The only catch though is that this is not done automatically, the new APIs allow memory stacking but to achieve this game developers will need to specifically optimize games as such.
One of the more interesting features of multiple GPU configuration is how the system itself works. The reason why there is no memory stacking effect is because a complete copy of the game has to be run in each GPU, and frames are alternated between whatever number of cards you have. However, with low level APIs like Mantle having now entered the scene, the developer is granted the coveted ‘to-the-metal’ access to the GPU. That means, that with the proper optimization developers can now surpass the limitation of the old soft architecture and enable buffer stacking.
Hearing a lot of chatter tonight that an 8GB dual-GPU card is "really 4GB." Not true at all. pic.twitter.com/GVXIo608Mo
— Robert Hallock (@Thracks) February 1, 2015
Source: WCCF Tech