Wow, this back and forth hashing battle between Bminer and NebuTech has resulted in gpu hashing peformance more than doubling (close to triple?) in just several months for same hardware as before.  I do not recall any other miner on another algo that has resulted in such significant performance gain just due solely to software code.

This leaves one to contemplate how terribly inefficient the original code must have been possibly?  Or were devs intentionally holding back code improvements for later to allow for miner upgrades?