Business
SuperX Unveils the All-New SuperX XN9160-B200 AI Server, Powered by NVIDIA Blackwell GPU -- Accelerating AI Innovation by 30x as Compared to H100 Series with Supercomputer-Class Performance
SINGAPORE, July 30, 2025 /PRNewswire/ -- Super X AI Technology Limited (Nasdaq: SUPX) ("Company" or "SuperX") today announced the launch of its latest
About this update from Superx Ai Technology Limited
[{"type":"text","content":"SINGAPORE, July 30, 2025 /PRNewswire/ -- Super X AI Technology Limited (Nasdaq: SUPX) (\"Company\" or \"SuperX\") today announced the launch of its latest flagship product — the SuperX XN9160-B200 AI Server. Powered by NVIDIA's Blackwell architecture GPU (B200), this next-generation AI server is engineered to meet the rising demand for scalable, high-performance computing in AI training, machine learning (ML), and high-performance computing (HPC) workloads.\nThe XN9160-B200 AI Server is purpose-built to accelerate large-scale distributed AI training and AI inference workloads. It is optimized for GPU-supported tasks to support intensive GPU instances, particularly for training and inference of foundation models using reinforcement learning (RL) and distillation techniques, multimodal model training and inference, as well as HPC applications such as climate modeling, drug discovery, seismic analysis, and insurance risk modeling. Its performance rivals that of a traditional supercomputer, offering enterprise-grade capabilities in a compact form.\nThe launch of the SuperX XN9160-B200 AI server marks a significant milestone in SuperX's AI infrastructure roadmap, delivering powerful GPU instances and compute capabilities to accelerate global AI innovation.\nXN9160-B200 AI Server\nThe all-new XN9160-B200 features 8 NVIDIA Blackwell B200 GPUs, fifth-generation NVLink technology, 1440 GB of high-bandwidth memory (HBM3E), and 6th Gen Intel® Xeon® processors, unleashing extreme AI compute performance within a 10U chassis.\nBuilt for AI —— Cutting-edge Training Performance\nThe SuperX XN9160-B200 is powered by its core engine: 8 NVIDIA Blackwell B200 GPUs, equipped with fifth-generation NVLink technology to provide ultra-high inter-GPU bandwidth of up to 1.8TB/s. This significantly accelerates large-scale AI model training, achieving up to a 3x speed improvement and drastically shortening the R&D cycle for tasks like pre-training and fine-tuning trillion-parameter models. For inference, it represents a quantum leap in performance: with 1440GB of high-performance HBM3E memory running at FP8 precision, it achieves an astonishing throughput of 58 tokens per second per card on the GPT-MoE 1.8T model. Compared to the 3.5 tokens per second of the previous generation H100 platform, this represents an extreme performance increase of up to 15x.\nTh...