The Rohan Node (Heavy AI Compute)

**The Rohan Node** is the heavy cavalry of the Fin-Hawk system. It is a specialized, high-performance compute node designed to churn through massive language models and complex image/video generation. With a staggering amount of combined VRAM, this node acts as the primary inference engine. Built on a premium desktop chassis, it leverages late generation GPUs to provide robust vector database serving, context servers, and hosts the "God Node" capable of running 70B+ parameter models seamlessly in the background. **Hardware Specifications:** - **GPUs:** NVIDIA RTX 3090 24GB + Tesla P40 24GB (48GB Total VRAM) - **Compute:** AMD Ryzen 7 3800X (8 Cores) - **Memory:** 64GB DDR4 RAM - **Storage:** High-speed NVMe **Software Payload (Aviary OS Compute Tier):** - Ollama & KoboldCPP (Configured for multi-GPU split inference) - ComfyUI (Pre-loaded with Flux, SDXL, and Wan2.1 video pipelines) - Pre-configured vGPU passthrough for Proxmox VMs *Custom built-to-order desktop chassis. Extensive burn-in testing applied. Please allow a 2-3 week lead time for hardware sourcing, assembly, and provisioning.*