NVIDIA DGX B300
- Condition
- New
- Manufacturer
- NVIDIA
- CPU
- 2x Intel Xeon 6776P (64c/128t, 2.3GHz-3.9GHz, 350W)
- RAM
- 2000GB (DDR5 ECC REG)
- NIC card
- ConnectX-8 + BlueField-3 DPU + 1GbE RJ45
- GPU
- 8 x NVIDIA B300 SXM
- Accelerator manufacturer
- NVIDIA
- Accelerator model
- NVIDIA B300 Blackwell Ultra
- Installed GPU/APU count
- 8 GPUs
- Maximum GPU/APU count
- 8 SXM GPUs
- Memory per accelerator
- 288GB
- Total accelerator memory
- 2304GB
- GPU memory type
- HBM3e
- Accelerator memory bandwidth
- Up to 8TB/s per GPU
- Accelerator form factor
- SXM
- Accelerator interconnect
- Fifth-generation NVIDIA NVLink with two NVLink switches
- Accelerator-to-CPU interface
- PCIe 5.0
- Supported compute data types
- FP64, FP32, TF32, BF16, FP16, FP8, FP6, FP4 and INT8
- Accelerator virtualization
- MIG-backed NVIDIA vGPU, SR-IOV and GPU passthrough
- MIG support
- Up to 7 isolated GPU instances per GPU
- Software ecosystem
- NVIDIA CUDA, NVIDIA AI Enterprise, NVIDIA DGX OS, NVIDIA NGC, NVIDIA Mission Control, NVIDIA Dynamo and NVIDIA Transformer Engine
- Maximum system power
- 15kW
- Cooling
- Air cooling with front-to-back chassis airflow and redundant hot-swap fan modules
- Recommended workloads
- AI reasoning, agentic AI, generative AI and large-language-model training, fine-tuning and inference, multimodal AI, accelerated analytics, scientific computing and HPC
- Dimensions
- 442 x 482.6 x 904.2 mm (H x W x D)
- Weight
- 168 kg for the AC PSU configuration
Need a technical consultation on choosing a server?
Submit a requestNVIDIA DGX B300 is an integrated accelerated computing system designed for enterprise artificial intelligence, machine learning, generative AI, analytics, and high-performance computing (HPC) workloads. This new NVIDIA configuration is built around eight NVIDIA B300 SXM GPUs and provides a unified architecture for AI training, inference, and large-scale data processing in modern data centers.
Unlike a traditional GPU server assembled from individual components, NVIDIA DGX B300 combines GPU acceleration, high-speed interconnects, storage, networking, and a validated NVIDIA software stack in a single enterprise-ready platform. In this configuration, the system includes NVIDIA ConnectX-8 networking, NVIDIA BlueField-3 DPUs, and a 1GbE RJ45 management interface, allowing it to operate as a standalone AI compute node or as part of scalable DGX-based clusters.
GPU architecture for AI training and inference
NVIDIA DGX B300 is equipped with eight NVIDIA B300 SXM GPUs based on the NVIDIA Blackwell Ultra architecture. The eight-GPU configuration provides 2,304GB of total GPU memory, giving enterprises the capacity required for large language models, generative AI services, accelerated analytics, and HPC workloads.
The NVIDIA B300 GPU is designed for demanding AI workloads, including large-scale model training, fine-tuning, inference, recommendation systems, computer vision, analytics, and scientific computing. Tensor Cores with support for advanced low-precision and mixed-precision computing help accelerate neural network operations while maintaining the performance and efficiency required for production environments.
The eight GPUs are connected through fifth-generation NVIDIA NVLink technology and NVIDIA NVSwitch interconnects, creating a high-bandwidth GPU communication fabric. This architecture allows multiple accelerators to work together efficiently as a single computing domain for demanding distributed workloads.
The system is powered by two Intel Xeon 6776P processors, each with 64 cores and 128 threads, a 2.3GHz base frequency, up to 3.9GHz maximum frequency, and a 350W processor power rating. This configuration includes 2,000GB of DDR5 ECC registered memory, enabling efficient data preparation, orchestration, containerized workloads, and CPU-side processing tasks required by enterprise AI pipelines.
Primary use cases
- AI model training and fine-tuning. DGX B300 accelerates development and optimization of advanced AI models, including generative AI, natural language processing, computer vision, recommendation systems, and enterprise AI applications.
- Large-scale inference workloads. The platform is designed for deploying production AI services, intelligent applications, automated decision systems, chatbots, and real-time model execution.
- Data analytics and accelerated computing. DGX B300 supports GPU-accelerated analytics, scientific computing, simulations, and workloads that combine traditional HPC methods with machine learning.
- Multi-user AI environments. The platform can support isolated and efficiently scheduled accelerated workloads for multiple teams, helping organizations improve resource utilization across AI development, testing, and production environments.
- Enterprise AI infrastructure. DGX B300 provides a standardized foundation for organizations building internal AI platforms, research environments, scalable machine learning operations, and private generative AI infrastructure.
NVIDIA DGX B300 in enterprise infrastructure
NVIDIA DGX B300 is designed for organizations that require dedicated AI computing resources for continuous development, testing, and production deployment. The platform enables data science teams, AI engineers, and enterprise developers to work within a consistent hardware and software environment instead of managing separate accelerator servers and integration layers.
The system can be deployed as an individual AI node or integrated into larger NVIDIA-based environments. When planning an enterprise deployment, organizations should evaluate not only GPU performance but also network architecture, external storage requirements, workload distribution, power availability, cooling capacity, and data center readiness.
With eight NVIDIA B300 SXM GPUs, dual Intel Xeon 6776P processors, 2,000GB of DDR5 ECC registered memory, ConnectX-8 networking, BlueField-3 DPUs, and the NVIDIA DGX software ecosystem, this DGX B300 configuration provides a reliable foundation for organizations developing generative AI services, machine learning platforms, analytics solutions, and HPC applications that require predictable performance and simplified infrastructure management.
to Buy a Server
from Servermall
warranty
testing
We want to make sure that our equipment is fully reliable for you. That’s why all our equipment (even new) is subject to complete diagnostics, testing, and part replacement.
testing
We fully load our equipment with various utilities and software. Rest assured, it’ll work for a very long time.
Power supply
The components work at full capacity and are tested for several days. If we are not sure about something, we replace it.
of all equipment
Our regulations imply the same testing for both used and new equipment. Oh, and by the way, we complement the official manufacturer’s warranty with our own warranty.