GRIFO 32 — Single-root scale-up supernode
← All productsArchitecture specification

Single-root scale-up supernode

GRIFO 32

A 32-accelerator core module. The GRIFO PCIe fabric connects up to four modules to scale to 128 GPUs within a single root complex.

Discuss your workload

Best fit

High-density AI, mixed AI/HPC, memory-bound analytics, sovereign inference and heterogeneous acceleration.

StatusArchitecture in industrialisation and validation

Evidence classArchitecture specification

Verifiable facts

  • Up to 32 full-length PCIe accelerators per core module
  • Up to four GRIFO 32 modules linked by PCIe fabric: 128 GPUs within one root complex
  • Single-root PCIe Gen5 switched fabric
  • Liquid-cooled by design
  • Qualified GPU, FPGA, ASIC and DPU integration paths
  • Scale up first, then scale out
  • Larger domains remain projected until validated

What Billi Dynamics provides

  • Proprietary system architecture and accelerator-domain engineering
  • Billi Dynamics PCIe fabric, management and thermal technologies
  • Accelerator and I/O qualification with system validation
  • Deployment engineering for the target facility

What the customer defines

  • Workload and success criteria
  • Accelerator class and software stack
  • Data, security and sovereignty requirements
  • Power, cooling and data-centre constraints

Documentation

Specifications before slogans.

01

Technical architecture

Explains the single-root PCIe Gen5 fabric, accelerator-domain topology, control plane and validation boundaries behind GRIFO 32.

02

Product brochure

Introduces the product role, intended workloads, system-level design and the evidence currently available for the flagship architecture.

Request a technical briefing