





Supermicro AS-8126GS-NB3RT NVIDIA HGX B300 8-GPU AI Server – 8U, Dual AMD EPYC 9005, Gold Series
SKU: AS-8126GS-NB3RT
- 8× NVIDIA B300 SXM (Blackwell Ultra) on HGX B300 – 2.3 TB HBM3e, NVLink 1.8 TB/s per GPU
- Dual AMD EPYC 9005 – Gold preset: 2× EPYC 9575F (128 cores in total, 3.3 GHz base)
- 24 DIMM slots – up to 6 TB DDR5-6400 (Gold preset: 3 TB)
- 8× E1.S NVMe + 2× M.2 – Gold preset: 61 TB NVMe (8× 7.68 TB) + 2× 1.92 TB boot
- 8× 800G ConnectX-8 SuperNIC ports + up to 2× BlueField-3 (Gold: 2× dual-port 200G)
- Gold Series preset ships faster: 6–12 weeks · 6× 6600 W Titanium PSUs (3+3)
Price on request – B2B quote within 1 business day. Lead time: 6–12 weeks (Gold Series).
- GPUs8× NVIDIA B300
- GPU memory2.3 TB HBM3e
- CPUs2× AMD EPYC 9005
Estimated delivery 6–12 weeks
Order confirmed within 24hTracked deliveryTo all 27 EU countries
Warranty & supportAfter-sales help
B2B invoiceEU reverse-charge VAT
Request a quote
Configure and request your quote
Tell us the configuration, quantity and timeline. A hardware specialist checks availability and sends you a written quote.
- Written B2B quote within 1 business day, excl. VAT
- Estimated delivery 6–12 weeks
- Invoice payment; EU reverse charge for VAT-registered businesses
- NVIDIA data-centre GPUs are export-controlled: we ask for end-user details before an order
Thank you – your request is on its way
We have received your quote request and sent a confirmation to your e-mail. A hardware specialist will reply within 1 business day.
The Supermicro AS-8126GS-NB3RT pairs the NVIDIA HGX B300 8-GPU board with two AMD EPYC 9005 processors in an 8U air-cooled chassis. Its Gold Series version, AS-8126GS-NB3RT-01-G2, is a complete ready-to-deploy inference node: two EPYC 9575F CPUs, 3 TB of DDR5-6400, 61 TB of local NVMe, eight 800G ConnectX-8 ports and two dual-port 200G adapters – rack it, connect power and networking and start serving models.
With 288 GB of HBM3e per GPU and Blackwell Ultra's doubled attention performance, it is built for large-scale LLM inference, reasoning models and retrieval-augmented generation with big local datasets. Mineshop.eu supplies the Gold preset or a custom build with EU delivery and B2B invoicing.
Explore AI & GPU Servers and compare AS-A126GS-TNBR (HGX B200, AMD) with SYS-822GS-NB3RT (HGX B300, Intel).
Key features
- 8× NVIDIA B300 SXM (Blackwell Ultra) on HGX B300 – 2.3 TB HBM3e, NVLink 1.8 TB/s per GPU
- Dual AMD EPYC 9005 – Gold preset: 2× EPYC 9575F (128 cores in total, 3.3 GHz base)
- 24 DIMM slots – up to 6 TB DDR5-6400 (Gold preset: 3 TB)
- 8× E1.S NVMe + 2× M.2 – Gold preset: 61 TB NVMe (8× 7.68 TB) + 2× 1.92 TB boot
- 8× 800G ConnectX-8 SuperNIC ports + up to 2× BlueField-3 (Gold: 2× dual-port 200G)
- Gold Series preset ships faster: 6–12 weeks · 6× 6600 W Titanium PSUs (3+3)
Workloads and use cases
- Turn-key inference node: Gold preset ships with CPUs, 3 TB RAM, 61 TB NVMe and all NICs installed and pre-tested.
- RAG and vector search: 61 TB of local E1.S NVMe keeps embeddings, document stores and model weights next to the GPUs.
- Reasoning / long-context LLMs: 2.3 TB of HBM3e and 108 PFLOPS dense FP4 for large MoE and reasoning models.
- AMD-based AI clusters: EPYC 9575F high-frequency host CPUs with 8× 800G scale-out per node.
NVIDIA HGX B300 8-GPU performance
System-level figures published by NVIDIA for the eight-GPU board; real application performance depends on model, precision and software.
| GPU board | NVIDIA HGX B300 8-GPU |
|---|---|
| GPUs | 8× NVIDIA B300 SXM (Blackwell Ultra) |
| Memory per GPU | 288 GB HBM3e |
| Total GPU memory | 2.3 TB HBM3e (8 × 288 GB) |
| FP4 Tensor Core | 144 PFLOPS sparse / 108 PFLOPS dense |
| FP8 / FP6 Tensor Core | 72 PFLOPS (sparse) |
| INT8 Tensor Core | 3 POPS (sparse) |
| FP16 / BF16 Tensor Core | 36 PFLOPS (sparse) |
| TF32 Tensor Core | 18 PFLOPS (sparse) |
| FP32 | 600 TFLOPS |
| FP64 / FP64 Tensor Core | 10 TFLOPS |
| NVLink | 5th-gen NVLink + NVLink Switch, 1.8 TB/s per GPU, 14.4 TB/s total |
| Networking bandwidth (NVIDIA reference) | 1.6 TB/s |
Gold Series presets – faster delivery
Supermicro Gold Series systems are pre-configured and pre-tested best-sellers with key components already installed, which shortens lead times. For HGX systems the GPU allocation still applies – current lead time to Europe is about 6–12 weeks.
| Gold Series SKU | CPUs | Memory | Storage | AI fabric | Extra networking |
|---|---|---|---|---|---|
| AS-8126GS-NB3RT-01-G2 | 2× AMD EPYC 9575F (64 cores each, 3.3 GHz base) | 3 TB DDR5-6400 | 8× 7.68 TB E1.S NVMe + 2× 1.92 TB M.2 NVMe | 8× 800GbE / XDR800 (ConnectX-8) | 2× dual-port 200GbE / NDR200 |
AS-8126GS-NB3RT full specifications
| Model | Supermicro AS -8126GS-NB3RT (GPU A+ Server), Gold Series version -01-G2 |
|---|---|
| Form factor | 8U rackmount, chassis CSE-GP807TS-R000NP |
| Motherboard | Super H14DSG-OM |
| Processors | Dual AMD EPYC 9005 series; up to 384C/768T in total; up to 500 W TDP (air-cooled) |
| GPU | NVIDIA HGX B300 8-GPU (8× B300 SXM, 288 GB each) |
| GPU interconnect | NVIDIA NVLink with NVSwitch (GPU-GPU); PCIe 5.0 x16 CPU-to-GPU |
| Memory | 24 DIMM slots; up to 6 TB DDR5-6400 ECC RDIMM (1DPC); 1.1 V |
| Drive bays | 8× front hot-swap E1.S NVMe |
| M.2 | 2× M.2 NVMe (M-key) |
| Expansion slots | 2× PCIe 5.0 x16 FHHL (option: 2 additional FHHL with extra parts) |
| Networking | 8× OSFP 800G ports (NVIDIA ConnectX-8 SuperNIC), 2× RJ45 10GbE, dedicated BMC/IPMI port |
| Front I/O | 1× USB 3.0 + 1× USB 2.0 Type-A, 1× VGA, 1× mini-DisplayPort |
| Cooling | Up to 12 hot-swap heavy-duty fan modules |
| Power | 6× 6600 W redundant (3+3) Titanium Level (96 %) PSUs; 5900 W @ 200–207.9 VAC up to 6600 W @ 239–240 VAC |
| Security | TPM 2.0, Silicon Root of Trust (NIST 800-193), signed firmware, Secure Boot, secure firmware updates, automatic firmware recovery, remote attestation, system lockdown |
| Management | BMC/IPMI, SuperCloud Composer, Supermicro Server Manager, Super Diagnostics Offline, Thin-Agent Service, SuperServer Automation Assistant, DCMS |
| BIOS | AMI 64 MB UEFI |
| Certified OS (per Supermicro) | ESXi 8.0U3 / 9.0, Oracle Linux 9.6, RHEL 10.0 |
| Dimensions (H×W×D) | 356 × 449 × 950 mm (13.8 × 17.6 × 37.4 in) |
| Weight | 119 kg net / 137 kg gross |
| Shipping package | 670 × 730 × 1280 mm (26.38 × 28.74 × 50.39 in) |
| Operating environment | 10–35 °C, 8–90 % RH non-condensing |
Configuration options
Every HGX server is configured to order. Typical options are listed below; final compatibility and part numbers are confirmed in your quote.
| Component | Base | Options |
|---|---|---|
| Processors | 2× AMD EPYC 9005 | Gold: EPYC 9575F; custom: EPYC 9555 / 9655 / 9755 / 9965 (up to 192C per CPU) |
| Memory | 24 DIMMs | 768 GB – 6 TB DDR5-6400; Gold preset 3 TB |
| NVMe storage | 8× E1.S bays | Gold: 8× 7.68 TB (61 TB); custom 3.84 / 7.68 / 15.36 TB |
| Boot | 2× M.2 NVMe | Gold: 2× 1.92 TB (RAID 1 ready) |
| East-west AI fabric | 8× ConnectX-8 on board | XDR 800G InfiniBand or 800GbE – optics/cables quoted separately |
| North-south network | 2× FHHL slots (+2 optional) | Gold: 2× dual-port 200GbE/NDR200; BlueField-3 DPU options |
| Operating system | None / pre-installed | Ubuntu Server LTS, RHEL 10, Oracle Linux, VMware ESXi |
| Software | Optional | NVIDIA AI Enterprise, inference serving stack (vLLM / TensorRT-LLM / NIM), cluster management |
| Services | Optional | Burn-in test report, rack integration with 800G optics and switches, onsite installation, onsite warranty (3/5 years) |
Delivery and lead time
- Estimated lead time: 6–12 weeks for the Gold Series presets; custom configurations up to ~30 weeks depending on GPU allocation.
- Quote: written B2B quote excluding VAT within 1 business day, including the confirmed delivery window.
- Shipping: palletised freight to your data centre in all 27 EU countries; delivery to other countries on request and subject to export-control checks.
- Before shipping: optional burn-in test with report, firmware update and OS pre-installation.
Data-centre checklist
- Power: 6× 6600 W Titanium PSUs in 3+3 redundancy on 200–240 VAC – plan two independent feeds/PDUs. The exact power budget is part of the quote.
- Space: 950 mm deep 8U chassis – a rack at least 1200 mm deep.
- Weight: 119 kg net – use a lift for installation and check rack and floor load ratings.
- Cooling: front-to-back airflow, inlet air 10–35 °C; hot/cold-aisle containment recommended.
- Network: plan leaf switches, optics and cables for the GPU fabric (8× 800G per server) plus management and storage networks.
Buying from Mineshop.eu
- New systems sourced through European IT distribution, with manufacturer warranty and onsite service options.
- B2B invoicing excluding VAT, EU reverse charge for VAT-registered businesses, prices in EUR.
- Hardware specialists who help you size GPUs, memory, storage and networking for your models.
- One supplier for the complete stack: servers, GPUs, 400G/800G networking, server memory and NVMe storage.
Ordering process
- Click Request a quote and tell us the configuration, quantity, delivery country and timeline.
- A hardware specialist confirms the configuration with you – by email or a short call.
- You receive a written quote within 1 business day, with lead time, service plan and payment terms.
- You confirm with a purchase order; we handle export-control and end-user paperwork.
- Your server is built, tested and delivered; optional rack integration and onsite installation.
Compare Supermicro HGX servers
| Model | GPU board | GPU memory | CPUs | Chassis | Lead time |
|---|---|---|---|---|---|
| SYS-A22GA-NBRT | HGX B200 8-GPU | 1.4 TB HBM3e | 2× Intel Xeon 6900P | 10U air-cooled | On quote (custom build) |
| AS-A126GS-TNBR | HGX B200 8-GPU | 1.4 TB HBM3e | 2× AMD EPYC 9005/9004 | 10U air-cooled | On quote (custom build) |
| SYS-822GS-NB3RT | HGX B300 8-GPU | 2.3 TB HBM3e | 2× Intel Xeon 6700P | 8U air-cooled | Gold Series: 6–12 weeks |
| AS-8126GS-NB3RT (this server) | HGX B300 8-GPU | 2.3 TB HBM3e | 2× AMD EPYC 9005 | 8U air-cooled | Gold Series: 6–12 weeks |
Frequently asked questions
What is the price of the Supermicro AS-8126GS-NB3RT?
We quote the AS-8126GS-NB3RT per configuration – the fully-loaded Gold preset (-01-G2) or a custom build. HGX B300 pricing moves with GPU allocation, so you receive a written B2B quote, excluding VAT, within 1 business day of your request.
What is included in the AS-8126GS-NB3RT-01-G2 Gold Series preset?
Two AMD EPYC 9575F CPUs (128 cores in total), 3 TB DDR5-6400, eight 7.68 TB E1.S NVMe drives, two 1.92 TB M.2 NVMe boot drives, eight 800G ConnectX-8 ports, two dual-port 200G adapters and six 6600 W Titanium power supplies – pre-installed and pre-tested.
Is the AS-8126GS-NB3RT better than the SYS-822GS-NB3RT?
Both use the same HGX B300 8-GPU board and 8U chassis. Choose the AS-8126GS-NB3RT for AMD EPYC 9005 hosts and the storage-heavy Gold preset; choose the SYS-822GS-NB3RT for Intel Xeon 6700P hosts, up to 8 TB of memory (32 DIMMs) and RHEL AI certification.
How fast can it be delivered?
The Gold Series preset currently has the shortest lead time – about 6–12 weeks to Europe. Custom builds can take up to ~30 weeks depending on GPU allocation. Your quote confirms the date.
What does it need in the data centre?
A 200–240 V supply for six 6600 W Titanium PSUs (3+3), 8U in a rack at least 1200 mm deep, a rack rated for 119 kg and 10–35 °C inlet air. Power budget, cords and optics are listed in the quote.
Can you deliver several servers as a cluster?
Yes. We quote multi-node HGX B300 clusters including 800G InfiniBand or Spectrum-X Ethernet switches, optics and cabling, rack integration and onsite installation.
NVIDIA, HGX, NVLink, ConnectX and BlueField are trademarks of NVIDIA Corporation. Supermicro is a trademark of Super Micro Computer, Inc. Specifications from the manufacturers; subject to change without notice.
Warranty
Terms of Sale
Shipping Information


