8 GPU HGX B200 AI server
Eight NVIDIA B200 GPUs with 180 GB HBM3e each, NVLink and NVSwitch, 800 Gb/s or 400 Gb/s fabric ports, in an 8U air cooled chassis, build to order.
GPU: NVIDIA HGX B200
- GPUs
- 8 x NVIDIA B200, 180 GB HBM3e each, NVLink and NVSwitch
- CPU
- 2 x AMD EPYC 9005 or 2 x Intel Xeon 6
- Memory
- Up to 3 TB DDR5 RDIMM
- Storage
- 2 x M.2 boot, up to 8 x U.2 or E3.S NVMe
- Networking
- 8 x 800 Gb/s ConnectX-8 (east west), 2 x 400 Gb/s BlueField-3 (north south)
- Power
- Redundant PSUs, about 14 kW per node planning figure
Overview
HGX B200 is the proven Blackwell eight GPU platform and is available to build without allocation. Each B200 has 180 GB of HBM3e, and the eight GPUs share an NVLink and NVSwitch domain at 1.8 TB/s per GPU, so the node holds 1.4 TB of GPU memory. It runs the same software stack as HGX B300 and joins the same 800 Gb/s or 400 Gb/s fabrics.
Choose it when a cluster is already tuned for Blackwell, when inference memory per GPU is not the binding constraint, or when delivery timing matters more than the extra dense FP4 throughput of B300. The node is offered with ConnectX-8 at 800 Gb/s for new fabrics and with ConnectX-7 at 400 Gb/s for sites that are extending an NDR InfiniBand or 400 GbE installation.
Host options match the rest of the HGX range: dual AMD EPYC 9005 or Intel Xeon 6, DDR5 memory up to 3 TB, mirrored boot and up to eight NVMe data drives, BlueField-3 frontend. Plan for about 14 kW per node and eight rack units. Burn in includes NVLink topology checks and a fabric loopback test, documented in the acceptance report.
Use cases
- LLM training clusters
- Inference serving fleets
- Extending existing Blackwell deployments
- Fine tuning and RLHF pipelines
Specifications
| GPUs | 8 x NVIDIA B200 on the HGX B200 baseboard |
|---|---|
| GPU memory | 180 GB HBM3e per GPU, 1.4 TB per node |
| GPU interconnect | Fifth generation NVLink, 1.8 TB/s per GPU, NVSwitch |
| FP4 performance | 144 PFLOPS sparse, 72 PFLOPS dense per node |
| CPUs | 2 x AMD EPYC 9005 series or 2 x Intel Xeon 6 |
| System memory | 24 or 32 x DDR5 RDIMM, up to 3 TB |
| Boot | 2 x M.2 NVMe, mirrored |
|---|---|
| Data | Up to 8 x U.2 or E3.S NVMe, 7.68 TB to 30.72 TB each |
| RAID | Software RAID or optional hardware RAID for boot volume |
| GPU fabric | 8 x 800 Gb/s ConnectX-8, OSFP; 8 x 400 Gb/s ConnectX-7 available for existing NDR fabrics |
|---|---|
| Frontend | 2 x 400 Gb/s, BlueField-3 or ConnectX-7 |
| Management | Dedicated 1 GbE BMC port, TPM 2.0 |
| Form factor | 8U rackmount, air cooled; 10U variants available |
|---|---|
| Power supplies | Redundant 3,000 W class PSUs, 200 to 240 V input |
| Planning power | About 14 kW per node |
| GPU power | Configurable up to 1,000 W per GPU |
Figures describe the system class. Exact configuration, weights and power draw are confirmed on the quotation.
What's included
- Rails, power cords and cable management arm
- Burn in with multi hour GPU stress and NVLink validation, acceptance report
- NVIDIA end user registration handled with the order
- Operating system imaging, driver and fabric manager baseline on request
- 3 year warranty, options to 5 years with advance replacement
Compliance
Related
Related systems
8 GPU HGX B300 AI server, air cooled
Allocation required- GPUs
- 8 x NVIDIA B300, 288 GB HBM3e each
- CPU
- 2 x AMD EPYC 9005 or 2 x Intel Xeon 6
- Memory
- Up to 3 TB DDR5 RDIMM across 24 or 32 slots
- Chassis
- 10U, about 14 kW, air cooled
8 GPU HGX H200 AI server
Available on request- GPUs
- 8 x NVIDIA H200 SXM, 141 GB HBM3e each
- CPU
- 2 x AMD EPYC 9005 or 2 x Intel Xeon 6
- Memory
- Up to 3 TB DDR5 RDIMM
- Chassis
- 8U, about 10 kW, air cooled
8 node HGX B300 AI pod, 64 GPUs
Allocation required- GPU nodes
- 8 x 8 GPU HGX B300 servers, 64 x B300
- Fabric
- 64 x 800 Gb/s ports, non blocking
- CPU and control
- 2 x CPU nodes for data services
- Chassis
- Multi rack, about 125 kW, air cooled
ConnectX-7 400 Gb/s adapter
- Speed
- 400 Gb/s single port, or 2 x 200 Gb/s
- Host interface
- PCIe Gen5 x16
- Offloads
- RDMA and RoCE, GPUDirect RDMA
- Form factor
- PCIe HHHL card with OSFP or QSFP112 port
8 node HGX B200 AI pod, 64 GPUs
- GPU nodes
- 8 x 8 GPU HGX B200 servers, 64 x B200
- Fabric
- 64 x 800 Gb/s ConnectX-8 ports
- CPU and control
- 2 x CPU nodes for data services
- Chassis
- Multi rack, about 125 kW, air cooled
Related guides
Buying data center GPUs, export and registration
The compliance steps that come with every HGX, H200 NVL, Server Edition and NVL72 purchase, explained so your order ships on time.
In this guide
- Why a GPU purchase comes with paperwork
- Which products are covered
HGX vs PCIe GPU servers
The real differences between an HGX node and a PCIe GPU server, and a simple rule for choosing between them.
In this guide
- Two kinds of 8 GPU server
- The interconnect is the point
Sizing CPU, memory and storage around GPUs
Six rules of thumb that keep the GPUs fed, with the arithmetic for an 8 GPU node, a PCIe inference server and a 16 node training pod.
In this guide
- The GPUs are the expensive part, so feed them
- The six rules
Compare it with the next class up or down
The nearest classes we quote are the 8 GPU HGX B300 AI server, air cooled and the 8 GPU HGX H200 AI server. If neither is quite right, tell us the workload and we specify the system between them.
- (858) 613-1818
- sales@computermodules.com
- Monday to Friday, 8am to 5pm Pacific