“Using Spin-Hall MTJs to Build an Energy-Efficient In-memory Computation Platform,” a paper that is the product of research conducted in ECE, has received the best paper award at the 2019 International Symposium on Quality Electronic Design (ISQED).
The paper is an outcome of collaborative research conducted by multiple ECE research teams led by professors Ulya Karpuzcu, Jian-Ping Wang, and Sachin Sapatnekar.
Through their join efforts the teams have created a specialized in-memory computation system called SHE-CRAM, a computational RAM using spin-Hall-effect-based nanomagnets, that performs efficient data processing within the memory.
In conventional computing, data is brought from the memory to a CPU for processing, or through near-memory processing, to a computational unit at the edge of the memory. Both of these approaches are prohibitively expensive for big-data applications due to the high cost of transporting data outside the memory.
The creation of SHE-CRAM is the coming together of novel spintronics-based spin-Hall storage technologies, memory circuits, computation algorithms, and memory architectures. It has been demonstrated that SHE-CRAM is over 2000x faster and at least 130x more energy-efficient than state-of-the-art near-memory processing for applications such as 2-D convolution and neural computation.
The paper authors are Masoud Zabihi, Zhengyang Zhao, Zamshed Chowdhury, Michael Resch, Thomas Peterson, Mahendra DC, Jian-Ping Wang, Ulya Karpuzcu, and Sachin Sapatnekar. The best paper award will be presented on March 6, at the ISQED luncheon.