- The Intel 386 (1985) was the first 32-bit x86 processor with a 16-byte instruction prefetch queue to improve performance by fetching instructions before they are needed.
- The prefetch queue circuitry includes an incrementer using Manchester carry chain and carry skip for fast pointer updates, and a limit check circuit using dynamic logic to stop prefetching at the code segment boundary.
- An alignment network rotates data to support unaligned 32-bit memory accesses, and a sign extension circuit converts 8-bit or 16-bit signed values to 32 bits.
- The prefetch unit contains approximately 7,400 transistors, less than 3% of the entire 386, highlighting the complexity required for backward compatibility and unaligned memory support.
- Instructions flow from the Bus Interface Unit to the prefetch queue, then opcodes go to the decoder via an 8-bit bus while data bytes travel via a 32-bit datapath.