| MOQ: | 1 |
| Price: | USD 5110/set |
| Standard Packaging: | Carton with waterproof warpped |
| Delivery Period: | Quantity (pieces) 1-10 for around 15days, > 10, to be negotiated |
| Payment Method: | T/T,MoneyGram,Credit Card,PayPal,Western Union,Cash,Escrow; |
| Architecture | NVIDIA Grace Blackwell |
|---|---|
| Processor | NVIDIA Grace 20-core Arm CPU 10 Cortex-x925 cores 10 Cortex-A725 cores |
| Rear Ports | 4 USB4 Type-C (20 Gb/s) 1 HDMI 2.1a (multi-channel audio) 1 RJ-45 10 Gigabit Ethernet 2 ConnectX-7 (for connecting to other ThinkStation PGX nodes) |
| Wireless Networking | Wi-Fi 7 Bluetooth 5.3 LE |
| Ethernet | 10 Gigabit Ethernet |
| Security | Self-encrypting NVMe (AES SED) TPM 2.0 NVLink-C2C Secure Zone NVIDIA Firmware Recovery AMI Setup Password UEFI Secure Boot |
| Graphics Card | NVIDIA Blackwell Architecture Blackwell CUDA Cores 5th-Generation Tensor Cores 4th-Generation Ray Tracing Cores Embedded NVENC Video Encoder Embedded NVDEC Video Decoder |
| Artificial Intelligence | 1,000 TOPS (trillion operations per second) 1 PFLOP (FP4 precision) |
| Unified System Memory | 128 GB LPDDR5x 256-bit bus 273 GB/s bandwidth |
| Storage Capacity | 1 TB or 4 TB NVMe M.2 |
| Scalability | 2 ThinkStation PGX nodes can be connected via the integrated NVIDIA ConnectX-7 to process models with up to 405B parameters |
| Power Consumption | 240 watts |
| Operating System | NVIDIA DGX Operating System Ubuntu Linux Pro running on the NVIDIA base operating system |
| Pre-installed NVIDIA AI Software Stack | NVIDIA NIM NVIDIA Blueprint PyTorch (deep learning framework) Jupyter (interactive computing environment) Ollama (large language model runtime) |
With more than 20 years of expertise in China's ICT industry, we empower global enterprises with cutting-edge solutions in servers, storage, networking, and artificial intelligence. As a trusted partner officially certified by Dell, Huawei, Inspur, Lenovo, Microsoft, and other leading brands, we have built a strong team of 20+ engineers covering architecture design, O&M support, and project delivery.
In addition to standard x86 solutions, we also provide China-developed indigenous servers featuring domestic CPUs such as Huawei Kunpeng ARM, Hygon C86, Phytium, and Loongson. These servers are widely adopted in key industries (government, finance, telecom, energy) for their secure, controllable, and fully localized IT infrastructure. Leading vendors including Lenovo, Huawei, and Inspur have established mature product lines in this segment, offering a reliable alternative to traditional Western platforms.