Menu Close

NSF recipient Alex Schaefer shares his doctoral research focus

NSF recipient Alex Schaefer’s primary research focus centers on technology-optimized memory architecture and accelerator co-design that exploits those differences to expand the design space across the compute stack. As a second year PhD student in Professor Robert Radway’s group, Schaefer hopes to apply his background in circuit design and computer architecture to the integration of emerging memory technologies, which address system-level demands under a different cost structure than conventional memories.

As Schaefer explains, the research entails creating hardware accelerators tailored for specific problems, co-designing them alongside non-traditional memories built on emerging device technologies. Memory walls represent one of the problems this research addresses. For example, “computing systems suffer from an energy and performance bottleneck caused by transferring data between off-chip dynamic random access memory (DRAM), static random access memory (SRAM),” Schaefer says, adding, “A lot of time and energy is spent transferring data from DRAM to SRAM. So off-chip to on-chip.”

The role of emerging technologies such as RRAM (Resistive RAM) serve as a potential solution to the above problems, Schaefer continues. RRAM is denser than SRAM and can be “fabricated above the silicon substrate used for logic circuits.”

The above process directly correlates to manufacturing advantages, Schaefer explains. Standard DRAM, for example, calls for distinctly different manufacturing processes at separate fabs. When integrating the emerging memories above, the logic substrate removes the need for separate fabs, reducing the off-chip data transfer delay and energy, Schaefer emphasizes.

Schaefer’s prior work as a doctoral student includes developing a software tool for design space exploration. In doing so, Schaefer addressed tool deficiencies. Existing exploration tools were expansions of traditional SRAM and DRAM exploration tools. 

 Applying these older models to emerging memories produced inaccuracies,” Schaefer explains. To address the inaccuracies, Schaefer helped build, “specialized models for resistive memories to improve modeling accuracy, allowing researchers to effectively compare performance across different memory devices and capacities.”

When Schaefer isn’t working, he enjoys staying physically active by swimming and playing basketball with members of his research lab. In addition, Schaefer, who loves music, collects vinyl records. He played drums in the University of Virginia’s Jazz band, where he completed his undergraduate degree. “I haven’t found a place for my drums in Philadelphia, but I play both acoustic and electric guitar.” In fact, Schaefer is currently teaching his fiancée how to play.