Super X AI Technology Limited has announced the launch of its latest flagship product, the SuperX XN9160-B300 AI server. Powered by Nvidia's Blackwell GPU (B300), the XN9160-B300 is designed to meet the growing demand for scalable, high-performance computing across AI training, machine learning (ML), and high-performance computing (HPC) workloads. Designed for extreme performance, this system integrates advanced networking capabilities, scalable architectures and energy-efficient designs to support mission-critical data center environments.
The SuperX XN9160-B300 AI server is dedicated to accelerating large scale distributed AI training and AI inference workloads, providing extreme GPU performance for intensive, high-demand applications. Optimized for GPU-supported tasks, it excels in training and inference on basic models such as reinforcement learning (RL), distillation technology, and multimodal AI models, as well as high performance for HPC workloads such as climate modeling, drug discovery, seismic analysis, and insurance risk modeling.
Designed for enterprise-scale AI and HPC environments, the XN9160-B300 combines supercomputer-level performance with energy-efficient, scalable architecture to deliver mission-critical capabilities in a compact data center-enabled form factor.
The launch of the SuperX XN9160-B300 AI server marks key milestones in SuperX's AI infrastructure roadmap, delivering powerful GPU instances and calculating the capabilities to accelerate global AI innovation.
XN9160-B300 AI Server
The SuperX XN9160-B300 AI server delivers extreme AI computing performance within an 8U chassis and features a high-speed network with Intel Xeon 6 processors, 8 NVIDIA BLACKWELL B300 GPUs, up to 32 DDR5 DIMMs, and up to 8 x 800 gb/s in-band.
Also Read: Aithority Interview with Speakup CEO Tim Morrs
High GPU power and memory
The XN9160-B300 is built as a highly scalable AI node and features an NVIDIA HGX B300 module housing 8 NVIDIA BLACKWELL B300 GPU. This configuration provides peak performance for Blackwell Generation, specifically designed for the following ERA AI workloads:
Importantly, the server provides 2,304GB of unified HBM3E memory for eight GPUs (288GB per GPU). This huge memory pool eliminates memory offloading, supports large model housing, and is essential for managing the vast key/value caches required for high components, long-term generation AI and large language models.
Extreme reasoning and training throughput
The system leverages the B300 Ultra's excellent FP4/NVFP4 accuracy and second-generation transformer engine to achieve a monumental performance leap. According to Nvidia, the Blackwell Ultra makes a decisive leap to Blackwell by increasing NVFP4 computing by 50% and adding 50% HBM capacity per chip, allowing for larger models and faster throughput without compromising efficiency.[1] Scaling is easy thanks to eight 800GB/s OSFP ports for InfiniBand or dual 400GB/s Ethernet. These ports allow for the fast, low-latency communication required to connect servers to vast AI factories and SuperPod clusters. The fifth generation nvlink interconnect allows eight onboard GPUs to communicate seamlessly and act as a single powerful accelerator.
Robust CPU and Power Foundation
The GPU complex is supported on a robust host platform with dual Intel Xeon 6 processors, providing the efficiency and bandwidth required to deliver data to the accelerators. The memory subsystem is equally formidable, using 32 DDR5 DIMMs that support speeds up to 8000MT/s (MRDIMMs), ensuring that the host platform does not bottleneck GPU processing.
For mission-critical reliability and sustainable performance, the XN9160-B300 is equipped with a 12x3000W 80 plastionic titanium redundant power supply, ensuring extremely high energy efficiency and stability under continuous peak loads. The system also includes comprehensive storage options, including multiple high-speed PCIEGEN5 X16 slots and eight 2.5-inch Gen5 NVME hot-swap bays.
Market positioning
The XN9160-B300 is built for organizations pushing the boundaries of AI. Maximum scale, next-generation models, and ultra-low latency are core requirements.
- Hyperscale AI Factory: For cloud providers and large enterprises, it builds basic models of sign parameters and uses highly demanding advanced current AI inference engines.
- Science simulation and research: Exuscar scientific computing, advanced molecular dynamics, and the creation of comprehensive industrial or biological digital twins.
- Financial Services: Deployment of complex, large-scale language models for real-time risk modeling, high-frequency trading simulations, and financial analysis with ultra-low latency demands.
- Bioinformatics and Genomics: To accelerate protein structure prediction at scales requiring large-scale genomic sequencing, drug discovery pipelines, and B300's immeasurable memory capabilities.
- Global System Modeling: Very detailed disaster forecasting for national weather and government agencies that require extreme calculations for global climate and weather modeling.
Also Read: Cognitive Product Design: Enhance the Power of Non-Technical Users through Natural Language Interaction with AI-Native PLM
[To share your insights with us as part of editorial or sponsored content, please write to psen@itechseries.com]
