In running ALLPATHS-LG, a short read genome assembler, on a Dell PowerEdge R920 with four Xeon E7-4860v2 processors running at 2.60GHz, I did one run with hyperthreading on and one with hyperthreading off. This was mainly to see if the increased cache size on the physical CPU core with hyperthreading off had a positive effect on performance.
I'll take all the anticipation away: Nope. Hyperthreading wins.