ASUS NVIDIA HGX B300 8-GPU 9U AI Training Server with 10x NVMe Bays

The ASUS NVIDIA HGX B300 8-GPU 9U AI Training Server (SKU: ASUS NVIDIA HGX B300 10-Bay SFF NVMe 8-GPU 9U AI Training Server) is an ultra-high-density system designed for large-scale AI training, deep learning, and HPC environments. It uses a 9U rack form factor and integrates NVIDIA HGX GPU architecture with enterprise compute and memory capabilities.

This platform supports up to two 6th Generation Intel Xeon Scalable processors with up to 350W TDP per CPU. As a result, it delivers high core counts and strong compute performance for AI and data-intensive workloads. In addition, it includes 32 DDR5 DIMM slots (16 per CPU). It supports up to 8TB ECC RDIMM/3DS RDIMM memory. Therefore, it enables large memory capacity and high bandwidth for model training and analytics.

The system integrates 8× high-performance GPUs based on the NVIDIA HGX B300 platform. As a result, it provides massive parallel processing capability for AI training, large language models, and scientific computing. The architecture is optimized for GPU-to-GPU communication and high-throughput workloads.

For storage, the server supports 10× 2.5″ SFF NVMe hot-swappable drive bays. It also supports PCIe-based storage controllers. As a result, it delivers high-speed, low-latency storage for large datasets and training pipelines.

For networking, the system includes 2× 10GbE LAN ports via the Intel X710-AT2 controller. In addition, it supports further expansion through PCIe Gen5 slots. Therefore, it enables integration into high-bandwidth data center networks.

The platform supports PCIe Gen5 expansion with 5 slots, including 4× PCIe Gen5 x16 and 1× PCIe Gen5 x8 (RAID dedicated). As a result, it provides flexibility for additional networking, storage, or accelerator components.

For power and reliability, the server includes 10× 3200W 80 PLUS Titanium hot-swappable power supplies. This ensures stable operation under extreme workloads and provides redundancy for high availability. It also includes a dedicated RJ45 management port for remote monitoring and control.

This system is fully tested and ready for reuse. It is ideal for AI training clusters, deep learning, HPC, and enterprise data centers requiring maximum GPU density and performance.

Key Features:

• 8× NVIDIA HGX GPUs for large-scale AI and deep learning workloads
• Dual 6th Gen Intel Xeon Scalable processors with high TDP support
• 32 DDR5 DIMM slots supporting up to 8TB ECC memory
• 10× 2.5″ NVMe hot-swappable drive bays for high-speed storage
• PCIe Gen5 expansion with multiple x16 slots for scalability
• 10× 3200W Titanium redundant power supplies for high availability