you can notice how it struggles on high resolutions, almost all kepler gpu's have that problem (against AMD)
It struggles because of the 128bit wide bus (see how much extra it can give with the rams above 1800Mh), the 768 Kelper shaders are not the bottlenecks in this story. I already stated in some other threads, that it's simply not fair to bench these low-mid 128bit cards with the 4XAA utlra detail @ high resolutions, but I also understand that Wizzard wants to stay consistent between different generations. Perhaps including one 2XAA+4XAF medium resolution (900p, 1050p or maybe a 1080p) bench could prove to be more useful with cards like this.
The main reason behind that Nvidia "struggles" on ultra-high res is also mostly due to the 256 bit wide bus on the 660-680, but they were able to stop pushing things more and only release the mid-high GK104 because it was enough to compete against AMD that time.
nvidia make kepler the simplest they could, that way gpus have lower power consumption and don't loose potential on normal resolutions. it's an intelligent way to make a gpu, but in long terms, people who bought a 600 series may be dissapointed. time will tell. i don't see this gpus like a gt9500 or 9800gt that even today are running some the latest games (wich do not require dx11)
I have to disagree, Kepler is built for gaming and it's performing beautifully in that regard, so let's see what the Titan will be capable of, then we will know for sure finally.
.................
I only asked my question because I did measure difference with earlier cards sporting extra memory (even if it was not that big), and I'm really curious what Nvida is doing which results in no benefit anymore with more Ram. Bf3 or Skyrim is clearly able to load up and use more than 1GB memory, yet we don't see any change in resolutions like 1200p or below. Is it some kind of a new memory management in the drivers or some new solution in the chip itself? I'm curious and I would like to know