Harvard Architecture

Harvard Architectures

• The Harvard architecture is a computer architecture that separates the storage and h(pathways for program instructions and data. Unlike the Von Neumann architecture, where instructions and data share the same memory and bus, the Harvard model uses two separate memories and two separate buses :

■ Instruction (Program) Memory (for program code)

■ Data Memory (for data storage)

• This separation allows the CPU to fetch an instruction and access data simultaneously, improving speed and performance.

Features of Harvard architecture

1. Separate memories :

■ Instructions are stored in a dedicated memory.

■ Data is stored in a separate memory unit.

2. Separate buses :

■ The instruction bus carries program code to the CPU.

■ The data bus carries data between the CPU and the memory.

3. Parallelism :

■ The CPU can fetch the next instruction while simultaneously reading or writing data.

■ This reduces bottlenecks (the "Von Neumann bottleneck").

4. High performance :

■ Since instruction fetch and data access do not compete for the same bus, the system is faster and more efficient.

Advantages of Harvard architecture

• Faster execution due to simultaneous instruction and data access.

• No memory contention since separate buses are used.

• Improved throughput, making it suitable for digital signal processing (DSP) and embedded systems.

Disadvantages of Harvard architecture

• More complex design as it requires separate memories and buses.

• Higher cost compared to Von Neumann systems.

• Less flexible, because program code and data are stored separately and cannot be easily interchanged.

Comparison between Harvard and Von Neumann architectures

Review Questions

1. Define Harvard architecture. How does it differ from Von Neumann architecture ?

2. Explain the role of separate instruction memory and data memory in the Harvard architecture. State two advantages and two disadvantages of the Harvard architecture.

3. Give a comparison between Harvard and Von Neumann architectures.