What you are describing is the calculation of a sample size to compare two proportions. What you meant to ask is for a given difference in proportions what is the sample size needed to be sure that if the difference is significant you will be at least 80% certain that this significant difference is real
To answer this you will have to have an estimate of the proportions. If the old design produces 20% failures and you think the new design will produce 15% failures then these become the proportions you will use for the sample size calculation.
.
