Ask HN: What is the fastest sequential FLOPS?

I perform molecular dynamics (MD) simulations using a large number of parallel processor cores. As more atoms are added to a system, more cores can be added to handle them. Computational power is still mostly following Moore's law, so in the next few years, vastly larger systems will be capable of being simulated.

However, one problem is that the speed of an individual processor core hasn't improved very much in a while. This poses a scientific problem in that it becomes very difficult to simulate systems past a few microseconds. Special purpose computers such as Anton [ http://en.wikipedia.org/wiki/Anton_(computer) ] have been built using highly-specialized ASICs that can reach into the millisecond timescales, but I'm afraid scientific discovery in the seconds range (for MD) is not going to ever happen unless we either 1) suddenly discover a new technology for increasing transistor speed or 2) develop ground-breaking simulation algorithms.

My question is: what is the fastest sequential FLOPS that a general purpose, commodity processor can currently achieve?