Powered by two Intel® Xeon® 6 processors, the QuantaGrid D75H-10U is based on the NVIDIA HGX B300 platform and  features eight OSFP ports, supporting 800G Ethernet or InfiniBand.

With 72 PFLOPS FP8 for training and 144 PFLOPS FP4 for inference, QuantaGrid D75H-10U is ideal for hyperscalers to build large high-speed scale-out GPU clusters to accelerate LLM training and inference. It delivers breakthrough performance for the most complex workloads including agentic AI, AI reasoning, and real-time video generation forevery data center.

Seamless Data Processing and High-Performance AI at Scale

The QuantaGrid D75H-10U supports NVIDIA BlueField®-3 DPUs for North-South data transfer to eliminate bottlenecks in data processing. The system also supports 8x E1.S NVMe storage drives for 1:1 GPUDirect® Storage, and 1:1 GPUDirect® RDMA, to ensure high-speed cross-node communication for the most extreme AI workloads.

The D75H-10U stands as an ideal solution for AI inference, AI training, and data analytics, providing a robust and flexible architecture designed for resource-intensive workloads.

Advanced Serviceability

• Front-Access PSUs and SSDs to Reduce Downtime: Enabling maintenance, troubleshooting, and component swaps without requiring rear rack access, significantly improving service efficiency.

• Toolless Modular Design: CPU and GPU sleds can be serviced individually to maximize system uptime, allowing key components to be easily accessed and removed directly from the system chassis while remaining in the rack.