Featuring liquid-cooling plus NVIDIA’s BlueField-3 or NVIDIA ConnectX-7 for network communications, the Supermicro GPU SuperServer ARS-111GL-NHR-LCC. With extremely low-latency, this platform is designed for high-performance computing, large language models (LLM) and Generative AI, plus AI, Deep Learning and Inference. Outfitted with an NVIDIA Grace Hopper Superchip integrated with an H100 GPU, the system delivers up to 576GB of coherent memory per node along with 480GB of LPDDR5X memory and 96GB of HBM3 memory provided by the H100 GPU, ideal for large language model applications.
CPUs with a Thermal Design Power range of up to 1000W are supported via liquid-cooling with a direct to chip cold plate, which is optional. There are 2x x4 NVMe slots on the system board for support of one or two M.2 drives that can be used to boot the system. Two PCIe 5.0 x16 full-height, full-length slots for support of network controllers. With a dedicated RJ45 GbE BMC LAN port, the system can be easily managed using a number of Supermicro software utilities.
Compose the exact configuration you looking for to build the perfect server.
Just add components and quantities you would like to incorporate into the system
