Taking a look at the architecture powering advanced AI factories, NVIDIA has released a video providing the first overview of Eos, its newest datacenter-scale supercomputer. An extremely large-scale NVIDIA DGX SuperPOD, Eos is where NVIDIA developers create AI breakthroughs using accelerated compute infrastructure and fully optimized software.
Nvidia’s new supercomputer: Eos
Eos is built with 576 NVIDIA DGX H100 systems, NVIDIA Quantum-2 InfiniBand networking and software, delivering a total of 18.4 exaflops of FP8 AI performance. Unveiled at the Supercomputing 2023 trade show, Eos (named after the Greek goddess who is said to open the gates of dawn each day) reflects their commitment to advancing AI technology, according to NVIDIA.
Each DGX H100 system in Nvidia’s supercomputer is equipped with eight NVIDIA H100 Tensor Core GPUs. Eos has a total of 4,608 H100 GPUs ($30,000 each). As a result, Eos can manage the largest AI workloads to train large language models, recommender systems, quantum simulations and more.
Eos’ architecture is optimized for AI workloads that require ultra-low latency and highly efficient interconnection across a large set of accelerated compute nodes. This makes it an ideal solution for organizations looking to scale their AI capabilities.
Based on NVIDIA Quantum-2 InfiniBand with In-Network Computing technology, the network architecture on Nvidia’s new supercomputer supports data transfer rates of up to 400 Gb/s, facilitating the rapid movement of large data sets required for training complex AI models.
At the heart of Eos is the DGX SuperPOD architecture powered by NVIDIA’s DGX H100 systems. The architecture is designed to provide AI and computing with tightly integrated full-stack systems capable of computing at massive scale.
Ranked 9th in the TOP 500 list of the world’s fastest supercomputers, the Nvidia supercomputer Eos pushes the boundaries of AI technology and infrastructure.
