I think you’re placing too much emphasis on the von Neumann architecture. It’s not essential to anything, and AFAIK was only intended as a short-term engineering decision, pending redesign once we have more experience and understanding. This separation of CPU and memory is now our biggest performance bottleneck; hence why we shift so much on to GPUs, DSPs, etc. which aren’t von Neumann.
I think you’re placing too much emphasis on the von Neumann architecture. It’s
not essential to anything, and AFAIK was only intended as a short-term engineering
decision, pending redesign once we have more experience and understanding. This
separation of CPU and memory is now our biggest performance bottleneck; hence
why we shift so much on to GPUs, DSPs, etc. which aren’t von Neumann.