MGX 8 GPU AI server
NVIDIA MGX reference design server with eight PCIe GPUs, RTX PRO 6000 Blackwell Server Edition or H200 NVL, BlueField-3 and ConnectX-7 networking, in 4U.
GPU: NVIDIA RTX PRO 6000 Blackwell Server Edition or H200 NVL
- GPUs
- 8 x RTX PRO 6000 Blackwell Server Edition or 8 x H200 NVL, PCIe Gen5
- CPU
- 2 x AMD EPYC 9005 or 2 x Intel Xeon 6
- Memory
- Up to 2 TB DDR5 RDIMM
- Storage
- 2 x M.2 boot, up to 8 x E1.S or U.2 NVMe
- Networking
- 4 x 400 Gb/s ConnectX-7 (east west), 1 x BlueField-3 DPU (north south)
- Design
- NVIDIA MGX modular architecture, validated GPU, DPU and NIC placement
Overview
NVIDIA MGX is a modular reference architecture that fixes the placement of GPUs, DPUs and network adapters so that systems from different builders behave the same way and qualify for NVIDIA certified status. This eight GPU MGX node follows that design with RTX PRO 6000 Blackwell Server Edition or H200 NVL cards, a BlueField-3 DPU on the frontend and ConnectX-7 adapters for the GPU fabric.
The practical benefit is predictable topology: every GPU pair has a direct PCIe path to its adapter, the DPU offloads storage, networking and security from the host, and the layout is already validated for NVIDIA AI Enterprise and for cloud native inference stacks. It is a good fit for service providers and enterprises that will run many nodes and want them identical.
Configuration follows the PCIe server classes: dual AMD EPYC 9005 or Intel Xeon 6, DDR5 to 2 TB, mirrored boot and E1.S or U.2 data drives, about 6.5 kW per node in four rack units. Choose RTX PRO 6000 for FP4 throughput and video engines, or H200 NVL for HBM capacity and NVLink bridges.
Use cases
- NVIDIA AI Enterprise deployments
- Cloud and service provider inference
- Standardized multi node fleets
- DPU offloaded storage and security
Specifications
| GPUs | 8 x NVIDIA RTX PRO 6000 Blackwell Server Edition (96 GB GDDR7) or 8 x H200 NVL (141 GB HBM3e) |
|---|---|
| GPU interconnect | PCIe Gen5 switches with balanced GPU to NIC pairing per the MGX design; NVLink bridges with H200 NVL |
| CPUs | 2 x AMD EPYC 9005 series or 2 x Intel Xeon 6 |
| System memory | 24 or 32 x DDR5 RDIMM, up to 2 TB |
| Boot | 2 x M.2 NVMe, mirrored |
|---|---|
| Data | Up to 8 x E1.S or U.2 NVMe |
| Expansion | Shared storage over the DPU frontend network |
| GPU fabric | 4 x 400 Gb/s ConnectX-7, InfiniBand or RoCE, one per GPU pair |
|---|---|
| Frontend | 1 x BlueField-3 DPU, 2 x 200 Gb/s or 1 x 400 Gb/s, with offload for storage and security |
| Management | Dedicated 1 GbE BMC port, TPM 2.0 |
| Form factor | 4U rackmount, air cooled, MGX compliant |
|---|---|
| Power supplies | Redundant 2,700 W class PSUs, 200 to 240 V input |
| Planning power | About 6.5 kW per node |
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 RTX PRO 6000 PCIe AI server
- GPUs
- 8 x RTX PRO 6000 Blackwell Server Edition
- CPU
- 2 x AMD EPYC 9005 or 2 x Intel Xeon 6
- Memory
- Up to 2 TB DDR5 RDIMM
- Chassis
- 4U, about 6.5 kW, air cooled
8 GPU H200 NVL PCIe AI server
Available on request- GPUs
- 8 x NVIDIA H200 NVL, 141 GB HBM3e each
- CPU
- 2 x AMD EPYC 9005 or 2 x Intel Xeon 6
- Memory
- Up to 2 TB DDR5 RDIMM
- Chassis
- 4U, about 6.5 kW, air cooled
BlueField-3 DPU, 400 Gb/s
- Speed
- 400 Gb/s, 1 x 400 Gb/s or 2 x 200 Gb/s
- Compute
- Arm cores with onboard DDR memory running the DOCA software stack
- Offloads
- Software defined networking
- Host interface
- PCIe Gen5 x16
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
Choosing a GPU server for LLM inference
A practical method for sizing an inference server from the model, the quantization and the concurrency you expect, with worked examples.
In this guide
- Start with the model, not the server
- Step 1: how much GPU memory the weights need
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
How to request a quotation
A short guide to getting a complete, accurate quotation, from the details we need to what the quotation contains.
In this guide
- Why there are no prices on this site
- Three ways to ask
Compare it with the next class up or down
The nearest classes we quote are the 8 GPU RTX PRO 6000 PCIe AI server and the 8 GPU H200 NVL PCIe 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