I swear, somewhere on this board, a link was posted regarding the manufactuing processes of AMD and Intel. I believe that it was also mentioned that AMD chips are produced in a different manner that Intel's, and that manner allows less-than-perfect chips to be marked at a lower speed and FSB, meaning that they are still able to run, but not at the intended speed. This would also help keep costs down, since instead of scraping parts that are still working (just not at the original speed), they can just remark them and ship them out. Remember, if it's a processer in a box, don't worry, cause if it fails, you get a warrenty that will replace it for three years (AMD or Intel). So, point is, regardless of failure rates, they both have very slim chances of failing once they leave the Fab. Both Intel and AMD need to keep scrap (i.e. failures) to a minimum so that they can keep the costs lower and keep that razor-thin profit margin. As to in Fab scrap rates, I'd say that, at least for the high end chips, it's pretty high, while the chips that have been fabbed for a few months are pretty much automatic passes.