Efficient Design and Communication for 3D Stacked Dynamic Memory

Abstract

As computer memory increases in size and processors continue to get faster, memory becomes an increasing bottleneck to system performance. To mitigate the slow DRAM memory chip speeds, a new generation of 3D stacked DRAM will allow lower power consumption and higher bandwidth. To communicate between these chips, this paper proposes the use of ring based standing wave oscillators for fast data transfer. With a fast clocking scheme, multiple channels can share the same bus to reduce TSVs and maintain similar memory latencies. Simulations with the new clocking scheme and data transfers are performed to show the improvements that can be made in memory communication. Experimental results show that a ring based clocking scheme can obtain two times the speed up of current stacked memory chips. Variations of this clocking scheme can also provide half the power consumption with comparable speeds. These ring-based architectures allow higher memory speeds without compromising the complexity of the hardware. This allows the ring-based memory architecture to trade off power, throughput, and latency to improve system performance for different applications.