2019 ACM Gordon Bell Prize Awarded to Eth Zurich Team for Developing Simulation That Maps Heat in Transistors
December 03, 2019
Team was awarded for their new framework for simulating the transport of electrical signals through nanoscale materials (such as the silicon atoms used in transistors).
ACM, the Association for Computing Machinery, named a six-member team from the Swiss Federal Institute of Technology (ETH) Zurich recipients of the 2019 ACM Gordon Bell Prize for their project, "A Data-Centric Approach to Extreme-Scale Ab initio Dissipative Quantum Transport Simulations."
The ACM Gordon Bell Prize tracks the progress of parallel computing and rewards innovation in applying high performance computing to challenges in science, engineering, and large-scale data analytics.
The ETH Zurich Team wrote a variation of OMEN that is Data Centric (DaCe OMEN). The ETH Zurich team introduced DaCe OMEN, a new framework for simulating the transport of electrical signals through nanoscale materials (such as the silicon atoms used in transistors). To better understand the thermal properties of transistors, the team simulated how electricity would be transported through a two-dimensional slice of a transistor consisting of 10,0000 atoms. The ETH Zurich researchers simulated the 10,000-atom system 14 times faster than an earlier framework that was used for a 1,000- atom system. The DaCe OMEN code they developed for the simulation has been run on two top-6 hybrid supercomputers, reaching a sustained performance of 85.45 Pflop/s on 4,560 nodes of Summit (42.55% of the peak) in double precision, and 90.89 Pflop/s in mixed precision.
The ETH Zurich researchers estimate that cooling can consume up to 40% of the total electricity needed for data centers, amounting to cumulative costs of many billions of dollars per year. The team used their simulation to develop a map of where heat is produced in a single transistor, how it is generated, and how it is evacuated.
The ETH Zurich team also built a graphical interface for the DaCe OMEN framework that includes a visualization of dataflow in lieu of a simple textual description. Anyone running the code can use the image representation to interact with the data directly.
Winning team members include Alexandros Nikolaos Ziogas, Tal Ben-Nun, Timo Schneider and Torsten Hoefler, from ETH Zurich's Scalable Parallel Computing Laboratory, as well as Guillermo Indalecio Fernández and Mathieu Luisier from ETH Zurich's Integrated Systems Laboratory.