Bitmain-supported Tsinghua Team Defends Championship in ASC 2018 Student Supercomputer Challenge
On May 9, the ASC 2018 Student Supercomputer Challenge (ASC18) final ended at Nanchang University, with the victory of the defending champion, Tsinghua University. As a high performance computing chip producer, Bitmain has been in close partnership with the winning team since 2016, offering it financial and hardware support.
The ASC was launched in 2012. Over the past 7 years, this annual competition has attracted over 5,500 college participants and grown into the world’s largest and most participated high performance computing (HPC) competition for college students, ranking among the top three HPC competitions, along with Student Cluster Competitions of ISC High Performance in Germany and SC in the US.
This year’s champion is an experienced team with the right mix of class levels from the Department of Computer Science and Technology, Tsinghua University. In preparation, the team invited teachers and students specialized in natural language processing for discussions over the MS MARCO challenge. They examined each application in detail and used performance testing and bottleneck analysis to identify performance bottlenecks and explore optimization methods. Based on the requirements and devices from the ASC Organizing Committee, the team devised a system of six servers collectively equipped with 16 GPUs. The design stood out in all the six applications, winning the team the championship.
The team members included LI Beichen (leader), FENG Guanyu, XIE Lei, YU Jiping, CHEN Yu, WANG Miao, LIU Jiachang and TANG Shizhi, with ZHAI Jidong and HAN Wentao as advisors.Bitmain is a leading HPC chip company, specialized in products for blockchain and AI. We dominate the international blockchain industry with a market share of roughly 80%, selling about 50% of our products abroad. Our core competency lies in high performance chip design, as we have the world-leading 7nm chip production capacities. Bitmain started exploring the field of AI in late 2015, and in 2017, launched its first AI Chip, BM1680. The acceleration chip is dedicated to cloud deep learning, the first of its kind available for sale in the world. This is followed by the second-generation chip BM1682, released in March 2018. BM1682 has five times the performance of its predecessor, well ahead of similar products worldwide. These AI products can be used in various fields, ranging from security monitoring and data centers, to supercomputing and robotics.