossible. I was saying 20% improvement with drivers for the GTX 480, and people were saying that was too high. I agree now, even with a new architecture, its hard to get that much of an improvement.
Considering that this architecture is all about making the cypress and older architecture and is heavily based on it, it should be really efficient from the get go.
Where did you get this 50% figure from, I would like to see it. I have never heard of such improvements outside crossfire or a freaky issue or a flaw which decreases image quality.
Or they pass on a good deal on a 5870 or 5850. Such a crappy naming scheme also really hurts the reselling of used cards. A $259 dollar 6870, makes it look like a 5870 has to be priced around 150 or so to be priced properly, even though it is the faster card.
If they think scheduling




Reply With Quote
Bookmarks