How do these 3D architectures scale regarding yield? (The naïve expectation would be that adding vertical layers should have exponential impact on yield.) Is this commercially viable, near term?