I ran Folding@Home at Google on hundreds of thousands of fast Xeon cores for over a year. I concluded at the end that unbiased MD simulations are not an effective use of computer time.