According to a brief blog post, Google promises that the reduced pause times, which drop from a peak of 272 ms to just 50 ms, will result in notably better performing web apps, such as the WebGL versions of Google Maps or browser games. Subjectively, this writer found the WebGL version of Maps to be performing slightly better with a Chrome 17 nightly build rather than the stable Chrome 15, but I would not necessarily pin those marginal and very subjective impressions of improvements down on the change of the GC just yet. I am not convinced that the effect is as substantial as Google's data suggests
However, for those of us who believe benchmarks, there is a "stress test" that measures the effect of the new GC. Chrome 15 finished the (10 second) test with a score of 6, and Chrome 17 with a score of 34 on my test system. However, Chrome was beaten by the current Opera 11 stable build with 46, by the nightly build of Firefox 11 with 48 and, most dramatically, by the current Firefox 8 stable build with 126, which was able to paint more than twice the number of frames than Chrome 15 did - 617 frames versus 278 frames.
You can download the current developer version of Chrome here.