Supermicro GPU Servers Built for Modern Data Centers
A flexible AI processing environment makes it easy to incorporate AI into IT workflows. Many applications involve small to large AI models and different data batch sizes. Simultaneously, AI product development and IT deployment require a wide range of AI processing capabilities on demand. NVIDIA® Virtual Compute Server (vCS) and Red Hat® Virtualization running on Supermicro GPU systems with NVIDIA HGX™ A100 provide that flexibility, cost-effectiveness, and responsiveness to run AI workloads for developers and deployment.
We describe in this paper the capabilities and flexibility of Supermicro’s systems with NVIDIA HGX A100 and NVIDIA vCS combined with Red Hat Enterprise Linux® to virtualize NVIDIA A100 GPUs to run independent AI workloads. Multiple developers, some needing a small GPU and some needing multiple GPUs, can share the same HGX systems. The same systems also support AI workloads incorporated into IT applications at the same time.