© 2025 ITCT Shop. All rights reserved.
Description
Description
The NVIDIA DGX H100 is a powerful AI workstation built for heavy AI workloads, deep learning training, and large-scale model inference. It comes with 8 NVIDIA H100 GPUs, each with 80GB of super-fast HBM3 memory, connected through high-speed NVLink, making it perfect for running large AI models quickly.
It uses 2 Intel Xeon Platinum 8480+ CPUs (56 cores each) and 2TB of DDR5 memory, providing plenty of processing power and memory for demanding tasks. Storage is handled by 8 NVMe SSDs (3.84TB each), giving a total of ~30TB fast storage for AI datasets.
The DGX H100 supports popular AI frameworks such as vLLM, Megatron, DeepSpeed, HuggingFace Transformers, TensorRT-LLM, and Triton Inference Server, making it flexible for research and production.
Networking is extremely fast with high-speed ConnectX-7 NICs and NVLink + Infiniband for connecting multiple systems together. The system also includes secure boot, TPM 2.0, RAID support, and redundant 3000W power supplies for reliability. It runs NVIDIA DGX OS 6, ready for container-based AI workflows.
Key Features:
| Specifications |
|---|
| 8GPU, HGX H100 SXM5, RoHS NVIDIA DGX H100 System (x1) |
| 8GPU, HGX H100 SXM5, RoHS EWCSC (x1) |
| 0% 3 YRS LABOR, 3 YRS PARTS, 1 YR CRS UNDER LIMITED WARRANTY |
| CPU-Intel-Xeon-8480+ (x2) |
| Intel Xeon Platinum 8480+, 2P, 56C, 2.0GHz, 350W TDP |
| MEM-DR564MC-ER64 (x32) |
| 64GB DDR5-4800 ECC RDIMM 2Rx4 – Total: 2TB system memory |
| HDS-25N4-003T8-E1-TXD-NON-008 (x8) |
| SSD 2.5″ NVMe PCIe Gen4 3.84TB, 1DWPD, TLC NAND, 7mm – Total: ~30TB usable NVMe |
| GPU-H100 SXM5 Gen5: 900GB/s via NVLink 4.0 / NVSwitch |
| 80GB HBM3 per GPU (Total: 640GB HBM3) |
| TF32: 32,896 TFLOPS | FP8: 67,000 TFLOPS (aggregate system) |
| Networking |
| ConnectX-7 NICs: 2x 400GbE or 8x 100GbE |
| NVIDIA NVLink and In niBand Switch Fabric for multi-node DGX SuperPods |
| TPM: Secure Boot with UEFI and TPM 2.0 |
| RAID: Software RAID 0/1 for boot drives |
| Power Supply: Redundant 3000W (Titanium-rated) hot-swappable PSUs |
| SFT-DGXOS-SINGLE (x1) |
| NVIDIA Base Command OS (DGX OS 6) with container orchestration |
| LLM Framework Support: vLLM, NeMo Megatron, DeepSpeed, HuggingFace Transformers, TensorRT-LLM, Triton Inference Server |
| Interconnection: Full-bandwidth NVLink and In niBand-based scale-out support |
Brand
Brand
Nvidia
Reviews (0)
Shipping & Delivery
Shipping & Payment
Worldwide Shipping Available
We accept:
Visa
Mastercard
American Express
International Orders
For international shipping, you must have an active account with
UPS,
FedEx,
or DHL,
or provide a US-based freight forwarder address for delivery.
Additional Information
Additional information
Related products
NVIDIA DGX H200 (AI Supercomputer – 8× H200 SXM5 GPUs, 2× Intel Xeon 64C, 2TB DDR5, 30TB NVMe)
USD550,000
Reviews
There are no reviews yet.