Pre-Order QuantaGrid D75H-10U with B300
Built on the next-gen DGX B300, offering 11× inference and 4× training improvements. Lock in your deposit today to ensure early access
Breakthrough Performance for Massive Scale
The QuantaGrid D75H-10U, powered by NVIDIA’s HGX B300 platform, delivers 72 PFLOPS FP8 for LLM training and 144 PFLOPS FP4 for real-time inference. With 800G networking and air-cooled efficiency, it’s built for hyperscale AI workloads.
38K+
FP8 Training Power
Highest throughput for large-scale LLM training pipelines.
800G
East-West Bandwidth
Enabled by PCIe Gen 6 and ConnectX-8 readiness for seamless clustering.
10U
Air-Cooled Efficiency
QCT's new air-cooled powerhouse, perfect for demanding data centers.
Prebuilt Configuration: D75H-10u L6
The D75H-10U L6, with 8x NVIDIA B300 GPUs, 2TB DDR5, and 30TB+ NVMe storage, is built for instant AI integration.
Compute & Graphics (HGX B300)
- GPUs: 8x HGX B300 NVL8 GPU’s
- Processors: Dual Intel® Xeon® 6767P 2.4GHz (64-core each)
Memory & Storage
- System Memory: 2TB DDR5 (32x 64GB 6400MHz)
- Data Storage: 8x 3.84TB PCIe Gen5 NVMe Server SSD
- Boot Drives: 2x 1.92TB NVMe m.2 SSD Boot Drives
Networking & Chassis
- NIC: Nvidia ConnectX-7 200Gb/s Single Port SFP
- Form Factor: Brand New 10U Chassis (Air Cooled)
Expansion
- 4x FHHL PCIe Gen5 x16 Slots
- Future-proof for CXL 2.0
Build Tomorrow’s AI with B300
Be the first to harness NVIDIA B300’s 11x inference and 4x training performance. Reserve your D75H-10U for late October and shape AI’s future.
Secure Your QuantaGrid D75H-10U
We are currently accepting deposits to reserve your B300 server for the upcoming Late October delivery window. Fill out the form below to initiate the deposit process and finalize your order details with a Dataknox specialist.
