Memory to Memory Architecture

• 
• For all vector operation, operands are fetched directly from main memory, then routed to the functional unit
• Results are written back to main memory
• Large startup time