Transferring data in blocks from the main memory to the cache memory enables an interleaved main memory unit to operate at its maximum speed. True/False. Explain.

GATE 1990 · Computer Organization and Architecture · Cache Memory · medium

Answer: TRUE — Block transfers cause the cache controller to issue a burst of consecutive-address requests that map to successive interleaved banks, keeping all banks busy and allowing the interleaved main memory to operate at its maximum (pipelined) throughput.

  1. How interleaved memory achieves peak bandwidth: In a k-way interleaved memory, consecutive address words reside in banks 0, 1, ..., k-1 cyclically. A memory controller can issue a new read to bank i while bank i-1 is still completing its access. After an initial latency of one full memory cycle (t_m), one new word arrives every t_m/k time units — the maximum possible rate.
  2. Why block transfer enables this maximum speed: When the cache controller requests a full block (say 8 words starting at address A), it issues reads to addresses A, A+1, A+2, ... in rapid succession. These addresses map to successive banks, so the interleaved memory can pipeline all of them. The entire block is delivered at near-maximum memory bandwidth. If only one word at a time were fetched (non-block), the interleave advantage would be lost for each separate access.