Super X AI Technology released its new SuperX XN9160-B200 AI Server, powered by NVIDIA’s Blackwell B200 GPU. The server is designed to address the growing need for high-performance computing in AI training, machine learning, and high-performance computing (HPC) applications.
The XN9160-B200 is built to handle large-scale distributed AI training and inference workloads, supporting tasks such as reinforcement learning, distillation techniques, and multimodal model processing. It is also optimized for HPC applications, including climate modeling, drug discovery, seismic analysis, and insurance risk modeling. The server’s performance is comparable to traditional supercomputers, housed within a 10U chassis.
Equipped with eight NVIDIA Blackwell B200 GPUs and fifth-generation NVLink technology, the XN9160-B200 offers an inter-GPU bandwidth of up to 1.8TB/s, enabling a 3x speed improvement in large-scale AI model training. For inference, it delivers 58 tokens per second per card on the GPT-MoE 1.8T model at FP8 precision with 1440GB of HBM3E memory, a 15x increase over the 3.5 tokens per second of the H100 platform. The system includes two Intel® Xeon® 6710E processors (64 cores, 2.40 GHz, 205W), 32x 96GB DDR5 5600 RDIMM memory, a 960GB SSD system disk, and a 3.84TB NVMe U.2 storage disk.
The server features a multi-path power redundancy system with 1+1 redundant 12V and 4+4 redundant 54V GPU power supplies to ensure continuous operation. It also includes an AST2600 intelligent management system for remote monitoring and undergoes over 48 hours of stress testing, cold and hot boot validation, and high/low-temperature aging screening. SuperX provides a three-year warranty and technical support for the server.
Networking capabilities include eight CX7 MCX75310 IB Cards (400G OSFP), one BCM957608-P2200G (dual 200G QSFP56), and one BCM957412A4120AC (dual 10G SFP+). The server measures 440mm (H) x 448mm (W) x 900mm (D).
The XN9160-B200 targets large tech companies for training and deploying foundation models, academic and research institutions for scientific simulations, finance and insurance sectors for risk modeling, pharmaceutical and healthcare industries for drug screening and bioinformatics, and government and meteorological agencies for climate modeling and disaster prediction.
