Thanks kaktus for sharing :up:
Thanks kaktus for sharing :up:
What's "EPG"?Quote:
Originally Posted by kaktus1907
@the amd slides benchmark slides.
obviously they did not want to show themselves loosing in any benchmark. FC2 was in one but not the other, probably because its still slower in that game. then again, there were not very many games where nvidia has such an optimization as that game. if you dont see it on the chart, its probably too old, or it didnt win.
well all in all good.
but 6970 aka cayman needs to be much better as I expect it to go almost toe to toe on 5970
I'd say 6950 will be maybe 10% faster than 5870 and 6970 maybe 30-40% faster.
Quote:
Originally Posted by Pantsu
I think you are expecting a bit too much.
I keep my bets lower than that.
Hd5870 < Hd6950 - 5 - 10%
< Hd6970 - 25 - 30%
Until we see real numbers it's pretty much priceless. :D
Its tough to expect much more out of the same process, but keep in mind that AMD used a 30% smaller core and got similar performance... and Cayman is larger than Cypress by a good amount, so I bet its performance is a lot higher than people thinkQuote:
Originally Posted by andres1
Nice. Looks like the same size as HD4850.Quote:
Originally Posted by FischOderAal
Which is actually about where the last generation stood.Quote:
Originally Posted by Pantsu
5850 = 10%+ faster than HD4890
5870 = 30-40%+ faster than HD4890
Of course this varied from game to game.
Quote:
Originally Posted by Macadamia
so much less transistors ... i hope rumors are true ... super low die size .. super low pricing maybe??? :D
25% less silicon and same performance = huge proffit margin making gpu ... nvidia doesnt have anything to counter that .... they need a shrink asapQuote:
Originally Posted by RaV[666]
Quote:
Originally Posted by zerazax
exactly ... ;) i expect big things from antiles and cayman
Leaked results from someone on my forums with HD6870 for 3D06 and Vantage it looking very interesting and i think it real not fake for sure, let to see what about 6870 performance :D
http://img828.imageshack.us/img828/8135/0img5366.jpg
http://img831.imageshack.us/img831/6176/3img5392.jpg
http://img257.imageshack.us/img257/2289/g6000.gif
http://img684.imageshack.us/img684/6666/capture3pl.jpg
http://img168.imageshack.us/img168/3834/zumsi.jpg
http://img828.imageshack.us/img828/7785/capture2pl.jpg
http://img233.imageshack.us/img233/9096/0capture4.jpg
Source : Click
is the 128bit a bug with gpuz database, or is it reading it wrong, or is that the actual bitrate?
why afterburner shows max clock speed 775mhz in that graph? is it 6850 or 6870 ?
More performance with less hardware is very interesting.Quote:
Originally Posted by Sn0wm@n
Quote:
Originally Posted by MrMojoZ
yep ... + this will allow them more chips per waffer so pricing should still be interesting
they had the same advantage with evergreen and didnt go for perf/$. maybe they will this time, one can only hope.:rolleyes:Quote:
Originally Posted by Sn0wm@n
How could they when market decided cards were more value than recommended price from AMD.Quote:
Originally Posted by Chumbucket843
Quote:
Originally Posted by Chumbucket843
35% less transistors would surely help that dont you think?
and also they had a buggy tsmc node to work with ... and maybe they worked out the kinc now ....
But then again, bart xt is more or less 25% faster clocked than a 5850, which makes up for most of the difference. I doubt the 6970 will have significantly faster clocks because of its size(900mhz max) vs cypress(850). Basically everything has to come from core improvements and added components.Quote:
Originally Posted by zerazax
Clocks appear to match polygon counts on both cards(5850 and 6870).
If anyone wants to compare to stock HD5870 here is my run at the same CPU clocks (bar 2 extra cores which only influence CPU score):
http://service.futuremark.com/compare?3dmv=2622805
GPU-Z's 128-bit reading is cause GPU-Z is a database. I'm sure once its updated, it will read it right
You're right that clocks make up a big part of it, but it depends on what other things they put into Cayman. If, as rumored, Cayman has more architectural changes, then we may see even more improvement. Cypress after all was pretty inefficient compared to Juniper (we see that Juniper x 2 often beat Cypress handily, despite having effectively the same numbers)Quote:
Originally Posted by tajoh111
They did. That Fermi was 6 months late and a dissapointment was hardly ATI's fault (TSMC's early 40nm woes didn't help supply either). Simple supply vs. demand at work.Quote:
Originally Posted by Chumbucket843
Plus, ATI had no target price point since nothing was there to compete. Since ATI has a target card it wants to attack (the GTX 460), they have a target price point now
agreed, the clockspeed difference makes up alot of the perf difference.Quote:
But then again, bart xt is more or less 25% faster clocked than a 5850, which makes up for most of the difference. I doubt the 6970 will have significantly faster clocks because of its size(900mhz max). Basically everything has to come from core improvements and added components.
as a whole package its pretty good, but what about when both are overclocked (max typical on stock cooling), that could really close in the gap between them.
i think whats best is the potential this has for mobile chips and fusion products.
when mm2 matter on fusion, this can let them pack in more with good yields.
and for mobile stuff they seem to have a better perf per watt
What concerns me, the 6850 heatsink has 0 heatpipes, while the 6870 heatsink has 3?
From HardwareLuxx:
6850
http://i28.photobucket.com/albums/c2...12-950x633.jpg
6870
http://i28.photobucket.com/albums/c2...12-950x633.jpg
doesn't deserve the name 6870.. more like 5890.