MSI
MSI EDGEXPERT-32SEU Nvidia DGX Spark GB10 128GB 4TB AI Mini DT PC
24-Month Warranty- GB10 Grace Blackwell SuperchipProcessor
The NVIDIA DGX B200 is an enterprise AI platform that combines eight Blackwell GPUs in a single 10U system, designed for every AI workflow from training to inference. With 1,440 GB of total GPU memory, fifth-generation NVLink and NVIDIA's ready-made software stack, it appeals to organisations that want to run large language models, recommendation systems and generative AI applications on a single system.
The NVIDIA DGX B200 is an enterprise AI platform that combines eight Blackwell GPUs in a single 10U system, designed for every AI workflow from training to inference. With 1,440 GB of total GPU memory, fifth-generation NVLink and NVIDIA's ready-made software stack, it appeals to organisations that want to run large language models, recommendation systems and generative AI applications on a single system.
The DGX B200 is NVIDIA's Blackwell-generation DGX system and is positioned as the core building block of the "AI factory" architecture. The same system can be used for model training, fine-tuning and high-volume inference, reducing the need to build separate infrastructure for different workloads.
The system has eight NVIDIA Blackwell GPUs; total GPU memory is 1,440 GB and total HBM3e bandwidth is 64 TB/s. The GPUs are interconnected with fifth-generation NVLink and NVSwitch. According to the NVIDIA datasheet, the system delivers 144 petaFLOPS (sparse) / 72 petaFLOPS (dense) of FP4 Tensor Core performance. Power is provided by six 3.3 kW power supplies; maximum system power is 14.3 kW.
GPU-intensive workloads such as large language model pre-training and fine-tuning, real-time LLM inference, recommendation systems, computer vision, speech recognition and scientific simulation.
The NVIDIA DGX software stack and the enterprise support options that come with NVIDIA AI Enterprise and Base Command standardise installing, updating and adding the system to a cluster.
It requires 10U of rack space and up to 14.3 kW of power; high-density racks, suitable cooling and power distribution should be planned in the data centre. Multiple systems can be scaled in a DGX SuperPOD architecture over InfiniBand or Ethernet networking.
Suitable for banks, telecom operators, public institutions, research centres and technology companies developing AI products that want to build their own AI infrastructure in-house.
| Marka | NVIDIA |
|---|---|
| Model | DGX B200 |
| Product type | Raf tipi yapay zekâ sunucusu (10U) |
| GPU | 8 × NVIDIA Blackwell GPU |
| Total GPU memory | 1.440 GB HBM3e |
| GPU memory bandwidth | 64 TB/sn (toplam) |
| FP4 Tensor Core performance | 144 PFLOPS (sparse) / 72 PFLOPS (dense) |
| GPU interconnect | 5. nesil NVLink, NVSwitch |
| Power supplies | 6 × 3,3 kW |
| Maximum system power | 14,3 kW |
| Form factor | 10U raf tipi |
| Software | NVIDIA DGX OS, NVIDIA AI Enterprise, NVIDIA Base Command |
| Processor, system memory, storage, networking | Listed in the NVIDIA datasheet; provided with the current configuration at the quotation stage |
The NVIDIA DGX B200 is supplied by special order. Once quantity, network and storage configuration, installation and NVIDIA support options are determined, a corporate quotation including price and delivery time is prepared. Stating your intended use and delivery country in the quote form speeds up the process.
| Brand | NVIDIA |
|---|---|
| Model | DGX B200 |
| Status | New · Teklif ile tedarik |
| Warranty | Garanti ve servis koşulları teklif aşamasında üretici ve tedarik kanalına göre netleşir. |
| SKU | AI-NVIDIA-DGX-B200 |
Warranty: Garanti ve servis koşulları teklif aşamasında üretici ve tedarik kanalına göre netleşir.
In case of a fault you can create a service request and track the process on the servis takip page.
MSI
Pny
Acer
NVIDIA GB10 Grace Blackwell süper çip · 128 GB LPDDR5x birleşik bellek · Kompakt mini PC kasası