Equipment supply
TRcompute sources GPU servers and coordinates supply against an agreed specification. Send a bill of materials, or start with the models you run, the capacity you need and your installation site.
We help turn your brief into a configuration for quotation: accelerators, server platform, CPU, memory, storage and network interfaces. If you already have part numbers, we can work from those and discuss alternatives where needed.
The quotation sets out the proposed equipment, quantity, commercial terms and delivery assumptions. Installation, software configuration and ongoing support are scoped separately when requested.
Configuration briefs
These illustrative starting configurations show the level of detail needed for a brief. GPU models, counts, memory and the final system specification are confirmed in your quotation, subject to workload review and availability.
| Planning brief | Example starting configuration | What to review |
|---|---|---|
| Inference server | 1 GPU with 48 GB memory; 16 CPU cores; 128 GB system RAM; 2 × 1.92 TB NVMe drives. | Model size and precision, concurrency, latency target and serving framework. Confirm network, chassis and site power requirements. |
| Training server | 4 GPUs with 80 GB memory each; 64 CPU cores; 512 GB system RAM; 2 × 3.84 TB NVMe drives. | Model, dataset, training method and batch size. Review GPU topology, storage capacity, network and cooling requirements. |
| Multi-node system | 2 nodes, each with 4 GPUs, 80 GB per GPU, 64 CPU cores and 512 GB RAM. Shared storage and network specified separately. | Application scaling, inter-node communication and storage throughput. Agree switches, cabling, rack layout and site capacity. |
Performance depends on the application and software stack. Include any benchmark or acceptance criteria that the proposed system must meet.