In a bid to broaden its AI research and development initiative, Japanese multinational information technology and electronics company NEC has revealed today that it has teamed up with Supermicro to build Japan’s largest supercomputer for advanced AI research.
SuperMicro, an American-based information technology company, announced today that NEC had selected more than 116 Supermicro GPU servers to build the supercomputer. The system is expected to be powered by two 3rd Gen Intel Xeon Scalable processors, each with eight NVIDIA A100 80GB GPUs, to satisfy the needs of expanding AI training requirements.
The SYS-420GP-TNAR server, which is tailored for large-scale distributed learning applications, is part of Supermicro’s new system. Each server has eight NVIDIA A100 80GB Tensor Core GPUs, 1TB of RAM, and two Intel Xeon Platinum 8358 CPUs (32 cores, 2.6 GHz).
One 1.9TB NVMe SSD and four 7.6TB NVMe SSDs make up each server’s local storage, and five NVIDIA ConnectX-6 ports and one user interface are used to link the servers to one another made out of a dual-port NVIDIA ConnectX-6 interface for storage.
The addition of Supermicro’s 580 PFLOPS machine will enhance the performance of NEC’s supercomputer for AI research, speed up the creation of more sophisticated AI algorithms, and make it the biggest supercomputer in the Japanese market.
Speaking on the partnership, Charles Liang, President and CEO of Supermicro, said the firm will continue to cooperate with leading global customers to provide advanced rack-level server solutions, allowing customers to achieve business goals in a faster and more efficient way.
On the other hand, Takamino Kitano, senior AI platform architect at NEC, said that the Supermicro GPU server is highly scalable and customizable, allowing the system to evolve flexibly with the development of AI research in the future. NEC will continue to work closely with Supermicro to advance AI research further.
“NEC Corporation is leading the AI revolution by conducting leading-edge research using Supermicro GPU servers,” said Takatoshi Kitano, “The Supermicro GPU server is highly extensible and customizable, allowing the system to flexibly evolve as future AI research develops. We continue to work with Supermicro closely to advance our AI research.”