I'm guessing you meant he GDDR5 memory in the PS4 right? Or the eSRAM. Surely those are ahead of PC in terms of bandwidth but those alone don't mean that they are ahead. The processing power of the CPU is far behind what we can have in PC's (SC2 still uses CPU for the most intense part afaik), also the SSD's (be it SATA 3 or PCI-E) and many more factors to consider.
I'm talking about what AMD calls hUMA. Kaveri will be a weak incarnation of that (early reports is somewhere around an HD 7730 or 7750). It's orthogonal to GDDR5 (you can do hUMA with whatever memory), although it's the kind of thing that scales well with bandwidth so that's why PS4 went with super fast memory and Xbox One has that 32MB of super fast memory as well.
I'm aware the consoles will be largely inferior in raw numbers, my point is that a fully shared memory design means that offloading computations to the GPU is way easier and more efficient (no copying data back and forth to dedicated VRAM, CPU can get direct pointers in GPU memory and vice-versa), this means that console games can be optimized in ways that aren't feasible on PC right now. Even on the current-gen the raw numbers on PC vs the difference in quality vs consoles in games was way off, I think the difference will be even more pronounced on next-gen.