Overview
Unlike the Von Neumann architecture, the Harvard architecture uses separate memory buffers and buses for program instructions and data. This allows the CPU to fetch instructions and access data at the same time.
Advantages
- Increased Throughput: Simultaneous access to instructions and data eliminates the Von Neumann bottleneck.
- Security: Prevents data from being executed as code, mitigating certain types of buffer overflow attacks.
Modern Implementation
Most modern high-performance CPUs use a 'Modified Harvard Architecture,' where they have separate L1 caches for instructions and data but share a common main memory.