I don't imagine a shared memory pool is going to change anything other than reducing the cost of a card. And that cannot be discounted considering Nvidia (as far as we know) would still have to duplicate memory for an X2/GX2 style card.
When using AFR, each GPU is still going to have to independently render its frame. And it doesn't automagically allow easy use of other methods of multi-GPU rendering.
I'm with jawed on this. For ATI this is a HUGE from a cost standoint and a possible jumping off point for further multi-GPU integration.
From a consumer standpoint it'll allow for cheaper cards than the Nvidia alternatives, unless Nvidia has a monolithic GPU that is at or near the same speed as an R700 X2 type of card.
For the enthusiast who will spend oodles of cash for the fastest there is. There really aren't any benefits for them over and above the current way multi-GPU on a card is done, IMO.
Unless ATI is being especially coy when they made comments a few weeks back that multi-GPU rendering will still be basically the same with some "tweaks." Here, I'm assuming the major tweak would be shared memory pool.
In other words until I see differently. I like that there is progress being made. But I'm still not going to be buying a multi-GPU solution if it relies on AFR as the primary rendering scheme.
Regards,
SB