Panel 3: The Path from HPC to AI in Taiwan's NCHC
Authors: Ce-Kuen Shieh (National Center for High-Performance Computing, Taiwan)
Abstract: Artificial Intelligence (AI) has been regarded as one of the competitive edges of economy. The acceleration of compute cycles, which is often regarded as high-performance computing, is traditionally considered as the strength of scientific advance in research. It is now turned into the powerhouse of AI. NCHC was commissioned by Taiwan government to develop a new generation of acceleration platform, namely Taiwan Cloud Computing (TWCC). TWCC cleverly integrates 2016 Tesla V100 GPUs into a HPC system and, by the end of this year, the system will be ready with capacity of more than 8 PF performance and 50 PB storage. The energy awareness of the system built will lead to an exceptional power usage effectiveness of 1.2. This significant development will support and bridge Taiwan’s academics and industry as the compute backend of AI innovation. By plug-in into the cloud infrastructure, NCHC is able to upscale AI with HPC and to benefit the two worlds of science and economics and satisfy both scientific query and real life needs. In this presentation, I will give an overview of Taiwan AI strategy and the role of NCHC play in the strategy. Further, I will give you how we develop TWCC to support the Taiwan’s eco-system of AI, which includes stakeholders of general users, domain experts, system developers and operators and conclude in our future plan for the next 3-year horizon.
Back to 2nd ATIP Workshop on International Next-Generation Computing Programs and Workforce Development Archive Listing
Back to Full Workshop Archive Listing