Who We Serve
India's value-for-money IT hardware brand
14 years. 28,000 sq ft. 100,000+ devices. From desktops to data center — designed, engineered, and manufactured in India.
Why choose RDPProducts
Desktops to data center, all Make in India
14 product categories across compute, AI, and data center. Deployment-ready from our 28,000 sq ft facility.
Download Product CatalogAI Solutions
Sovereign AI infrastructure
End-to-end AI compute under one sovereign umbrella. Designed here. Manufactured here. Supported here.
Talk to a Solutions Architect
Support
SLA-driven. Not ticket-driven.
Warranty. SLA. On-site service. Account management. Every commitment documented, every response time defined.
Download SLA Commitment
Company
Built on process, not promises
ISO 9001. PLI 2.0. SOP-led manufacturing. The systems behind every device we ship.
Our Story
R Series · Enterprise AI · 8U HGX Flagship
RDP-AS8126GS-NB3PT-01
8U Flagship AI Training Server with 8× NVIDIA HGX B300 GPUs, Fifth-Gen NVLink, Dual EPYC 9575F & 3TB DDR5-6400
Key Applications
- Foundation Model Training (GPT/Llama/Mixtral class)
- Multi-Trillion Parameter LLMs
- Vision-Language Models (CLIP, BLIP)
- Diffusion & Generative AI Training
- Reinforcement Learning at Scale
- Multi-Modal AI & Foundation Models
- Sovereign AI & Frontier Research
- Pre-training, Fine-tuning, RLHF
Key Features
- 8× NVIDIA HGX B300 SXM GPUs with fifth-gen NVLink fabric (1.8 TB/s GPU-GPU bandwidth)
- 1.4TB HBM3e aggregate GPU memory across the 8-GPU complex
- Dual AMD EPYC 9575F processors (Zen 5) — 128 cores total, optimized for AI host duties
- 3TB DDR5-6400 ECC system memory across 24 DIMM slots (12-channel per CPU)
- 8× NVIDIA ConnectX-8 XDR800 InfiniBand for cluster scale-out (one per GPU)
- Air-cooled and liquid-cooled (DLC) configurations · 3/3/1 warranty · Make in India
RDP-AS8126GS-NB3PT-01
Product SKU
| Model | RDP-AS8126GS-NB3PT-01 |
|---|---|
| Platform Reference | Supermicro AS-8126GS-NB3PT (HGX B300 reference design) |
| Origin | Designed and integrated in India by RDP Technologies Limited, Hyderabad |
GPU
| GPU Configuration |
|
|---|---|
| GPU Memory | 1.4TB HBM3e total (180GB per GPU × 8) |
| GPU-GPU Interconnect | NVLink (fifth gen) · 1.8 TB/s GPU-GPU bandwidth · all-to-all NVLink fabric |
| FP8 / FP4 Compute | Up to 144 PFLOPS FP8 · 288 PFLOPS FP4 dense per system |
| Cooling | Air-cooled standard · Direct Liquid Cooling (DLC) optional |
Processor
| CPU |
|
|---|---|
| Core Count | 64 cores per CPU · 128 cores total · 256 threads with SMT |
| Clock Speed | 3.3 GHz base · up to 5.0 GHz boost (frequency-optimized for AI host) |
| TDP | 400W per socket |
System Memory
| Memory |
|
|---|
Storage
| Boot Drive | 2× 1.92TB M.2 NVMe (RAID 1 boot) |
|---|---|
| Hot-Swap Bays | 16× 2.5" U.2/U.3 NVMe (PCIe 5.0) for AI training dataset cache |
| Max Raw Capacity | Up to 491TB hot-tier NVMe for training data |
| RAID | Software RAID · Optional hardware RAID HBA |
Networking & Cluster Fabric
| Compute Fabric | 8× NVIDIA ConnectX-8 XDR800 InfiniBand (one per GPU) — for multi-node cluster scale-out |
|---|---|
| Storage / Mgmt LAN | 2× 25GbE SFP28 onboard · 1× dedicated 1GbE IPMI |
| AIOM Slot | 1× AIOM (OCP 3.0) — additional 200/400GbE Ethernet or NDR InfiniBand |
| USB / Serial | 2× USB 3.0 rear · 1× VGA · 1× COM |
Expansion Slots
| PCIe Slots | 8× PCIe 5.0 x16 FHFL (one ConnectX-8 per slot, pre-populated) |
|---|---|
| Additional PCIe | 2× PCIe 5.0 x16 LP for storage / management cards |
| AIOM | 1× AIOM (OCP 3.0) for additional networking |
| M.2 | 2× PCIe 4.0 x4 · M.2 2280/22110 boot drives |
System BIOS & Management
| BIOS Type | AMI 256MB UEFI |
|---|---|
| Management |
|
Security
| Hardware | TPM 2.0 onboard · Silicon Root of Trust (NIST 800-193 compliant) |
|---|---|
| Security Features |
|
Chassis & Physical
| Form Factor | 8U Rackmount (flagship AI training chassis) |
|---|---|
| Dimensions (W × D × H) | 438 × 925 × 350 mm (17.2" × 36.4" × 13.8") |
| Weight | Approximately 105 kg (231 lbs) fully populated |
| Available Color | Black |
System Cooling
| Cooling Modes | Air-Cooled (standard) or Direct Liquid Cooling (DLC) (option) |
|---|---|
| Fans | 10× hot-swap counter-rotating cooling fans |
| Operating Temperature | 10°C to 32°C (50°F to 90°F) — AI-optimized thermal envelope |
| Operating Humidity | 8% to 90% (non-condensing) |
Power Supply
| Configuration | 6× 5250W Titanium Redundant (4+2) hot-swap |
|---|---|
| Total Capacity | 21,000W (4 active + 2 redundant) |
| Efficiency | 96% at 50% load · 80 PLUS Titanium certified |
| Input Voltage | 200–240V AC · 50–60Hz |
OS Support, Compliance & Warranty
| Operating Systems |
|
|---|---|
| AI Software Stack | NVIDIA CUDA · cuDNN · NCCL · PyTorch · TensorFlow · JAX · TensorRT-LLM (pre-installed) |
| Compliance & Certifications | BIS · CE · FCC · RoHS · UL · ISO 9001 · MeitY Recognised · PLI 2.0 · GeM Listed |
| Warranty | 3 years parts · 3 years labor · 1 year cross-ship (3/3/1 standard) |
| Support | PAN-India SLA-backed service · 24/7 priority support available · AI cluster commissioning service |
Download technical resources, manuals, and quick-reference guides for RDP-AS8126GS-NB3PT-01.
Request a quote for RDP-AS8126GS-NB3PT-01
Tell us your AI workload — foundation model training, multi-node cluster build, or sovereign AI deployment. Our AI architects will spec the right configuration and provide a quote with locked pricing and SLA commitment in writing.
Standard lead time: 5 business days for pre-validated configurations. Cluster deployments planned in coordination with your data center team.