System component summary for origami (v1.5 platform). (GPU specifications are detailed in the Physical to Logical Mapping section below). Reference: PCPartPicker Saved Build (v1.5).
| Component | Hardware Specification | System Allocation & Details |
|---|---|---|
| CPU | Intel Core i9-7940X (14 Cores / 28 Threads @ 3.10 GHz Base / 4.30 GHz Turbo) | LGA 2066 High-End Desktop (HEDT) Processor |
| Platform / PCIe Lanes | X299 Server Platform (44 PCIe CPU Lanes) | High-bandwidth PCIe lane allocation for multi-GPU compute |
| RAM / Memory | 128GB DDR4 Quad-Channel Memory (8x 16GB DDR4-3200) | High-speed system memory for concurrent F@H + LLM models |
| Power Supply (PSU) | Super Flower LEADEX VII XG 1300W 80+ Gold Certified | Dedicated 1300W continuous power delivery for multi-GPU workloads |
| Chassis / Case | High-Airflow Full-Tower Server Chassis | Optimized thermal dissipation for dual F@H + LLM compute |
| Cooling & Fans | Multi-Fan PWM High-Static Pressure Thermal Control (Automated dynamic thermal regulation via gpufanpin) |
Customized 3D-printed fan redirect ducts on the MEGACOOL side intake fans for targeted GPU airflow |
Mapping physical motherboard PCIe slots to PCI Bus Addresses, NVIDIA software indexes, GPU UUIDs, and active F@H / LLM compute roles on host origami.
| Motherboard Slot | GPU Model | PCI Bus ID | NVIDIA Index | GPU UUID | Target Workload / Assignment | Notes |
|---|---|---|---|---|---|---|
| PCIE1 (x16) | Asus PRIME SFF-Ready GeForce RTX 5070 (12GB GDDR7) | 0000:65:00.0 |
1 |
dd40d876.. |
F@H Folding Engine | Active (180W Cap) |
| PCIE3 (x16) | PNY NVIDIA RTX PRO 4000 Blackwell Workstation (24GB GDDR7) | 0000:17:00.0 |
0 |
6f99ccc4.. |
F@H Folding + Local LLM Inference | Active (145W Cap) |
| PCIE5 (x16) | Zotac GAMING SOLID OC GeForce RTX 5070 (12GB GDDR7) | 0000:66:00.0 |
2 |
18de45ee.. |
F@H Folding Engine | Active (180W Cap) |
System evolution plan for node origami. Reference: PCPartPicker origami v2 Target Build.
gpufanpin and preemption safety via fah-runaway.