Supermicro's Vera Rubin systems use liquid cooling for AI workloads
Supermicro just rolled out its NVIDIA Vera Rubin lineup, built for next-level AI and massive inference workloads.
The three new systems, Vera Rubin NVL72, HGX Rubin NVL8, and Vera CPU, use advanced liquid cooling to keep things running smoothly.
NVL72 is the flagship system
The flagship NVL72 packs 72 GPUs and 36 CPUs into one rack, delivering up to 3.6 exaflops of AI processing power, 75TB of fast memory, and lightning-fast bandwidth.
The custom CPUs and GPUs are designed for heavy-duty AI tasks like large language models.
HGX Rubin NVL8 can scale up with different CPU choices
If you need flexibility, the HGX Rubin NVL8 can scale up with different CPU choices (Vera, AMD, or Intel), while the compact Vera CPU system fits 2 CPUs and up to 6 powerful Blackwell GPUs in a small footprint.
Supermicro has announced expanded rack-scale manufacturing capacity and upgraded liquid-cooling capabilities to support deployment of Vera Rubin and HGX Rubin platforms.