Supermicro announces an end-to-end DCBBS blueprint built around NVIDIA’s Vera Rubin NVL4 platform for converged high-performance computing (HPC) and AI infrastructure. The solution is described as a total rack-based offering designed for performance on both AI workloads and FP64 (double-precision) simulations. According to the available reports, the system can scale up to configurations including as many as 1,152 NVIDIA Rubin GPUs and 576 NVIDIA Vera CPUs within liquid-cooled racks.
The announcement highlights a scalable unit rated at 3.2 megawatts, intended to support next-generation compute performance for research and deployment across clusters of varying size. The system is said to use DLC-2 Direct Liquid Cooling, with specifications mentioned that support high power density per rack (including figures described as 362 kW per level). While the sources provided repeat the same core product claims, they do not add detailed independent benchmarking, pricing, or customer deployment information in the text excerpts.
