GIGABYTE Introduces AI TOP ATOM Cluster for Scientific Computing
GIGABYTE has launched its AI TOP ATOM cluster, designed to scale local AI computation and scientific simulations. The four-node system offers enhanced memory capacity and data sovereignty for complex workloads.
GIGABYTE is introducing its AI TOP ATOM cluster, a solution aimed at addressing the growing demands of local artificial intelligence computing and scientific simulations. Built around a four-node architecture, the cluster enables the execution of memory-intensive workloads locally without compromising data security.
Each AI TOP ATOM node delivers 1 PFLOPS of AI performance and 128 GB of unified memory. A 200 GbE RoCE-compatible switch interconnects the nodes, facilitating distributed memory utilization. This modular architecture allows organizations to scale from one to four nodes as workload requirements evolve, while maintaining local deployment and full data sovereignty.
GIGABYTE collaborated with NVIDIA to demonstrate the cluster's capabilities through an AI-powered scientific computing workflow. This workflow leverages NVIDIA NemoClaw blueprints and Nemotron-3 Nano-30B-NVFP4 models. The system was used to simulate molecular dynamics, including the development of thermal interface materials for semiconductors.
The demonstration highlighted how a four-node AI TOP ATOM cluster can handle simulations involving over 30 million atoms for next-generation integrated circuit packaging research. This surpasses the typical limitations of standalone systems, which often face memory constraints around 10 million atoms. The solution extends GIGABYTE's offerings from AI development to demanding scientific computing applications.