There is rarely acknowledged connection between yield and capacity in the sense that in many cases the most direct way to fix a yield problem is to restrict certain processes to go through a limited subset of available equipment. I know that we resorted to this technique many times in my past doing yield enhancement. By artificially constraining the toolset or dedicating specific pairings of tools, you get less variability in the mfg parameters and get more predictable yield. You hide the yield problem in a capacity problem.
An interesting point for this discussion: AMD paid dearly for the right to have 28-nm products manufactured at other foundries, yet 28-nm capacity is constrained. I assume that before AMD took that step they made sure they could get the capacity they needed.
Better get those 28-nm capacity, if you're going to hit revenue goals. I wonder what's the reason for the shortage? High in demand?
Member of the Patexia Semiconductor Group (http://bit.ly/HeyQJO)
Blog Doing Math in FPGAs Tom Burke 13 comments For a recent project, I explored doing "real" (that is, non-integer) math on a Spartan 3 FPGA. FPGAs, by their nature, do integer math. That is, there's no floating-point ...