DRACO 8192× MI300A AI Supercomputer

Overview

The DRACO 8192× MI300A AI Supercomputer is a turnkey, liquid-cooled exascale-class HPC + AI supercomputer built on 8192 AMD Instinct MI300A APUs — the same accelerator architecture behind the world’s leadership-class FP64 systems. It delivers ~1.0 PB unified HBM3 of unified accelerator memory across a data hall with a non-blocking InfiniBand spine, fusing scientific simulation and frontier AI in one machine — on-premises, in INR, on a GST invoice.

Delivered as a single national-scale engagement, RDP co-designs the reference architecture, integrates, cools and burns it in, and hands over one validated supercomputer with one warranty and one support contract. Its defining strength: FP64 HPC leadership and AI fused in one APU with unified CPU-GPU memory, removing host-device copies.

Key highlights

• 8192× MI300A · ~1.0 PB unified HBM3 aggregate — exascale-class accelerator memory for the largest simulations and frontier AI.
• Infinity Fabric + non-blocking InfiniBand — AMD Infinity Fabric across the APUs and a non-blocking InfiniBand spine between nodes.
• Unified CPU-GPU APU memory — Zen4 CPU and CDNA GPU share one HBM3 pool, eliminating host-device transfers for HPC + AI.
• FP64 HPC leadership — the architecture of the world’s top FP64 supercomputers, with mixed-precision AI on the same nodes.
• 128 PB parallel NVMe parallel filesystem — exascale-grade storage for datasets, checkpoints and simulation output.
• Data hall, liquid-cooled, turnkey — delivered, integrated and validated; one engagement, one warranty.
• On-prem data sovereignty — data, models and codes stay in-country; 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)

• HPC + simulation: leadership-class FP64 scientific computing (CFD, climate, molecular dynamics, finite-element) at national scale.
• Frontier AI training: distributed training of the largest models across the 8192 APUs with 3D parallelism.
• Large-scale inference & fine-tuning: serve and fine-tune many large models in parallel.
• Engineering note: this is an AMD Instinct system — it runs the open ROCm/HIP software stack, not CUDA; we validate your HPC codes and AI models on ROCm before you commit. At exascale, network topology and parallelism strategy decide real performance; we run a scaling test on your codes, not a peak number.

AI workload positioning

This sits at the top of the HPC + foundation-model convergence: an exascale-class supercomputer for both scientific simulation and frontier AI. With ~1.0 PB unified HBM3 of unified accelerator memory on a non-blocking fabric, it is sized to sustain national-scale HPC/AI on-prem — the sovereign, owned alternative to hyperscale capacity.

Industry use cases

• Government & national labs — a sovereign exascale HPC + AI system on GeM-procurable infrastructure.
• Weather, climate & energy — national-scale FP64 simulation plus AI surrogates.
• Pharma & life sciences — large-scale molecular dynamics, drug discovery and AI.
• Aerospace, automotive & manufacturing — leadership CFD/FEA plus AI design.
• National research programmes — a shared exascale supercomputer.
• Sovereign neoclouds — a converged HPC+AI service at national scale.

Performance — and how to be sure

We don’t publish inflated peak numbers. The honest picture: 8192 MI300A APUs (~1.0 PB unified HBM3) on a non-blocking fabric are sized for exascale-class converged HPC + AI. Want certainty? Request a free benchmark — your HPC codes and AI models, including a scaling test — on a representative configuration before you commit; we’ll send back real performance and scaling efficiency.

Series & upgrade path

• DRACO (flagship supercomputer tier) — this.
• Scale ladder: from a few hundred accelerators to exascale-class systems (thousands of APUs).
• Architecture options: NVIDIA Grace-Hopper, AMD Instinct MI300X/MI300A — chosen to fit your codes and software stack.
• When to step up: extend the data hall — co-design the fabric, power and facility with an RDP architect.

On-prem vs cloud — the TCO case

At exascale, owning is a sovereign and economic decision: an always-on hyperscale cloud system dominates any national budget over a multi-year horizon, while on-prem keeps sovereign data, codes and capability in-country. RDP pricing is fixed in INR with a GST input-credit-eligible invoice — ask for a multi-year TCO and financing model.

Software & day-one readiness

Ships pre-integrated and validated: AMD ROCm, HIP, RCCL, the InfiniBand stack, Slurm and/or Kubernetes, with PyTorch (ROCm) and vLLM (ROCm) on a supported Linux LTS, plus monitoring and a scheduler. Optional managed operations and an HPC/MLOps platform.

Power, cooling & rack integration

A data hall liquid-cooled exascale system with the highest power density in the range — RDP co-designs facility power, CDU/manifold and water, the InfiniBand spine, floor layout and redundancy. (Exact power, BTU, flow and facility figures confirmed in the design package.) Full out-of-band management.

Deployment, warranty & support

• Made to order, integrated, racked, cabled, cooled and burned-in in India; project timeline confirmed at quote.
• Delivered as one engagement: nodes, leaf/spine fabric, cabling, PDUs, cooling, parallel storage, and the full HPC/AI software stack.
• Onsite warranty + AMC with pan-India coverage, data-hall SLAs and a dedicated escalation path (exact terms 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 a turnkey exascale HPC + AI supercomputer, made to order — talk to an RDP solution architect, co-design the architecture for your codes and models, get a multi-year TCO and financing plan, and run a benchmark before you commit. Request a quote to begin.

Specifications