hd 5800 series has been a huge success for ati/amd in the high end market... perhaps we can move our eyes to the next...
the hd 6xxx series... so anyone with any thoughts are welcome to post here...
personally... i think there are three things that i expect or would like to see when hd 6xxx series kicks in....
1) improvement on DX 11 titles over hd 5xxx series
2) impoved anti-aliasing on performance
3) less or no driver issues please??
those are three things which are the downside of hd 5xxx... and those are the strong point for nvidia..... performance in dx 10 titles is already good in 5xxx series so they should focus more on the tessellation and other dx 11 features... you couln't imagine buying a hd 5870 and played a game with dx11 tessellation on and 4x AA and see how much it drops in performance...
From the rumors I've deemed believable, the vast majority seem to believe that the 6xxx's will be a minor jump from the 5xxx's. They don't seem to be a different architecture or a smaller die therefore able to pack more transistors. I imagine it will just be a bigger GPU with minor architectural improvements. I wouldn't expect a jump like the 3xxx's to the 4xxx's or the 4xxx's to the 5xxx's. Honestly, I could see a similar jump from the 2xxx's to the 3xxx's again.
Correct Descendency. 6xxx is supposed to still be on 40nm, and we won't get another big jump until they switch over to 32nm or GF as the manufacturer. There will be tweaks and ??? that will bring performance up some, but nothing major. A beefed up tess engine and perhaps better Eyefinity support would be great, but the 5xxx cards are already pretty good. Moving a better card into the $200 price range would also be nice, something that competes better with the GTX460.
According to current info, the improvements will be mainly the uncore, which would mean better AA performance, and perhaps after tweaking the tessellation engine for bettertessellation performance. I think(hope) they will also slap on sideport that was supposed to be on the RV870.
A personal guess for the HD6870 would be 1920SPs with better tessellation performance & AA scaling, and Crossfire scaling that matches (or maybe even exceeds) SLI while maintaining the same power draw.
But its just all mindless speculation until the card is out.
With no die shrink I'm not sure you'll see more cores on the chips. Unless they built some spare ones in that they can now enable. The only way they can get more cores on there is to increase the size of the chip (not likely) or make some room as they rearrange things (possible). Tweaks, changes, possible more voltage to allow higher clocks, etc. Thats what will make up the bulk of the 6 series. Larger chips maybe, but I'm not sure they want to change their performance per watt figures.
i would imagine a small performance increase either from higher clocks or slightly more shaders (i would say closer to 1760 that 1920), possibly both. But with the improvements to the uncore that is supposed to be changed for NI i would suspect better tessellation (maybe %50) though i wouldn't imagine AA will get touched that much
as for Crossfire scaling, i would think about the same as now unless the new cards are easier to write drivers for (which i doubt)
I really expect the next generation of ATi cards to scale much better in CrossFire.
i doudt it. if they wanted, they can make 5xxx series scale better by providing us better drivers.......if they want to improve xfire scaling, that is the place to start...driver.....i doubt the hardware is the problem in xfire issue
there was a time when ATI fanboys round here, pre-dx11, were actually duped into believing the 3800s would do good under dx11 (because of dx10.1).
Not quite? Its because of the DX10.1 that it will be able to run more DX 11 code then the Nvidia DX10 only cards. See the original review of Assassin's Creed from Hardocp to see what was going on. To bad they changed how the test works so that Nvidia cards were faster again.