Neither Harvard nor Von Neumann are described in a scalable way. Maybe a contentious opinion, but I don’t think either copes with multiple CPU and virtual memory.

https://en.wikipedia.org/wiki/Harvard_architecture