King Abdullah University of Science and Technology (KAUST) and Cerebras Systems have joined forces to revolutionise seismic processing, leveraging the power of the Condor Galaxy AI supercomputer. Their achievements have earned them recognition as finalists for the 2023 Gordon Bell Prize.
The Gordon Bell Prize
The Gordon Bell Prize, named after the computer science and high-performance computing pioneer Gordon Bell, celebrates accomplishments in the field of high-performance computing (HPC). Now in its 35th year, this award recognises the contributions of KAUST and Cerebras Systems.
KAUST and Cerebras Systems’ research focuses on multi-dimensional seismic processing, a critical discipline for sustainable resource management and a low-carbon future. Their collaboration harnessed Cerebras Systems’ advanced technology to achieve seismic processing speeds and accuracy previously considered unattainable.
A crucial element is the development of a Tile Low-Rank Matrix-Vector Multiplication (TLR-MVM) kernel, tailored to exploit the advanced architecture of Cerebras CS-2 systems within the Condor Galaxy AI supercomputer. This resulted in seismic processing with production-grade accuracy and a record-breaking sustained memory bandwidth of 92.58 petabytes per second (PB/s), overcoming a traditional bottleneck in processor-rich architectures.
Andrew Feldman, the co-founder and CEO of Cerebras Systems, expressed his pride in this collaboration, emphasising its potential to bring about significant advancements across various fields, including climate modelling, computational astronomy, wireless communication, and seismic imaging.
KAUST’s research and collaborative efforts
Lead author Hatem Ltaief, Principal Research Scientist of KAUST’s Extreme Computing Research Centre (ECRC), highlighted the importance of fine-grained algorithmic innovations in overcoming memory challenges and unlocking the full potential of hardware.
ECRC Director David Keyes underscored the versatility of wafer-scale hardware, showcasing its potential beyond neural network training, and highlighting its significance in the history of Gordon Bell Prize winners.
KAUST’s collaborative efforts also include Professor Matteo Ravasi of Earth and Environmental Sciences and 2023 KAUST Computer Science PhD graduate Yuxi Hong, who has applied his high-performance computing skills to Lawrence Berkeley National Laboratory’s Exascale Computing Program.
