AT's article:
"In the case of memory allocations between 3.5GB and 4GB, what happens is unfortunately less-than-deterministic. The use of heuristics to determine which resources to allocate to which memory segment, though the correct solution in this case, means that the real world performance impact is going to vary on a game-by-game basis. If NVIDIA’s heuristics and driver team do their job correctly, then the performance impact versus a theoretical single-segment 4GB card should only be a few percent."
"The worst case scenario on the other hand would be to have the NVIDIA heuristics fail, or alternatively ending up with a workload where no great solution exists, and over 3.5GB of resources must be repeatedly and heavily accessed**. In this case there is certainly the potential for performance to crumple, especially if accessing resources in the slow segment is a blocking action."
I hope you guys are confident NV's driver team will not leave Maxwell in the dust once Pascal is out, else your SLI setups will be unable to handle games with 4k textures. You'd also hope more games don't fall into that "worse case" scenario in pushing the limits of vram. With more console ports and 4-6gb vram for consoles to abuse.. it will be a damn shame if you can't run 970 SLI at 1080p and enjoy ultra textures or AA.
NV's excuse doesn't fly. No company would not know what product they are developing, selling, especially when its only a few key products. I'm gonna call stupid on whoever decided to try and sell the 970 with full ROPs & 4gb. Stupid as in they think they'll get away with it.
** Ultra texture modded Skyrim, Arma 3 and poor frame time in Shadow of Mordor at 1080p with maxed textures. What else? Don't know cos review sites don't bother to investigate further. That bolded statement is an admission that in some games, your 3.5gb + 0.5gb 970 will tank. Just hope its not the games you play.
Read it again, slowly to comprehend, for those who are still defending NVs actions:
"The worst case scenario on the other hand would be to have the NVIDIA heuristics fail, or alternatively ending up with a workload where no great solution exists, and over 3.5GB of resources must be repeatedly and heavily accessed**. In this case there is certainly the potential for performance to crumple, especially if accessing resources in the slow segment is a blocking action."