| Key Features/Applications | AI training, AI inference, large language models, generative AI, scientific computing and high-performance computing |
|---|---|
| CPU | Dual Intel Xeon 6768P, 128 cores total |
| Chassis | 8U Rackmount |
| Drive | 1 x 960 GB M.2 D100P NVMe PCIe 4.0 SSD |
| RAM | 4 TB DDR5-6400 ECC RDIMM (16 x 256 GB) |
| GPU | 8 x NVIDIA B300 Blackwell Ultra, 2.3 TB HBM3e |
| Network Ports | 8 x 800 Gb/s OSFP plus 2 x 10GbE RJ45 |
| GPU Platform | NVIDIA HGX B300 NVL8 |
| GPU Architecture | NVIDIA Blackwell Ultra |
| Number of GPUs | 8 x NVIDIA B300 GPUs |
| GPU Memory | 288 GB HBM3e per GPU |
| Total GPU Memory | 2,304 GB / 2.3 TB HBM3e |
| GPU Interconnect | NVIDIA NVLink and NVSwitch |
| NVLink Generation | 5th Generation |
| GPU-to-GPU Bandwidth | Up to 1.8 TB/s |
| CPU Model | 2 x Intel Xeon 6768P |
| CPU Cores | 64 cores per processor / 128 total cores |
| CPU Frequency | 2.40 GHz |
| CPU Cache | 336 MB per processor |
| CPU TDP | 330 W per processor |
| System Memory | 4 TB DDR5-6400 ECC RDIMM |
| Memory Configuration | 16 x 256 GB |
| Boot Storage | 1 x 960 GB M.2 D100P NVMe PCIe 4.0 SSD |
| Additional Storage | Optional, configuration dependent |
| High-Speed Networking | 8 x 800 Gb/s OSFP |
| Standard Networking | 2 x 10GbE RJ45 |
| Management | IPMI / BMC / Remote Management (Redfish, IPMI) |
| Form Factor | 8U Rackmount |
| Cooling | High-density air cooling with redundant fans |
| Power Supplies | 6 x 6,600 W (3+3 redundant, Titanium efficiency) |
| Maximum Power Consumption | Approximately 15 kW at full load |
| Maximum Heat Output | Approximately 50,000 BTU/hr |
| Dimensions (W x H x D) | 449 x 356 x 950 mm (17.6 x 13.8 x 37.4 in) |
| Net Weight | Approximately 119 kg (263 lb) |
| Gross Weight | Approximately 137 kg (302 lb) |
| Operating Temperature | 10 to 35 C (50 to 95 F) |
| Operating Humidity | 8 to 90 percent, non-condensing |
| Warranty | 3 Years Parts and Labour |
| Power | 6 x 6,600 W Titanium Level, 3+3 redundant |
| Memory | 4 TB DDR5-6400 ECC RDIMM (16 x 256 GB) |
| Storage | 1 x 960 GB M.2 D100P NVMe PCIe 4.0 SSD |
| AOC | 8 x 800 Gb/s OSFP plus 2 x 10GbE RJ45 |
Actual product may vary slightly from the photography shown.
8U AI Server with 8x NVIDIA HGX B300 NVL8
Starting at$0.00
In Stock
- Key Features/Applications: AI training, AI inference, large language models, generative AI, scientific computing and high-performance computing
- CPU: Dual Intel Xeon 6768P, 128 cores total
- Chassis: 8U Rackmount
- Drive: 1 x 960 GB M.2 D100P NVMe PCIe 4.0 SSD
- RAM: 4 TB DDR5-6400 ECC RDIMM (16 x 256 GB)
- GPU: 8 x NVIDIA B300 Blackwell Ultra, 2.3 TB HBM3e
- Network Ports: 8 x 800 Gb/s OSFP plus 2 x 10GbE RJ45
Request a quotation for volume pricing
Request a QuotationShipped Configuration
The base build as despatched. Every component group can be specified differently against your requirement, and the revised price is confirmed in a written quotation.
- CPUDual Intel Xeon 6768P, 128 cores total
- Memory4 TB DDR5-6400 ECC RDIMM (16 x 256 GB)
- Storage1 x 960 GB M.2 D100P NVMe PCIe 4.0 SSD
- AOC8 x 800 Gb/s OSFP plus 2 x 10GbE RJ45
- GPU8 x NVIDIA B300 Blackwell Ultra, 2.3 TB HBM3e
- Power6 x 6,600 W Titanium Level, 3+3 redundant
Warranty and Onsite Service
3 Years Parts and Labour
Cover is the manufacturer warranty applicable to this product and region. Onsite service and extended terms can be added at quotation stage.
Configuration Summary
- CPUDual Intel Xeon 6768P, 128 cores total
- Memory4 TB DDR5-6400 ECC RDIMM (16 x 256 GB)
- Storage1 x 960 GB M.2 D100P NVMe PCIe 4.0 SSD
- AOC8 x 800 Gb/s OSFP plus 2 x 10GbE RJ45
- GPU8 x NVIDIA B300 Blackwell Ultra, 2.3 TB HBM3e
- Power6 x 6,600 W Titanium Level, 3+3 redundant
- Warranty3 Years Parts and Labour
Need help specifying this system?
Rack power, cooling capacity and network topology all change what will actually work on your site. Tell us the environment and we will come back with a configuration that fits it.
Talk to a specialistBuilt for the most demanding workloads
The NVIDIA HGX B300 NVL8 8-GPU Enterprise AI Server in the 4 TB configuration is a high-performance 8U system built around the NVIDIA HGX B300 platform. It is designed for large language model training and inference, generative AI, scientific computing, data analytics and other data-intensive enterprise workloads.
Eight NVIDIA B300 Blackwell Ultra GPUs are connected over fifth-generation NVLink and NVSwitch, giving up to 1.8 TB/s of GPU-to-GPU bandwidth across 2.3 TB of HBM3e memory. Dual Intel Xeon 6768P processors provide 128 cores, and eight 800 Gb/s OSFP ports handle east-west traffic in a scaled cluster.
Deployment considerations
This is a high-density platform. It draws approximately 15 kW at full load and rejects around 50,000 BTU/hr, so rack power, containment and cooling capacity should be confirmed before order. Configuration, condition and documentation are checked against the agreed specification before the system is released for shipment.
Related products
-
In Stock Now8U Server with 8x NVIDIA HGX B300 (AS -8126GS-NB3RT)
Starting at$833,440.43- Key Features/Applications: Large language model and generative AI training, conversational AI, scientific research, drug discovery, business intelligence and analytics
- CPU: Dual AMD EPYC 9005/9004 Series processors
- Chassis: 8U Rackmount
- Drive: 8 front hot-swap E1.S NVMe bays + 2 x M.2 NVMe
- RAM: 24 x DDR5 DIMM slots, up to 6 TB at 6400 MT/s
- GPU: NVIDIA HGX B300 8-GPU (8 x SXM, NVLink)
- Network Ports: 8 x OSFP 800GbE + 2 x 10GbE RJ45
-
In Stock Now8U Server with 8x NVIDIA H200 141GB (SYS-821GE-TNHR)
Starting at$476,420.46- Key Features/Applications: High performance computing, AI and deep learning training, conversational AI, industrial automation, retail, healthcare, drug discovery, finance and economics
- CPU: Dual 5th or 4th Generation Intel Xeon Scalable Processors
- Chassis: 8U Rackmount
- Drive: Up to 19 x 2.5 inch hot-swap NVMe, SATA or SAS bays plus 2 x M.2 NVMe slots
- RAM: 32 DDR5 DIMM slots
- GPU: NVIDIA HGX GPU baseboard, 8 x H200 141GB
- Network Ports: Flexible networking options
-
In Stock Now8U Server with 8x NVIDIA HGX B300 (SYS-822GS-NB3RT)
Starting at$799,732.58- Key Features/Applications: Conversational AI, high performance computing, scientific research, autonomous vehicle development
- CPU: Dual Intel Xeon 6700 Series processors with P-cores
- Chassis: 8U Rackmount
- Drive: 8 front hot-swap E1.S NVMe bays + 2 x M.2 NVMe
- RAM: 32 x DDR5 RDIMM slots, up to 8 TB
- GPU: NVIDIA HGX B300 8-GPU (8 x SXM, NVLink)
- Network Ports: 8 x OSFP 800GbE + 2 x 10GbE RJ45 + 1GbE dedicated BMC
-
In Stock Now4U Server with 4x NVIDIA RTX PRO 6000 Blackwell (SYS-422GA-NRT-01-G2)
Starting at$171,660.06- Key Features/Applications: High performance computing, AI and deep learning training, cloud gaming, design and visualisation, diagnostic imaging, VDI, media and video streaming, animation and modelling, 3D rendering
- CPU: Dual Intel Xeon 6960P Processors, 72 cores each
- Chassis: 4U, single node
- Drive: 1 x 960 GB M.2 NVMe plus 1 x 3.8 TB U.2 NVMe
- RAM: 16 x 64 GB DDR5-6400 RDIMM
- GPU: 4 x NVIDIA RTX PRO 6000 Blackwell Server Edition
- Network Ports: 2 x 10GbE RJ45 LAN ports
Reviews
There are no reviews yet.