Counting the DRAM row buffer as a separate cache layer is interesting because it has meaningful performance impact, but mainly because existence of such an thin

Counting the DRAM row buffer as a separate cache layer is interesting because it has meaningful performance impact, but mainly because existence of such an thing is good counter argument to people who think that DRAM chips have RAS/CAS multiplexed address pins only to save package pins.