Vietnam Welcomes Supercomputers To Boost AI Development | Scientific technology

Viettel’s supercomputer system

Recently, Viettel Cyber ​​Space Center (member of the Viettel group) operated its high performance computing system in its database center. This system can run 20 petaFLOPS, allowing Cyber ​​Space Center to run various complex training models simultaneously.

Viettel’s supercomputer uses computational acceleration technologies, network switches and synchronous software similar to those of the world’s top 500 supercomputers.

Viettel Group Deputy Managing Director Nguyen Dinh Chien said his organization is making a strong investment in AI to fulfill its role as a key force in establishing a digital society. This investment allows Viettel to easily attract talent and AI experts ready to take on important AI missions from the nation. It is also a tool to make Viettel a global pioneer in the implementation of AI solutions in life and to positively increase overall turnover.

Likewise, VinAI Research (member of VinGroup) just installed Southeast Asia’s most powerful artificial intelligence supercomputer system named NVIDIA DGX SuperPOD. This supercomputer, 10 times faster than the current NVIDIA DGX A100 system owned by Vin AI, will help in the development of self-propelled electric vehicles and medical care devices not only for national but also international demands.

VinAI Research Director Bui Hai Hung expressed hope that in May a cluster of 20 new DGX A100 systems will be connected to NVIDIA Mellanox HDR, an InfiniBand network at the speed of 200 GB / second. The emergence of DGX SuperPOD is said to provide an enabling environment to nurture world-class Vietnamese AI talent.

Since 2018, FPT Group has been one of the pioneers in Vietnam to use supercomputers in its AI research, focusing on language and image processing tasks. At this point, the DGX-1 was NVIDIA’s last supercomputer, and FPT Group was the first in the country to have such a powerful machine. Thanks to this, FPT Group was able to save a lot of costs and efforts in the realization of its projects, accelerating its research and creating useful products.

The FPT Group Board of Directors highlighted the critical role of supercomputers in AI research, saying they can speed up the research process 96 times compared to using the deep learning model on the traditional CPU machine. This means that a task that typically takes a month to complete can now be completed in one to two days, which is blazingly fast.

By Ba Tan, Tran Binh – Translated by Vien Hong


Source link

About Mariel Baker

Check Also

Getting up to date on the proton – sciencedaily

Scientists are developing a revolutionary theory to calculate what happens inside a proton traveling at …