You’re going to have to try a real scene that takes more work to render if you want to see a real comparison. A test that runs in a couple seconds is not going to tell you anything interesting.
fishy_cat for example is generally around twice as fast in 2.8.
Interestingly it’s 8% faster using 32x32 GPU tiles in 2.80 than using 128x128 tiles, even before enabling CPU+GPU rendering which speeds it up even more.
It’s not that easy to find good benchmark scenes that are compatible with both 2.79 and the current 2.80 beta and some of the old benchmark scenes are currently broken in 2.80.
Yes, don’t worry about it. I didn’t do any real benchmark, yet, but I actually have the impression that 2.8 is faster. At least it feels that way.
Otherwise, not surprising that there are corner cases where it isn’t true. Most real-world optimizations don’t make things always faster, only for typical use cases, while for exotic ones they may be slower.