If you use Kanotix you can even use kernel 3.9 with wheezy ;)
...then it should use benchmarks that measure latency. For example, try copying a 4 gigabyte file and start firefox at the same time, and measure how long it needs before firefox is fully usable.
What about measuring things that matter for a desktop and gaming machine such as worst frame performance and input-to-response latency, especially worst-case latency?
There's not much that is more annoying than typing in a text editor or word processor and having everything freeze solid for two seconds because a large disk job started running in the background.
Venture to say that I think ... my kernels - particularly i7 [for Sandy and Ivy Bridge] and Brazos [for AMD Athlon II from up] line will be faster, and far more responsive - especially under a load of 100% for each thread / core processor - than Steven kernels. Of course, it must be emphasized that Steven kernels are really good - btw I in recent years do not use BFS and BFQ.