Manufacturing News 23/05/2024, 15:09

The first DGX H100 servers of the AI Factory

NDO - Recently, FPT has imported the first DGX H100 servers system to the country, initially realizing a strategic cooperation project to promote the development of AI in Vietnam with NVIDIA to build an Artificial Intelligence Factory (AI Factory) worth 200 million USD.

The first DGX H100 servers of the AI Factory

Previously, at the end of April, FPT signed a comprehensive cooperation contract with NVIDIA to develop an AI solution ecosystem. According to the MoU, FPT plans to invest 200 million USD to build an AI Factory to provide a cloud computing platform for AI research and development and sovereignty in Vietnam. FPT's import of the DGX H100 server system is to gradually implement the memorandum of cooperation signed with NVIDIA, to develop the sovereign AI and Cloud ecosystem in Vietnam.

This is the first system import to ensure the progress of technology integration, build supercomputers for artificial intelligence development, and step by step implement the plan to build an AI Factory.

The remaining series of DGX-H100 systems will be brought to Vietnam shortly to complete technology, accelerate product research and development, and create a premise to build new generation AI and Cloud solutions and services.

Cooperating with NVIDIA in the project to build an AI Factory in Vietnam, FPT will provide GPU cloud services to help FPT's business customers globally access the most core resources to improve research capacity, accelerate the speed of AI application, especially Generative AI, achieve breakthroughs in improving productivity and customer experience.

Những máy chủ DGX H100 đầu tiên của Nhà máy trí tuệ nhân tạo ảnh 2

Close-up of the AI machine DGX-H100 licensed for import in Vietnam.

Each DGX H100 server contains 8 H100 GPUs connected via NVIDIA NVLink, combined with dual Intel Xeon Platinum 8480C chips, a total of 112 cores, and 2TB system memory for the most intensive AI jobs. Besides, there are two NVIDIA BlueField-3 DPUs to reduce load, accelerate and isolate network services, storage, and advanced security.

The DGX H100 possesses 6x higher performance and 2x faster network connection than the previous generation server, meeting the huge computational requirements with trillions of parameters of large language models (LLM), recommendation systems, medical research, and the field of climate science.

DGX H100 helps build a refined, systematized, and scalable artificial intelligence platform to help businesses process language, and analyze data, etc. This is also the foundation system that makes up Eos - NVIDIA's most powerful supercomputer, ranking 9th among the fastest supercomputers in the world. NVIDIA confirms: that DGX H100 can push the boundaries of technology and AI infrastructure.

The DGX H100 server is being used by corporations and technology companies around the world. Tesla develops self-driving car products on the H100 GPU platform. In Japan, start-up Xeureka applies DGX-H100 in the process of research and development of drugs, improving medical examination and treatment activities and treatment of diseases. In Ecuador, and Sweden, a series of financial, healthcare, legal, IT, and telecommunications projects apply AI on the DGX-H100 platform.


Sign up for more information