Nvidia Unveils Vera Rubin Supercomputing Platform With 7 Exaflops AI Performance Per Rack

According to BlockBeats, Nvidia announced on June 22 the Vera Rubin supercomputing platform designed for scientific computing and high-performance computing (HPC). A single-rack system delivers over 7 Exaflops of AI compute performance and 5 Petaflops of FP64 double-precision performance, equivalent to TOP500-level supercomputer capability in one rack.

The platform combines Rubin GPU and Vera CPU architectures with NVLink-C2C, ConnectX-9 SuperNIC, and BlueField-4 DPU technologies, supporting unified workloads for AI training, scientific simulation, and data-intensive research. Vera Rubin targets climate modeling, computational fluid dynamics, quantum chemistry, and energy exploration applications.