Edit this Page

PROJECT SPECIFICATION: MTHR8R41n

Attribute Specification
Revision Date: February 5, 2026
Project Status: Fabrication & Component Sourcing (Phase 1)
Classification: Custom Local AI Server & Assistant

1. PROJECT OVERVIEW

Codename: MTHR8R41n ("Motherbrain") Primary Function: Dedicated local AI inference server (LLM & Vision). Design Language: Industrial Mecha / Cyberpunk / "Muffinland" Brand Identity (Sunflower Yellow & Raspberry Red). Form Factor: Open-frame, wall-mounted "Canopy" Wedge.

2. MECHANICAL SPECIFICATIONS

2.1 Chassis Dimensions

  • Orientation: Vertical (Portrait) Wall Mount.
  • Height: 600 mm
  • Width: 460 mm
  • Depth (Top): 280 mm (Creates "Canopy" overhang for cooling/dust protection).
  • Depth (Bottom/Sides): 150 mm (Tapers down from top).
  • Front Interface: Open frame with tubular steel "Crash Bars" (Top/Bottom 1/3).

2.2 Component Layout

  • Top Zone (0-100mm): Exhaust plenum, DIN Rail for logic/distribution, Master Switches.
  • Middle Zone (100-450mm): Motherboard (Centered). "Engine Scoop" cutout for CPU Cooler protrusion.
  • Bottom Zone (450-600mm): Power Deck. PSU mounted sideways (Flush with 150mm wall depth).

3. HARDWARE MANIFEST

3.1 Phase 1: "The Awakening" (Current Config)

  • CPU: Intel Core i7-6850K (Broadwell-E, 6-Core, 40 PCIe Lanes).
  • Motherboard: X99 Platform (LGA 2011-v3).
  • Cooling: Cooler Master V8 GTS (Air).
    • Note: Protrudes ~25-30mm past chassis walls. Native LGA 2011 mount.
  • Memory: 32GB DDR4-3200 (4x 8GB HyperX Fury Kit).
    • Status: Pending Auction.
    • Config: Quad Channel. Target speed 2666MHz for stability.
  • GPU: AMD Radeon RX 580 (8GB VRAM).
  • Power Supply: 850W Unit (New).

3.2 Phase 2: "The Evolution" (Future Target)

  • Target Capability: ~96GB VRAM (Llama-3-70B / DeepSeek-V3).
  • GPU Array: 3x - 4x NVIDIA RTX 3090 (24GB each).
  • Power Distribution: Dual PSU.
    • Primary: 850W (System + GPU 1).
    • Secondary: EVGA SuperNOVA 1200W (GPUs 2, 3, 4).
  • Thermal Management: Custom Water Loop.
    • Radiator: 360mm or 420mm mounted in "Ceiling" (280mm depth zone).
    • Manifold: Parallel connection block.

4. SOFTWARE ARCHITECTURE

  • Operating System: Ubuntu 24.04 LTS.
  • Deployment: Docker Compose.
  • AI Backend: Ollama.
    • Driver Override: HSA_OVERRIDE_GFX_VERSION=8.0.3 (Force ROCm on Polaris).
  • User Interface: Open WebUI.
  • Network: LAN Exposed (Accessible by WORKBEAST/WORKDECK).

5. FABRICATION NOTES & WARNINGS

5.1 Critical Assembly Order

WARNING: The Cooler Master V8 GTS is 154mm wide.

  1. Install RAM First: All 4 DIMM slots must be populated before cooler installation.
  2. Mount Cooler: Once bolted, RAM slots are inaccessible without full disassembly.
  3. Fan Header: Connect CPU_FAN cable before bolting down the heatsink unit.

5.2 Structural Requirements

  • Backboard Material: 10mm Plywood or Aluminum Composite (ACM).
  • Motherboard Standoffs: Minimum 20mm elevation for cable routing behind board.
  • Wall Anchors: Must support estimated loaded weight of ~15kg (Phase 1) to ~25kg (Phase 2).

6. ACTION ITEMS

  • [ ] Procurement: Monitor and win auction for HyperX Fury RAM.
  • [ ] Fabrication: Cut Trapezoidal Side Walls (280mm -> 150mm taper).
  • [ ] Fabrication: Cut Backboard (600mm x 460mm).
  • [ ] Assembly: Test fit V8 GTS on motherboard (dry) to verify "Engine Scoop" depth.
  • [ ] Software: Initial Docker stack deployment on temporary RX 580 setup.