Won't happen in the next 5 years. Might not happen in the next 10 either. It's just too hard to justify using Ray Tracing in real time. A basic RT setup will get you reflections, refractions, hard shadows, and maybe funny shaped camera lenses. Do those effects matter that much? Certainly not enough to justify the huge engineering effort and performance hit they'll come with. And not when they can be faked pretty well with raster graphics in most cases. If you want more advanced effects like accurate depth-of-field, soft shadows, and global illumination you need to do distribution or monte-carlo ray tracing, which is so vastly expensive it won't even be possible with hardware acceleration. Rasterized graphics can fake this OK, and do it at a reasonable speed, so it's not likely that RT will make much sense in the next decade.
Ray Tracing does have the advantage that its cost is not very sensitive to geometric complexity. But its cost is very, very sensitive to the number of pixels and degree of anti-aliasing used. For that reason maybe RT will emerge as a real time method once we stop demanding higher and higher resolutions in our devices (because we've surpassed the eye already) and can no longer get improved image quality out of raster graphics.
Source: Me, I'm a Grad student in Computer Graphics.
