It's likely "real". By real, I mean that I just ran JSBench on my MacBook Air with Safari and Chrome (latest released version of each) and Safari came out significantly ahead (Chrome took 1.75x longer). They're claiming 2.53x faster than Chrome in that slide so that would mean an improvement of probably around 50% over the current Safari in terms of JS speed on this benchmark. So, it would seem that this particular benchmark already favors Safari's JS engine and that the next Safari will improve on the current Safari's performance.

Now, the issue with JSBench is that it's real-world JS, but only from 5 sources. Granted, they're 5 sources that are very high-trafficked, but it's still a small sample size (http://jsbench.cs.purdue.edu/).

Similarly, Apple showed off SunSpider benchmarks showing the next Safari being faster than Chrome. The current Safari on my MacBook Air beats Chrome, with Chrome taking 6.9% longer. So, part of it is that each browser maker has its preferred benchmarks.