A wafer is cut into many small chips. Tiny flaws land at random, and any chip that gets hit is thrown away. Small chips dodge the flaws much better than big ones, which is why young materials like diamond start with small chips.

With D₀ killer defects per unit area, the chance that a die of area A has none is e^(−AD₀) (Poisson). Murphy's model averages over wafer-to-wafer variation of D₀ and is less pessimistic for large dies. Good dies per wafer = dies per wafer × yield: yield falls exponentially with area.

Dies per wafer uses the standard edge-loss formula; the Monte Carlo drops uniform point defects on the usable disk (3 mm edge exclusion) and discards any die hit, which is exactly the Poisson model, and the test suite checks the two agree. Clustering (negative-binomial) raises large-die yield further [stapper1983]. Diamond defect densities are not yet published at production scale; chapter 2 and E0 discuss the evidence.