Sure... but how much of an advantage is this going to be for gaming ?
Small, but there, and repeatably measurable. Usually, you'll find it empirically documented in memory reviews.
Transfers of data from video card to/from system RAM take time on the CPU's memory controller. As video cards are added, and used, this load increases. Once there isn't enough time to spare per memory channel for the CPU, or one or more GPUs, performance will be lower than if there were more channels/bandwidth to spare, due to the hardware waiting on the memory.
The bridges on the cards don't have enough bandwidth to handle this sort of thing, and are used mainly for command/control duties. Multi-GPU users have always been a small enough part of the market to not make it worthwhile to R&D high quality all-hardware GPU SMP support.
The added L3 cache and CPU cores allows the drivers/OS a bit of breathing room, in terms of CPU time, which by itself can increase performance a little, compared to 4 cores and 6MB or 8MB (the driver is doing rendering setup for 2+ GPUs, after all, not just 1). But then, the added RAM channels/bandwidth also help alleviate some of the waiting the CPU and GPUs may be forced to do, as more time is spent shuffling data about different memories (CPU 1, video card 1, video card 2, video card 3...), compared to at most 2 (CPU 1, video card 1). It varies, anywhere from negligible to maybe 15% (I don't recall seeing 15% in a review, but have seen over 10%, so it's probable, IMO, with the right game, GPU, and monitor config). Not worth it at all for $1200-1500 typical builds where the person is trying to go crazy on a budget, since better LGA1150 OC gear, or a better video card model, will surely make up that difference, and possibly more. But, if you have the money, and the intent to spend it, it can be a potential performance gain...