Thailand’s National Science and Technology Development Agency's (NSTDA) upcoming supercomputer will harness hundreds of GPUs, making it the largest public high-performance computing system in Southeast Asia, says Nvidia.
Powered by 704 Nvidia A100 Tensor Core GPUs, the new system will be 30 times faster than the current TARA HPC system. According to information from Nvidia’s product page, the A100 is available in 40GB or 80GB variants and offers up to 294 times higher AI inference performance over traditional CPUs.
The new supercomputer will be hosted at the NSTDA Supercomputer Centre (ThaiSC) to drive research by engineers and computational and data scientists from academia, government, and industry sectors. It is expected to support research projects in areas such as pharmaceuticals, renewable energy, and weather forecasting.
“The new supercomputer at NSTDA will expand and enhance research in Thailand, speeding up the development of breakthroughs that benefit individuals and industries in the country,” said Dennis Ang the senior director of enterprise business for worldwide field operations in the SEA and ANZ region at Nvidia.
“NVIDIA A100 incorporates building blocks across hardware, networking, software, libraries, optimized AI models, and applications to enable extreme performance for AI and HPC,” said Ang.
“We chose NVIDIA A100 because it is currently the leading solution for HPC-AI in the market. Even more important is that many HPC-AI software applications are well supported by NVIDIA technology, and the list will keep growing,” explained Manaschai Kunaseth, chief of operations at ThaiSC.
When operational, the additional power of the new supercomputer will allow users at ThaiSC to scale up existing research projects. Specifically, the new supercomputer will accelerate innovation for Thailand’s efforts with more advanced modeling, simulation, AI, and analytics capabilities.
Kwanchiva Thangthai, of the National Electronics and Computer Technology Center’s Speech and Text Understanding Team, expects to see massive efficiency gains in speech recognition research pipelines. “We can gain competitive performance and provide a free-of-charge Thai speech-to-text service for everyone via AIForThai,” she said.
The supercomputer is expected to commence operation in the second half of 2022.
Image credit: iStockphoto/sdecoret