DRACO 8× B300 SXM GPU Server

Overview

The DRACO 8× B300 SXM GPU Server is RDP’s flagship single-node AI server — an Rack 10U HGX B300 (Blackwell Ultra) platform with eight NVIDIA B300 SXM (HGX B300, Blackwell Ultra) GPUs delivering 2,304 GB HBM3e of HBM3e on one NVLink+NVSwitch baseboard. It is the most GPU memory and the highest single-node training and inference capacity RDP builds, sized for trillion-parameter-class frontier models — on-premises, behind your firewall, in INR, on a GST invoice.

Engineered for organisations building national- or enterprise-scale AI capability in-house, it pairs the eight Blackwell Ultra GPUs with a 2× Intel Xeon 6 host, 4 TB DDR5 ECC and 120 TB NVMe, with 8× XDR InfiniBand for scale-out, liquid cooling, redundant power and full BMC/IPMI for lights-out operation.

Key highlights

• 2,304 GB HBM3e of HBM3e across 8× B300 SXM — the largest single-node GPU-memory pool RDP offers, for trillion-parameter-class training and serving.
• NVLink + NVSwitch fabric — full all-to-all bandwidth across all eight Blackwell Ultra GPUs for the most demanding tensor-parallel models.
• Blackwell Ultra FP4/FP8 — the current generation’s highest throughput for training and high-efficiency inference.
• 2× Intel Xeon 6 + 4 TB DDR5 ECC — high core count and memory bandwidth to feed eight flagship GPUs.
• Rack 10U, liquid-cooled, redundant PSU, BMC/IPMI — sustained clocks under full load, lights-out management.
• 120 TB NVMe + 8× InfiniBand XDR — very large local storage and the highest-bandwidth scale-out fabric; no egress fees.
• On-prem data sovereignty — training data and weights stay in-house; DPDP-friendly, air-gappable.
• Make-in-India OEM — predictable INR pricing, GST tax invoice (HSN 8471), pan-India onsite support, GeM-procurable.

AI workload fit (what it actually runs — honestly)

• Frontier training & fine-tuning: full pre-training and fine-tuning of trillion-parameter-class models, tensor- and pipeline-parallel across the eight NVSwitch-linked Blackwell Ultra GPUs.
• Inference: highest-efficiency FP4/FP8 serving of the very largest models, or many large models concurrently.
• RAG, vision, multimodal & agentic: the most demanding production pipelines on the 120 TB NVMe and large multi-agent systems.
• Engineering note: all eight SXM GPUs share an NVLink+NVSwitch fabric for full all-to-all bandwidth, and Blackwell Ultra’s 288 GB per GPU is what makes trillion-parameter single-node work practical; for scale beyond one node, units link over the 8× XDR InfiniBand fabric.

AI workload positioning

This is the top of single-node train-and-serve capability. With 2,304 GB HBM3e of HBM3e on an NVLink+NVSwitch fabric and 8× XDR InfiniBand, it is sized to sustain frontier-class training and the highest-efficiency inference on-prem — the owned alternative to renting the largest Blackwell cloud nodes continuously.

Industry use cases

• BFSI — train and serve the largest private models under data-residency rules.
• Healthcare & life sciences — on-prem training of large medical and imaging models, PHI in-house.
• Government & national programmes — sovereign frontier-class AI on GeM-procurable infrastructure.
• Neocloud / AI providers — the flagship building block for a premium GPU cloud.
• Large enterprise & telecom — national-scale in-house AI capability.
• Research & national labs — a flagship frontier training node.

Performance — and how to be sure

We don’t publish inflated peak numbers. The honest picture: 2,304 GB HBM3e of HBM3e across eight NVSwitch-linked Blackwell Ultra GPUs is sized for trillion-parameter-class training and the highest-efficiency serving on-prem. Want certainty? Request a free benchmark of your model and dataset on this exact configuration before you buy — we’ll send back real tokens/sec and training/fine-tune timings for your workload.

Series & upgrade path

• DRACO (flagship training tier) — this, RDP’s highest single-node platform.
• Step up in scale: for multi-rack systems, move to RDP Rack-Scale AI Systems and AI SuperClusters built from these nodes — talk to an architect about the fabric.
• Lower tiers: 4× / 8× B200 and H200 nodes for smaller budgets and footprints.

On-prem vs cloud — the TCO case

For sustained frontier-class training, owning beats renting decisively: an always-on flagship Blackwell Ultra cloud node is among the largest line items in any AI budget, and on-prem removes egress fees and keeps data and weights in-house. RDP pricing is fixed in INR with a GST input-credit-eligible invoice — ask for a 3-year TCO comparison.

Software & day-one readiness

Ships pre-configured to train and serve: NVIDIA driver, CUDA, cuDNN, NCCL, the InfiniBand stack, Docker and NVIDIA Container Toolkit, with PyTorch, vLLM / Triton / TensorRT-LLM on Ubuntu LTS. Optional Slurm/Kubernetes, managed AI-stack and observability setup available.

Power, cooling & rack integration

An Rack 10U liquid-cooled node with very high power draw — plan CDU/manifold and facility water, dedicated power and XDR InfiniBand cabling. (Exact PSU rating, BTU, flow, fabric cabling and rack-depth figures confirmed on the build sheet.) BMC/IPMI provides remote power, console and health monitoring.

Deployment, warranty & support

• Made to order, built, racked-and-stacked, cabled and burned-in in India; realistic lead time confirmed at quote.
• In the box: server, rails, power cables, quick-start, and the pre-installed AI software stack; fabric switches and cabling scoped at quote.
• Onsite warranty + AMC with pan-India coverage and an RMA/escalation path (exact term & response window confirmed at quote).

Why RDP

14 years of Make-in-India infrastructure and 300,000+ devices shipped. Indian OEM, INR pricing, GST tax invoice (HSN 8471), pan-India onsite engineers, GeM availability, and DPDP / sovereign-AI-ready deployment.

Buy with confidence

This is RDP’s flagship single-node AI server, made to order — talk to an RDP solution architect, get a configuration and 3-year TCO tailored to your workload, and benchmark your own model on it before you commit. Request a quote to begin.

Specifications