The start-to-finish time for the first transaction is about 13 clocks, but each bus (request, response, data) is only used by the first transaction for a fraction of those clocks. Pipelineing other transactions reduces bus idle time and results in a greater transaction throughput. This is especially important because not all transactions use the response and data buses, and those buses would be essentially wasted for several clocks if other transactions weren't able to use them.
This comment was marked helpful 0 times.
ruoyul
This reminds me of the washer/dryer example from the first half of the semester.
The start-to-finish time for the first transaction is about 13 clocks, but each bus (request, response, data) is only used by the first transaction for a fraction of those clocks. Pipelineing other transactions reduces bus idle time and results in a greater transaction throughput. This is especially important because not all transactions use the response and data buses, and those buses would be essentially wasted for several clocks if other transactions weren't able to use them.
This comment was marked helpful 0 times.
This reminds me of the washer/dryer example from the first half of the semester.
This comment was marked helpful 0 times.