Supermicro AS-A126GS-TNBR 10U GPU System
$270,000.00
Supermicro AS-A126GS-TNBR is a 10U air-cooled system supporting NVIDIA HGX B200 eight-GPU hardware and dual AMD EPYC 9005/9004 processors. It offers 24 DDR5 DIMM slots, PCIe 5.0 expansion and alternative storage layouts. Confirm the CPU generation, installed memory, drives and network adapters in the supplied configuration.
Explore Related
Related products, categories and research.
Supermicro AS-A126GS-TNBR is a 10U, air-cooled GPU system built for an NVIDIA HGX B200 eight-GPU platform and two AMD EPYC processors. Its host platform supports the EPYC 9005 and 9004 series. The processor generation affects the documented memory-speed limit, so the exact CPU selection matters when comparing configurations.
This is the standard AS-A126GS-TNBR model, recorded under seller SKU ARS-SRV-SUPERM-A126GSTNBR10U-0284. It should not be confused with an Intel-based SYS model or a separately configured G1 Gold Series package. A similar chassis photograph does not establish the same host hardware or supplied components.

Supermicro AS-A126GS-TNBR specifications
| Item | Platform specification |
|---|---|
| Chassis | 10U rackmount; air-cooled GPU platform |
| Motherboard | Supermicro H14DSG-OD |
| Processors | Dual AMD EPYC 9005/9004 series; up to 500 W TDP CPUs |
| Accelerators | NVIDIA HGX B200 eight-GPU SXM platform, 180 GB per GPU |
| GPU interconnect | NVLink with NVSwitch; PCIe 5.0 x16 CPU-to-GPU interconnect |
| Memory | 24 DDR5 RDIMM slots; up to 6 TB at 1 DIMM per channel |
| Memory speed limits | Up to 6400 MT/s with EPYC 9005; up to 4800 MT/s with EPYC 9004 |
| Expansion | 8 PCIe 5.0 x16 low-profile slots and 2 PCIe 5.0 x16 FHHL slots |
| Default front bays | 8 hot-swap 2.5-inch NVMe bays plus 2 hot-swap 2.5-inch SATA bays |
| Alternative storage | Optional 8 E1.S bays; 2 M.2 NVMe slots |
| Onboard networking | Dual 10GbE RJ45 with Intel X710; dedicated 1GbE BMC port |
| Power supplies | 6 × 5250 W Titanium; 3+3 redundancy subject to configuration and load |
Platform limits are not the installed bill of materials. Confirm the CPUs, accelerators, DIMMs, drives and adapters in the offered system. Available slots and bays do not establish populated capacity.
AMD EPYC host and memory choices
AS-A126GS-TNBR supports EPYC 9005 and 9004 processors with different published memory-speed limits. Match the processor models, approved DIMMs, population rules and firmware to one supported configuration. The maximum memory figures assume one DIMM per channel; a capacity and speed should not be combined from incompatible configurations.
Host DDR5 serves CPU-side work, including preprocessing and data staging. It is separate from B200 accelerator memory. Size both resources for the intended application.
AS-A126GS-TNBR versus related GPU servers
| Model | Host and accelerator distinction |
|---|---|
| AS-A126GS-TNBR | AMD EPYC host with NVIDIA HGX B200; this listing identifies the standard model. |
| AS-A126GS-TNMR | Separate AMD Instinct MI355X accelerator model. |
| SYS-A22GA-NBRT | Intel Xeon host with HGX B200. |
A G1 Gold Series offer specifies a separate component package. Do not assume its installed CPUs, memory or accessories are supplied with the standard model.
NVMe/SATA or E1.S storage layout
The default eight NVMe and two SATA bays form one front arrangement. Optional eight-bay E1.S storage is an alternative, not eight extra bays. Confirm the actual backplane, controller, cabling, carriers and approved drives. Connector shape and nominal interface alone do not establish compatibility.
Specify the two M.2 slots’ boot arrangement separately from application storage. Installed drives and RAID configuration require confirmation. Define protection and recovery for datasets, checkpoints and outputs.
GPU links and external fabric
NVLink and NVSwitch connect the eight B200 GPUs locally. External fabric adapters and switches connect servers. Onboard dual 10GbE does not establish an included 400G cluster network; identify the adapters, switches and compatible cabling in the offer.
Allocate add-in cards to supported low-profile or FHHL slots, checking electrical width, cooling and clearance. The dedicated BMC port provides a separate management path; verify optional management software entitlements.
Application capacity depends on precision, parallelisation, software and runtime memory, including cache and training state. No measured benchmark of the offered unit is claimed.
Power and delivery scope
The 3+3 power-supply redundancy qualification depends on configuration and application load. Plan supported power feeds and maintenance accordingly. The sum of supply ratings is not typical or measured consumption. Check rack depth, handling access, service clearance and cooling against the installation instructions.
What should an AS-A126GS-TNBR quotation identify?
Request the configured CPUs, memory, drives, fabric hardware, rails and cables. Record unit condition, software rights and service coverage in the written offer. Reference images and manufacturer options do not prove those items are included.
Manufacturer information
Use Supermicro’s exact-model specifications for supported limits and options. Browse related AI Servers or submit a configuration enquiry through the contact page.
Recently viewed
Add AI Robot Supplier as a preferred source on Google.