Turn a fixed rack of servers into a fluid pool of GPUs, memory, NVMe, and accelerators — claimed by any host on demand. Utilisation goes up. Refresh cycles move to the endpoint. Vendor lock-in evaporates.
A single Datafabrix DF-R4 rack, delivered as one purchase order, installed as one unit, supported as one system.
The specific metrics our customers ask about first — and where this rack is engineered to win.
Merchant silicon is called out explicitly. Datafabrix engineering is called out explicitly. No hand-wave categories.
| Layer | Component | Notes |
|---|---|---|
| Host servers | 6 × host servers with Datafabrix PCIe HBAs | Existing hosts or Datafabrix DF-X Xeon Server Cards |
| Fabric | 2 × Datafabrix PCIe Gen4 Fabric Backplane (DF-B4) | Redundant PCIe switching for HA |
| Composable EP | GPU shelf · NVMe pool · accelerator shelf | Endpoints that any host can claim through the control plane |
| Control plane | Management Server (DF-S13) | Zoning, policy, RBAC, fleet-wide composition |
| JBOF | JBOF Platform (DF-S03) · 24 × NVMe | Disaggregated flash pool visible to all hosts |
| DCIM | All 8 DCIM modules pre-instrumented | Includes composition telemetry: which host owns which endpoint, right now |
| Networking | Standard datacenter networking (customer choice) | Rack does not sit in the data-network path |
| Cooling | Air-cooled · standard rack thermal envelope | Standard datacenter cooling profile |
A server with a $30K GPU that runs at 20% utilisation is $24K of stranded capital. This rack lets any host claim any GPU on demand — utilisation moves from 20% to 60–80% overnight.
Traditional servers force you to refresh compute + I/O + accelerators together. Composable lets you refresh only the endpoints — a new GPU generation slides in, old NVMe stays, hosts unchanged.
One rack replaces a dozen separate departmental dev servers, each half-idle. Every user gets the resources they need, when they need them, and gives them back automatically.
Because the fabric is merchant PCIe and the endpoints are standard PCIe cards, you are never locked into one accelerator vendor. Migrate to next-generation silicon on your timeline, not theirs.
Internal GPU pool shared across data science, ML engineering, and product teams.
Elastic risk-analytics infrastructure — burstable to peak load, contracted at quiet times.
Retire a dozen legacy servers into a single composable rack with better utilisation and lower TCO.
Swap between GPU, DPU, and FPGA endpoints under the same hosts and workloads to benchmark objectively.
Our delivery discipline is the single biggest reason customers pick us over DIY assembly or a locked AI system. Here is what the timeline actually looks like.
One call to align on workload, site, and timeline. Written quote with BOM, power/cooling profile, and delivery date inside one business week.
Rack assembled and validated at our integration facility. Firmware loaded. DCIM modules provisioned. Burn-in tested against the specific workload class.
Rack ships fully assembled. On-site installation by our team or a certified partner. Cabled, powered, network-integrated, DCIM streaming to your operators.
Workload deployed on the rack. Acceptance tests. Handover to your operations team, with our engineering on standby for the first 90 days.
Each Datafabrix rack is a curated combination of solutions. Below are the solutions that ship inside this rack. Click any card for the block diagram, connection topology, and per-subsystem specs.
Tell us your workload, your site, and your timeline. We will respond within one business day.